Looking Back: Results of a Retrospective Survey of Delawareans Diagnosed with Cancer June 2007 This report is made possible with funding from the Delaware Health Fund, and with strategic leadership and guidance provided by the Delaware Cancer Consortium. Table of Contents EXECUTIVE SUMMARY................................................................................................1 1.INTRODUCTION...................................................................................................8 1.1. Cancer in Delaware..........................................................................................8 1.2. Study Objectives............................................................................................9 1.3. Study Design................................................................................................9 1.3.1. Delaware Cancer R.egistry.................................................................................10 1.3.2. Delaware Institutional Review Board.......................................................................10 2. METHODOLOGY...................................................................................................11 2.1. Identification of Study Population..........................................................................11 2.2. Physician Consent...........................................................................................12 2.3. Patient Recruitment.........................................................................................13 2.3.1. Selection of Patients for Recruitment.....................................................................13 2.3.2. Recruitment Procedures....................................................................................14 2.4. Data Collection.............................................................................................14 2.4.1. Survey Instrument.........................................................................................14 2.4.2. Fielding of Interviews....................................................................................15 2.4.3. Status of Study Participants..............................................................................15 2.5. Analytic Methods............................................................................................17 2.5.1. Weighting.................................................................................................17 2.5.2. Prevalence Estimates......................................................................................18 2.5.3. Comparison of Cancer Population With BRFSS Respondents....................................................18 2.5.4. Predictors of Early Versus Late Stage at Diagnosis........................................................19 3. RESULTS.......................................................................................................20 3.1. General Attributes..........................................................................................20 3.1.2. Socioeconomic Status......................................................................................22 3.1.3. Length of Residency.......................................................................................25 3.1.4. Occupational History......................................................................................26 3.1.5. Health Status.............................................................................................32 3.1.6. Diet......................................................................................................36 3.1.7. Alcohol Consumption.......................................................................................38 3.1.8. Tobacco Indicators........................................................................................40 3.1.9. Physical Activity.........................................................................................43 3.1.10. Health Care Access.......................................................................................45 3.1.11. Comparison of Risk Factors With the General Delaware Population..........................................48 3.2. Cancer-Specific Attributes..................................................................................52 3.2.1. Breast Cancer.............................................................................................52 3.2.2. Colorectal Cancer.........................................................................................57 3.2.3. Prostate Cancer...........................................................................................60 3.2.4. Lung Cancer...............................................................................................63 3.2.5. Diagnosis and Treatment...................................................................................65 3.2.5.1. Cancer Diagnosis........................................................................................65 3.2.5.2. Cancer Treatment........................................................................................67 4. DISCUSSION AND SUMMARY OF FINDINGS............................................................................71 4.1. Limitations.................................................................................................71 4.2. General Attributes..........................................................................................72 4.3. Cancer-Specific Attributes..................................................................................74 4.4. Future Directions...........................................................................................76 References.......................................................................................................77 APPENDICES Appendix 1: Physician Consent Letter Appendix 2: Informed Consent Packet Appendix 3: Survey Questionnaire with Reported Frequencies and Percentages Appendix 4: Interviewer Training Manual LIST OF TABLES Cancer Incidence and Mortality Rates for Delaware Compared With the United States Comparison of Weighting Factors in the Eligible Population of Delaware Residents With Cancer and the Interviewed Population Frequency and Percentage of Interviewed Study Participants for Each Cancer Site by SEER Stage at Diagnosis Demographic Variables by County Demographic Variables by Cancer Type Socioeconomic Variables by County Socioeconomic Variables by Cancer Type Length of Residency by County Occupation by County Occupation by Cancer High-Risk Industry by County High-Risk Industry by Cancer Type Health Status by Cancer Type Health Conditions by Cancer Type Dietary Variables by County Dietary Variables by Cancer Type Alcohol Consumption by County Alcohol Consumption by Cancer Type Tobacco Indicators by County Tobacco Indicators by Cancer Type Physical Activity Levels by County Physical Activity Levels by Cancer Type Health Care Access and Utilization by County Health Care Access and Utilization by Cancer Type Risk Factor Comparisons With the General Delaware Population Breast Cancer Risk Factors Breast Cancer Screening Test Comparisons With the General Delaware Population Colorectal Cancer Risk Factors Colorectal Cancer Screening Tests Colorectal Cancer Screening Test Comparisons With the General Delaware Population Prostate Cancer Risk Factors Prostate Cancer Screening Tests Prostate Cancer Screening Test Comparisons With the General Delaware Population Lung Cancer Risk Factors Occupational Lung Cancer Risk Factors Cancer Diagnosis Surgery Radiation Therapy Chemotherapy Hormone Therapy Time from Diagnosis to Treatment [For a complete, print copy of these report--including all the charts and graphs--send a written request to the Division of Public Health, Health Promotion and Disease Prevention Section, Suite 9, Thomas Collins Building, 540 S. DuPont Highway, Dover, DE 19901.] EXECUTIVE SUMMARY The Retrospective Survey of Delawareans Diagnosed with Cancer was designed and conducted in response to the Delaware Cancer Consortium’s recommendation to gain a better understanding of the factors related to incidence and mortality for the most common cancers in Delaware. Together, these four cancers accounted for 57 percent of all cancers diagnosed in Delaware and 52 percent of all deaths due to cancer during 1999–2003. Cancer Incidence in Delaware 1999-2003 Female breast 14% Colorectal 11% Lung and bronchus 16% Prostate 16% Other cancers 43% Cancer Mortality in Delaware 1999-2003 Female breast 7% Colorectal 10% Lung and bronchus 30% Prostate 5% Other cancers 48% Source: Delaware Cancer Registry, Delaware’s Division of Public Health From November 2005 to April 2006, 892 survivors of these four cancers participated in a 30–45- minute telephone interview and provided answers to questions about their health, lifestyles, cancer risk factors, screening practices and diagnosis and treatment outcomes. This report summarizes the information gathered from participants for each of the four cancers. The key findings are presented here. Health Status • The majority of study participants (92 percent) reported their general health status prior to diagnosis as good to excellent. Lung cancer participants were less likely to report good to excellent health compared to participants with breast, colorectal or prostate cancer. • The prevalence of chronic health conditions for study participants was similar to the prevalence of these health conditions in the general Delaware population. Six chronic health conditions were assessed: heart disease, diabetes, hypertension, high blood cholesterol, asthma and arthritis. Lung cancer respondents had higher rates of asthma, and breast cancer respondents had lower rates of asthma compared with the general Delaware population. A higher proportion of lung cancer participants reported having three or more health conditions than participants with breast, colorectal or prostate cancer. • Body mass index (BMI) was calculated from height and weight prior to diagnosis, as reported by study respondents. Respondents with all types of cancer had a BMI similar to that of individuals in the general Delaware population. About 40 percent of study respondents were overweight (BMI= 25 to < 30) and 24 percent were obese (BMI=30+). Occupation • Forty-two percent of study respondents reported ever working in one or more of five high- risk industries: 21 percent worked in the chemical industry, 3 percent worked in the pharmaceutical industry, 26 percent worked in manufacturing, 10 percent worked in agriculture and 17 percent worked in the construction industry. • About 60 percent of lung cancer respondents reported exposure to potential carcinogens in the workplace. Diet • Only 21 percent of study respondents reported eating five or more servings of fruits and vegetables per day at time of diagnosis. Another 40 percent ate at least three but fewer than five servings of fruits and vegetables per day. Fruit and vegetable consumption among study participants at time of diagnosis was similar to fruit and vegetable consumption reported by the general Delaware population. Breast cancer respondents were most likely, and lung cancer respondents were least likely, to meet the current recommendation. • About 23 percent of study respondents reported eating five or more daily servings of foods high in fat at time of diagnosis, and another 40 percent reported eating at least three but fewer than five daily servings of foods high in fat. Alcohol Consumption • About 60 percent of the study population reported having at least one drink in the 30 days prior to diagnosis. Eight percent were classified as heavy drinkers and 13 percent were classified as binge drinkers. • Breast cancer respondents were more likely to have reported having a drink in the 30 days prior to diagnosis and to binge drink than individuals in the general Delaware population. • Prostate cancer participants were twice as likely to report binge drinking, heavy drinking and having a drink in the 30 days prior to diagnosis when compared with individuals in the general Delaware population • Lung cancer respondents were two and a half times more likely to report heavy drinking, and three times more likely to report binge drinking, compared with members of the general Delaware population. Tobacco Use • Ninety-four percent of lung cancer patients either smoked at the time of diagnosis or were former smokers. Lung cancer patients were 27 times more likely to be current smokers and seven times more likely to be former smokers compared to the general Delaware population. • The association between current or former smoking status and the risk for breast, colorectal or prostate cancer was not significant. • About 67 percent of study participants who had ever smoked began before age 18. Physical Activity • Seventy-seven percent of the study respondents reported doing some leisure-time physical activity or exercise. About half of study participants met the requirement for moderate activity level, and 26 percent met the requirement for vigorous activity level. Health Care Access and Utilization • Overall, 98 percent of survey respondents had health insurance coverage at diagnosis. Lung cancer participants were less likely to have health insurance compared with participants with breast, colorectal or prostate cancer. • About 97 percent of the study population reported that their usual source of health care was in a doctor’s office or public health/community health clinic. Ninety percent of respondents reported seeing a doctor at least once a year. • A higher percentage of colorectal cancer respondents saw a doctor less than once a year (18 percent), compared with those with prostate cancer (12 percent), lung cancer (10 percent) or breast cancer (five percent). Screening • Use of screening tests was high among respondents with breast cancer. About 83 percent reported having received both a mammogram and a clinical breast exam in the two years prior to diagnosis. • A majority of women with breast cancer (59 percent) reported receiving their first mammogram before age 50 and then receiving one every year. • Annual clinical breast exams were reported by 85 percent of respondents. Monthly self- breast exams were reported by 56 percent of participants with breast cancer. • Among study respondents aged 50 and older with colorectal cancer, 18 percent had a fecal occult blood test in the two years prior to diagnosis and 55 percent had a sigmoidoscopy. • About 55 percent of men with prostate cancer aged 40 and older received a prostate-specific antigen test in the two years prior to diagnosis and 58 percent received a digital rectal exam. • Compared with the general population, study respondents were equally or more likely to have received mammograms, sigmoidoscopy and prostate cancer screening tests, but were 16 percent less likely to receive a clinical breast exam, and two and a half times less likely to report a recent fecal occult blood test. Diagnosis and Treatment • Overall, 34 percent of study respondents had their cancer detected due to experiencing a symptom and 51 percent of cancers were detected through screening tests. Respondents with breast and prostate cancer were more likely to have their cancer detected by a screening test compared to respondents with colorectal cancer. • The time from when cancer was detected by the patient or a health care provider until diagnosis differed by cancer type. Cancer detection refers to the first event that suggested that the respondent might have had cancer. This event could be a positive screening test, a clinical finding or the respondent experiencing a symptom. Diagnosis is the point when cancer is pathologically or clinically confirmed. Overall, about 60 percent of the respondents had less than one month between detecting their cancer and receiving a diagnosis. • The majority of breast, colorectal and lung cancer patients received their first treatment for cancer within one month of diagnosis. The majority of men with prostate cancer began treatment within three months of diagnosis. 1. INTRODUCTION The Delaware Cancer Consortium was originally formed as the Delaware Advisory Council on Cancer Incidence and Mortality in March 2001 in response to Senate Joint Resolution 2 signed by Governor Ruth Ann Minner. The advisory council, consisting of 15 members appointed by the governor, was established to advise the governor and legislature on the causes of cancer incidence and mortality and potential methods for reducing both. The advisory council was later expanded and its name changed to the Delaware Cancer Consortium (DCC) in SB102.1 The DCC began meeting in April 2001 with the shared understanding that their work would be focused on developing a clear and usable cancer control plan for action. In their report “Turning Commitment into Action,” (April 2002) the council issued the following recommendation with regard to increasing knowledge about cancer in the State of Delaware: “Conduct a retrospective survey of individuals with cancer or family members of patients to collect information on family history, occupation, lifestyle, diet, exercise, migration, etc. Include only those cancers for which the state is elevated in incidence or mortality. Obtain data necessary to determine which environmental factors might contribute to Delaware’s heightened cancer rates.” The Retrospective Survey of Delawareans Diagnosed with Cancer was designed and conducted in response to this recommendation. Recruitment for the study began in September 2005; data collection was completed in April 2006. 1.1. Cancer in Delaware During 1999–2003, Delaware’s five-year annual age-adjusted cancer incidence rate was 503.5 per 100,000, which is higher than the estimated U.S. rate of 478.1 per 100,000.2 Female breast cancer and prostate cancer were the most commonly diagnosed cancers, followed by lung and bronchus cancer and colorectal cancer.3 These four cancer types accounted for 57 percent of cancers diagnosed in Delaware during 1999–2003. An estimated 4,190 new cancer cases will be diagnosed in Delaware during 2006.4 Delaware’s five-year annual age-adjusted mortality rate was 206.9 per 100,000 during 1999– 2003, compared with the U.S. rate of 195.7 per 100,000.3,5 The death rate for lung and bronchus cancer was highest, followed by prostate, female breast and colorectal cancers. These four cancers accounted for 52 percent of cancer deaths in Delaware during 1999–2003. An estimated 1,690 Delaware residents will die from cancer during 2006.4 1.2. Study Objectives Incidence and mortality rates for lung and bronchus cancer and colorectal cancer are elevated in Delaware compared with the United States. Incidence and mortality rates for prostate cancer and female breast cancer in Delaware are similar compared with the United States. This study was designed to investigate the underlying factors that may effect positive change in incidence and mortality rates for these cancers. Specific study objectives are as follows: • Determine the prevalence of known risk factors for lung and bronchus, colorectal, female breast and prostate cancers, including environmental and occupational factors, in patients diagnosed with these cancers in Delaware. • Examine factors that may be related to elevated mortality rates for all cancers, such as stage at diagnosis, race, co-morbid conditions, health care access, screening behaviors and diagnosis/treatment patterns. • Determine the prevalence of behavioral risk factors, screening use and health care in cancer patients diagnosed with each of the four major cancers; compare these with prevalence rates in the overall Delaware population using data from the Centers for Disease Control and Prevention’s (CDC) Behavioral Risk Factor Surveillance System (BRFSS). • Examine geographic differences in cancer rates and risk factors (county level) and migration patterns of cancer patients. 1.3. Study Design This is a retrospective study of cancer survivors in Delaware to gather risk factor data prior to their cancer diagnosis and data on diagnosis, treatment and screening behaviors for each type of cancer. The target population for the study is all individuals aged 18 or older diagnosed with colorectal, prostate, female breast, and lung and bronchus cancers in Delaware during the five- year period, Jan. 1, 1999, to Dec. 31, 2003, and registered in the Delaware Cancer Registry (DCR). For the risk factor analysis, a case/control design was used, comparing the Delaware cancer survivors with the Delaware BRFSS 2002 sample population. The BRFSS is a state-based system of health surveys which generates information about health risk behaviors, clinical preventive practices and health care access and use, primarily related to chronic diseases and injury. The Delaware BRFSS survey sample was selected to represent the general population of Delaware. 1.3.1. Delaware Cancer Registry The Delaware Cancer Registry (DCR) is the state’s central cancer information center and is part of Delaware’s Division of Public Health. Delaware is one of 45 states supported by CDC’s National Program of Cancer Registries (NPCR). Delaware’s central cancer registry is population-based, and collects data on all cancer patients who are state residents at the time of diagnosis. The DCR collects information on newly diagnosed cancer cases, cancer treatment, cancer deaths, and follow-up data. As stated in the Delaware Cancer Control Act of 1980, the purpose of the registry is to ensure an accurate and continuing source of cancer data and specified tumors of a benign nature. Additionally, the confidentiality of patient information in the registry is required by this law. Quality assurance is performed by a Certified Tumor Registrar (CTR) to ensure that the registry includes complete and accurate data that conform to standards established by the National Program of Cancer Registries (NPCR) and the North American Association of Central Cancer Registries (NAACCR). 1.3.2. Delaware Institutional Review Board Before beginning work on this study, a research protocol was developed and submitted to the Division of Public Health in the Delaware Department of Health and Social Services (DHSS) for review. The study was reviewed and approved by the DHSS institutional review board (IRB) in April 2005 and by the ORC Macro IRB in May 2005. As part of the review process, the Delaware IRB required that the physicians of all eligible study participants be contacted and given the option to identify any patients who they thought should be removed from the study. 2. METHODOLOGY 2.1. Identification of Study Population For this study, a complete NAACCR version 10 dataset of eligible cancer patients was provided by the DCR according to the following inclusion and exclusion criteria: • Delaware residents at time of diagnosis, • Aged 18 and older at time of diagnosis, • Diagnosed between Jan. 1, 1999, and Dec. 31, 2003, • Diagnosed with: .. Female breast (ICD-0-3: 500–509), .. Prostate (ICD-0-3: 619), .. Colorectal (ICD-0-3: 180–189, 199, 209, 260), .. Lung and bronchus (ICD-0-3: 340–349), • Malignant tumors only (Behavior code = 3), • Alive, and • Excluded patients who expressly told the American Cancer Society (ACS) Quality of Life study they did not want to be contacted for research purposes or were cognitively impaired (seven individuals). For the remainder of this report, the term “breast cancer” will be used for female breast cancer and the term “lung cancer” will be used for lung and bronchus cancers. There were 7,006 individuals eligible to be interviewed for the study. Since the NAACCR database included identification numbers for doctors only, the DCR provided a database with addresses and contact information for each of the physician identification numbers. Using the databases provided by the DCR, a separate database with patient tracking information was created and merged with the physician database to obtain name and contact information for each patient’s physician. Each patient record contained physician IDs for up to five physicians, as defined for NAACCR data submissions, but not all fields had valid entries. The first physician to be contacted was selected according to the following hierarchy for patients with a valid physician ID code in any of the five fields: 1) NAACCR Item #2470—follow-up physician (physician currently responsible for the patient’s medical care), 2) NAACCR Item #2460—managing physician (physician responsible for the overall management of the patient during the diagnosis and/or treatment of this cancer), 3) NAACCR Item #2480—primary surgeon (physician who performed the definitive surgical procedure), 4) NAACR Item #2490—MD3 (another physician involved in the care of the patient), and 5) NAACR Item #2500—MD4 (another physician involved in the care of the patient). Using the physician ID number for matching, separate datasets were created for individuals with physician codes that were matched to the physician database (6,289) and for individuals with physician codes that were not matched (713). A third dataset was created with physician names and contact information for the 663 physicians that were matched to the study cancer patients. 2.2. Physician Consent Before contacting any patients for the study, a letter was mailed to their physicians requesting they identify any patients who should not be contacted due to reasons such as they were deceased, terminally ill, not cognitively or emotionally capable of participating in the study or not informed that they had cancer. Prior to the initial mailing, physicians with high patient loads (responsible for 20 or more patients) were identified. During June 8–10, 2005, these physicians were contacted by telephone to prepare them for the upcoming mailing and request their support in identifying patients who should not be contacted. Any physicians who expressed concern when the ACS Quality of Life study was conducted were also contacted at this time. A personalized letter and list of patients for each physician was prepared and mailed on July 5 (see appendix 1). The following were excluded from the initial mailing: physician codes that linked to a cancer center or registry (30 codes) and physicians with a note in the cancer registry that they were retired or deceased or had moved. Physicians were asked to identify patients who should not be contacted for the study and were asked to fax their patient lists to the study center by the end of July. As the fax responses were received, the physician contact database was updated to record receipt of the fax and the date of receipt. Approximately two weeks following the mailing, all physicians for whom no fax response had been received were called to confirm they had received the letter and to remind them to identify any study patients who should not be contacted. A computer-assisted telephone interview (CATI) script was developed and programmed for the calls. All calls were conducted by trained interviewers at the ORC Macro CATI call center in Burlington, VT. As a result of these calls, approximately 240 physicians requested the letter and patient list be resent to them via fax. These faxes were prepared and sent from ORC Macro’s Bethesda, MD, office. Responses were received from 242 physicians, which included 33 of the 67 physicians with high patient loads. Of these 33 physicians, 11 refused to give permission to contact their patients. The remaining 34 physicians with high patient loads who did not respond were contacted to confirm they had received the letter and patient list. The objective of the physician consent process was to provide each patient’s physician the opportunity to inform us if they did not want one or more of their patients to participate in the study. It was clearly stated in the letter that if we did not hear from them, we would presume that their patients could be contacted. We made a concerted effort to confirm that physicians did receive our letter and to remind them to notify us of any patients who should not be contacted for the study. Non-response from a physician who we were able to contact and verify receipt of our letter was considered a passive consent. We did not attempt to recruit any patients whose physician could not be contacted. 2.3. Patient Recruitment 2.3.1. Selection of Patients for Recruitment In order to determine the set of patients eligible for recruitment, the physician responses and results from the CATI follow-up calls to physicians were merged with the patient dataset. A new dataset was created that included the physician response and CATI follow-up information, along with stratification variables and contact information for each patient. A patient’s eligibility status for recruitment was determined according to the following criteria: Patients eligible for recruitment: • Physician fax received, checked yes, • Physician fax received, nothing checked and no other comments, • CATI follow-up with physician confirmed letter received, • Physician contacted and faxed a second letter, and • Patients who consented from the prior study. Patients ineligible for recruitment: • Physician fax received, checked no, and • Physician fax received, patient deceased. Patients possibly eligible for recruitment: • Physician fax received, note that the patient is not theirs or not followed, • Physicians who were not sent the mailing (retired or deceased), • Physicians with invalid addresses (returned to sender), • Physicians without phone numbers who were not contacted, • Physicians with phone numbers who could not be contacted, and • Patients with a physician code of the cancer center or registry. The recruitment status of the original 7,006 patients eligible to participate in the study was determined as follows: • 4,410 patients eligible to recruit (physician contacted), • 1,106 patients ineligible to recruit (physician contacted, refusal or patient deceased), and • 1,490 patients possibly eligible to recruit (unable to contact or locate physician). All breast (1,307), colorectal (791) and lung (318) cancer patients eligible to be recruited were selected for initial recruitment, along with a subset of prostate cancer patients (1,336). Because we were expecting a 20 percent response rate and there were 2,075 prostate cancer patients eligible for recruitment, we assumed we would receive a sufficient number of positive responses if we selected a smaller subset of the eligible participants. Therefore, 739 (35.6 percent) Caucasian patients with local (stage 1) prostate cancer were not selected for recruitment. 2.3.2. Recruitment Procedures All individuals selected for recruitment were initially sent an informed consent packet via U.S. mail between September and December 2005. The informed consent packet contained the following materials: • A letter from DHSS requesting participation in the survey. • A two-page survey information sheet describing the background and purpose of the study, eligibility, study plan, benefits and risks of participation, confidentiality, rights for withdrawal and contact information for DHSS and the study team. • Two copies of the consent form, one for the patient’s records and one to be signed and submitted directly to the study team. • A stamped return envelope with the address of the study team. The materials provided in the informed consent packet are included in appendix 2. Packets were mailed to a total of 3,687 potential participants. In addition, follow-up telephone calls were placed to all lung and some colorectal cancer patients who did not initially return their consent forms. This subset of patients received follow-up calls as we had fewer potential respondents with these cancer types, and wanted to have as many participate as possible. A total of 1,261 consent forms were returned, a response rate of 34 percent. Of these, 988 agreed to participate. Information from the consent forms was entered into the patient tracking database as the forms were received. Patient responses were classified into four groups: refused, unable, deceased and agreed to participate. Eleven individuals were removed from the study because they informed us they did not have cancer (seven individuals), were diagnosed after 2003 (two individuals), or were ineligible because they were not diagnosed in Delaware (two individuals). Three hundred thirty-eight individuals either declined or were unable to participate. Of these, 64 were unable to participate (36 from consent forms and 28 from telephone follow-up), and 274 refused to participate (204 from consent forms and 70 from telephone follow-up). The most common reasons given for nonparticipation were: too sick, too old, too tired, hard of hearing, not interested, no time, too busy, traveling or away. The Social Security Death Index (SSDI) was used to further research lung cancer patients whose consent forms were not returned and all patients whose consent forms were returned due to invalid addresses. Deceased individuals, as indicated by physicians, relatives or through SSDI research were reported to the DCR for further confirmation and updating the DCR database. Since the sample size from recruitment of the eligible participants was sufficient, we did not attempt to further research and/or recruit any individuals who were possibly eligible for recruitment. 2.4. Data Collection 2.4.1. Survey Instrument The survey instrument included questions regarding behavioral risk factors and diagnosis, treatment and screening practices specific to each type of cancer. The questionnaire was designed to take about 30–45 minutes to complete and consisted of a core set of questions to be asked for all cancer patients and separate modules with questions specific to each type of cancer. To ensure compatibility for comparison with BRFSS data, survey questions designed to collect information regarding demographics, behavioral risk factors, screening behaviors and health care access were worded the same as on the BRFSS questionnaire. A copy of the survey instrument is included in appendix 3. The survey instrument was programmed using Computers for Marketing Corporation software that allowed automatic control of skip and fill logic and performed range checking and other editing as the interviews were conducted. In early October 2005, the survey instrument was pilot tested and timed with eight of the initially recruited participants, two for each type of cancer. The questionnaire was further refined based on feedback from the pilot interviews and operationally tested prior to fielding. 2.4.2. Fielding of Interviews Interviews were conducted via CATI at ORC Macro’s Vermont call center from November 2005 through April 2006. Due to the nature of the majority of the survey questions, which required respondents to recall behaviors and screening practices prior to their cancer diagnosis, we did not allow proxy interviews with family members or relatives. All interviews were conducted directly with study participants in order to obtain the most reliable survey data. Interviewers with prior training in sensitive-subject surveys such as the BRFSS were selected to conduct the interviews for this study. Female interviewers were used to conduct surveys of female participants, and male interviewers conducted the surveys of male participants. A full day of intensive training was conducted specifically for this study to ensure familiarity with the survey instrument and particular areas of the instrument that needed to be reviewed or emphasized. An interviewer training manual (provided in appendix 4) was prepared for the full-day interviewer training session, which was held on Oct. 18, 2005. 2.4.3. Status of Study Participants Interviews were completed with 892 of the 988 individuals who returned their consent forms and agreed to participate. Eight of these individuals were diagnosed with two different cancers. Twenty-five of the 988 individuals who agreed to participate voluntarily dropped out of the study. Fifteen refused to be interviewed when they were called, two started but did not complete the interview and eight were unable to be interviewed when called. Sixty-one individuals who consented to participate were dropped from the study because they could not be reached after 15 or more attempts, and two individuals were dropped due to death during the interview period. 2.5. Analytic Methods All analyses for this study were conducted using Statistical Analysis Software (SAS) version 9.1. Two analytic datasets were created for the analyses. The patient-level dataset contained one record for each of the 892 completed interviews and included all variables collected in the first nine sections of the interview for all cancer types. The cancer-level dataset contained a total of 900 records, one record for each cancer type, and included the variables collected in the cancer modules and the diagnosis and treatment section of the interview. 2.5.1. Weighting The main objective of this study was to measure the prevalence of cancer risk factors among Delawareans with the four most common cancers. Ideally, the study population should reflect all individuals diagnosed with these four types of cancer in the Delaware population. However, due to the logistics of conducting the study and the consent process, the study was conducted only among cancer survivors. We interviewed 892 of the 7,006 individuals known to be alive at the beginning of the study period. Because data were collected from anyone who was willing to participate in the study, the individuals who participated were not a random sample of all individuals eligible to participate. Therefore, the distributions of basic demographic variables, such as age, race and gender, were not necessarily the same for study participants compared with the entire eligible population. In order to account for differences in these basic variables, we adjusted the results to reflect the entire eligible population using a method called post-stratification weighting. The potential confounders shown in the table below differ in the interviewed population compared with the eligible population. The variables were used to weight the study responses to the distribution of these variables in the total eligible population. Although race was reported during the interview, we used the race coded by the DCR for weighting purposes in order to be consistent for all eligible participants. There were few differences between the two sources, and these occurred mainly for individuals who reported more than one race in the interview. 2.5.2. Prevalence Estimates For each of the interview questions, a raw prevalence and a weighted prevalence were computed. The responses to open-ended questions and “Other” responses were reviewed and categorized where possible. For all cancer sites combined, the weighted prevalence estimates and their 95 percent confidence intervals (CI) are presented for the entire Delaware population and by county at diagnosis. For the data collected in the cancer-specific modules of the interviews, weighted prevalence for each of the variables are calculated, and data stratified by county are included. A chi-square test was conducted for each variable to identify significant differences among counties. These analyses were conducted for the following variables: • Demographic variables, including age at diagnosis, race, Hispanic ethnicity, gender, marital status, education, income, occupational status and length of residency in the county, Delaware and the United States. Since age, race and county at diagnosis were included in the interview questions and also included as part of the cancer registry data, these variables were compared to verify and assess level of agreement. Where differences occurred, the variable as coded in the cancer registry was used for all analyses. • Risk factors for cancer, including diet, alcohol consumption, tobacco use, physical activity, environmental exposures and the personal, family and occupational history questions from the cancer-specific modules. • Factors affecting cancer prognosis, including co-morbid conditions, health care access and timely and appropriate diagnosis and treatment questions from the cancer-specific modules. In addition to the analysis by county type, an analysis of each variable by cancer type was conducted for variables answered by all respondents, as part of the core section of the interview. The weighted prevalence and 95 percent confidence interval for each variable by cancer type is presented along with the results of the chi-square test to identify differences. Analyses of racial differences were not conducted due to insufficient numbers of patients in racial/ethnic groups other than Caucasians. 2.5.3. Comparison of Cancer Population With BRFSS Respondents The survey questions were designed whenever possible to follow the same format as questions asked in the BRFSS. This allows for comparison of the prevalence of risk factors in the study population of cancer patients with responses provided by a random sample of Delaware residents. From the BRFSS survey questions, a series of binary behavioral risk factors have been developed. Individuals are categorized as either being at risk or not at risk for each behavior based on definitions developed by CDC. Wherever possible, these risk factor definitions were used for this analysis. For each risk factor, a multivariate logistic regression controlling for age at diagnosis, race, gender and county at diagnosis was used to compare the two populations as specified in the following model: Log odds of cancer = a + ß1(risk factor) + ß2(age at diagnosis) + ß3(gender) + ß4(race) + ß5(county) The odds ratio for each risk factor and the 95 percent confidence interval are presented for each comparison. An odds ratio with a value of 1.00 for any risk factor indicates that the odds of the outcome, in this case, cancer, is exactly the same for individuals with or without the risk factor. An odds ratio is considered significant if the value of 1.00 is not contained in the confidence interval. In addition, age-adjusted, weighted prevalence rates and 95 percent confidence intervals were calculated for each level of the binary risk factors. Because the study population of cancer patients is much older than the general Delaware population, the potential confounding effect of age is reduced when comparing age-adjusted rates computed using the same standard million population. An age-adjusted rate is a weighted average of the age-specific (crude) rates, where the weights are the proportions of persons in the corresponding age groups of a standard million population. The 2000 U.S. Census standard million population was used for the calculation of the age-adjusted rates in this analysis.6 Confidence intervals were computed using the method developed by Fay and Feuer.7 2.5.4. Predictors of Early Versus Late Stage at Diagnosis We proposed to examine the association between screening test use, health care access and utilization, insurance status, socioeconomic status and race on diagnosis of cancer for later versus early-stage cancers. However, there were insufficient numbers of respondents with late-stage disease to conduct this analysis. The table below displays the distribution of interviewed cases by stage. The criterion for sufficient numbers was based on the analysis of Peduzzi et al., who recommended there be at least 10 people with the outcome of interest (in this case, late-stage disease) for every variable entered into the logistic regression model.8 As the data indicate, there were not enough cases to conduct the multivariate logistic regression modeling that was originally proposed. 3. RESULTS This section of the report describes the results from the study and is divided into two main sections. In the first section, results are presented for all variables asked in the core section of the interview, which were not specific to a particular type of cancer. These included demographics, length of residency, behavioral risk factors and health care access. The second section describes results for risk factor and screening variables from each of the cancer modules and the variables asked in the diagnosis and treatment section. Frequencies and both unweighted and weighted percentages for each interview question are provided in the survey questionnaire (appendix 3). 3.1. General Attributes 3.1.1. Demographics This module of the interview asked about race, ethnic background, marital status, education level and income level and briefly ascertained occupational history. All questions, other than occupational history, were taken from the BRFSS. Age at diagnosis and gender for each respondent were obtained from the cancer registry database. Age was categorized as less than 50 years, 50 to less than 65 years, 65 to less than 80 years and 80 years or more. Race was classified into three categories: Caucasian, African- American and other race, which includes Asian, Pacific Islander, American Indian, and Alaska Native. For individuals who reported more than one race, the preferred race as indicated by the respondent was used. Hispanic ethnicity was asked as a separate question. The number of Hispanics in the general Delaware population is low, and only eight of the study respondents reported Hispanic ethnicity, so it was not used as a separate racial/ethnic category. The majority of study respondents (approximately 80 percent) were between the ages of 50 and 79. Although more females (466 respondents or 52 percent) were interviewed than males (426 respondents or 48 percent), there were more males (53 percent) in the full study population; thus the weighted percentages reflect this higher proportion of males. The racial make-up of the study population was mostly Caucasian (83 percent), with 18 percent African-American and only 0.7 percent belonging to the other race category. Again, these percentages reflect the racial distribution of the full study population. See section 2.5.1 for an explanation of the weighting for these variables. The county-level analysis of age at diagnosis and gender shows no significant differences in these attributes of the study population by county. For race, significant differences were found. Ninety-one percent of study respondents living in Sussex County at time of diagnosis were Caucasian compared with 79 percent in Kent County and 81 percent in New Castle County. The analysis of age at diagnosis by cancer type shows significant differences in the age distributions of the study population for each cancer. Breast cancer patients were youngest: 23 percent were younger than age 50 compared with 11 percent of colorectal cancer patients, seven percent of lung cancer patients and only three and a half percent of patients with prostate cancer. The data suggest that more lung cancer patients were male than female (54 percent compared with 46 percent), although the difference is not significant. A slightly higher percentage of prostate cancer patients were African-American (21 percent) compared with 15 percent of colorectal patients, 13 percent of breast cancer patients, and 11 percent of lung cancer patients, although these racial differences were not statistically significant. 3.1.2. Socioeconomic Status Respondents were asked about their marital status, education levels and income at the time of their cancer diagnosis. Marital status was classified into two groups: those who were married or part of an unmarried couple; and those who were never married, divorced, separated or widowed. Educational attainment was categorized into four groups: less than a high school education, a high school graduate or equivalent with no further education, less than four years of college and four or more years of post-secondary education. Annual household income was categorized as less than $15,000, $15,000 to less than $50,000 and $50,000 or more. The annual household income of $50,000 corresponds to the median income level in Delaware between 2002 and 2004.9 The majority of study respondents were either married or part of an unmarried couple (76 percent). There were no significant differences in marital status by county of diagnosis. More than half of the study respondents (60 percent) were educated beyond the high school level, and 44 percent reported annual household income of $50,000 or more at the time of their diagnosis. Differences in educational and income levels were observed by county. More respondents living in New Castle County at time of diagnosis received a college degree (37 percent) compared with those in Sussex County (30 percent) and Kent County (24 percent). The annual household income level for more than half of respondents living in New Castle County at time of diagnosis was above the median of $50,000, while only 35 percent of residents of Sussex County and 29 percent of residents in Kent County reported incomes above the median. Differences in marital status, educational level and income level were observed by cancer type. More prostate (88 percent) and colorectal (74 percent) cancer patients were married or partnered compared with lung (70 percent) and breast (67 percent) cancer patients. Respondents with prostate cancer were more highly educated and reported higher income levels. Thirty-nine percent received a college degree compared with 33 percent of breast cancer respondents, 30 percent of colorectal cancer respondents and 16 percent of lung cancer respondents. Fifty-one percent of patients diagnosed with prostate cancer were above the median income level compared with 41 percent of breast cancer patients, 39 percent of colorectal cancer patients and 31 percent of lung cancer patients. 3.1.3. Length of Residency Study participants were asked about where they had lived and how long they lived there. These questions were developed specifically for this interview to determine patterns of migration and acculturation in the study population. Respondents were asked to report how long they had lived in their county at diagnosis, Delaware and the United States, as well as whether they were born in the United States and the language most often spoken in their home. Thirty-three respondents were born in a foreign country, and eight respondents reported speaking a language other than English at home. Length of residency was classified into four groups: five years or less, six to 20 years, 20 years or more (excluding lifetime residents) and lifetime residency. About 50 percent of study participants reported living in their county at diagnosis for 20 years or more, and about 22 percent of respondents reported living in their county at diagnosis all their lives. Similar results were observed for length of state residency. The majority of study participants lived in the United States all their lives (93 percent). There were no differences in length of U.S. residency observed at the county level. There were reported differences in length of county and state residency at the county level. Residents of New Castle County at time of diagnosis lived in their county and Delaware the longest compared with residents of Kent County and Sussex County. More than 80 percent were either lifetime residents or lived in the county more than 20 years compared with 73 percent of Kent County residents and 47 percent of Sussex County residents. The same pattern was observed for length of residency in Delaware, with higher percentages of New Castle County residents living in Delaware longer compared with residents of Kent County or Sussex County. There were no observed differences in residency patterns by cancer type. Respondents with each type of cancer followed similar patterns as the entire study population. 3.1.4. Occupational History Respondents were first asked whether they had ever worked outside their home, and if they answered yes, they were asked to list the three jobs they held for the longest time. Each of the three jobs was classified into an industry and occupation group using the National Center for Health Statistics scheme for coding death certificates,10 which is based on the North American Industry Classification System (NAICS) and Standard Occupational Classification (SOC) codes. There are 20 and 23 broad classifications for industry and occupation, respectively. The information provided by the respondent was used to code industry based on the place of work and occupation based on the type of work. If the information provided was not descriptive enough to classify the job for either industry or occupation, it was coded as “unable to determine.” After the coding was completed, one variable was created for each occupational category to represent whether a study respondent had reported working in that occupation for any of the three jobs. Since the objective of this question was to determine the individual’s occupation, the industry codes were used to aid in verifying the high-risk industry questions asked later and are not presented here. Twenty study participants reported never working outside their home. Of the remaining 872 respondents, 355 provided three job titles, 248 provided two job titles, 246 provided one job title and 23 did not provide any job title. We were unable to code at least one occupational category for 26 study participants. The most common occupations reported were management (15 percent), education (15 percent), office and administrative support (26 percent), production (14 percent) and sales (13 percent). Together these five occupational groups included 84 percent of the study population. The following differences were observed by county: • Respondents living in Kent County at time of diagnosis reported more computer/ mathematical jobs, jobs in the life, physical and social sciences and jobs in construction/extraction compared with Sussex County or New Castle County, • More participants living in New Castle County at time of diagnosis reported working in military or transportation/moving jobs compared with Sussex County or Kent County, and • More production jobs were reported by participants living in New Castle County or Sussex County at time of diagnosis compared with Kent County. There were three significant occupational differences found by cancer type: 1) respondents with prostate cancer reported fewer management jobs, 2) respondents with prostate cancer reported more jobs in installation, repair and maintenance, and 3) respondents with colorectal cancer reported more computer and mathematical jobs. For all other occupations, the distribution for each cancer was similar to that of the entire study population. In addition to the general question asked about jobs, respondents were asked whether they had ever worked in one of the following high-risk industries: chemical, pharmaceutical, manufacturing, agriculture and construction. If they answered yes, they were asked for up to three job titles and the length of time worked in each job. Each job title was coded using the SOC classification system as described above. Individuals were then classified into either high-risk or low-risk categories depending on whether the coded occupation might have involved direct exposure to carcinogens. One variable was created with three levels: 1) never worked in a high-risk industry, 2) any work in a high-risk industry in a low-risk job, and 3) any work in a high-risk industry with a high-risk job. Twenty-one percent of study respondents reported ever working in the chemical industry, three percent in the pharmaceutical industry, 26 percent in manufacturing, 10 percent in agriculture, and 17 percent in the construction industry. In each of these industries, the majority of study participants worked in high-risk occupations, with the exception of the pharmaceutical industry, where the percentage of low-risk and high-risk jobs was similar. The only significant county- level difference was for the chemical industry. More respondents living in New Castle County at time of diagnosis were ever employed in the chemical industry (26 percent) compared with Kent County (10 percent) or Sussex County (16 percent). With the exception of the pharmaceutical industry, differences were observed by cancer type for participants working in each of the other high-risk industries. However, because males are more likely to work in high-risk industries, gender was a confounding factor in this analysis. The analysis was rerun controlling for gender, and significant differences by cancer were no longer present except for in the farming industry. More males with colorectal cancer (25 percent) and lung cancer (24 percent) worked in the farming industry as compared with males with prostate cancer (11 percent). No differences were observed for females working in the farming industry. 3.1.5. Health Status Several interview questions addressed health status; all were taken from the BRFSS. Participants were asked about their general health in the year prior to their diagnosis and categorized into those who reported excellent, very good or good health and those who reported fair or poor health. Respondents were asked whether they had ever been told by a health practitioner that they had the following conditions: heart disease, diabetes, hypertension, high blood cholesterol, arthritis or asthma. Individual variables were created for each condition, and a co-morbidity index ranging from zero to six was created. Each study participant was asked to report his or her height and weight just before diagnosis, and the most he or she ever weighed. These variables were used to calculate the respondent’s body mass index (BMI) using the formula: BMI = weight (kg)/height (m)2 For weight just before diagnosis and highest reported weight, BMI was categorized into three groups: <25, 25 to <30 and 30+. These groups correspond to the CDC definitions for healthy weight, overweight and obesity.11 The majority of study participants reported good to excellent health in the year prior to diagnosis (92 percent). A higher proportion of respondents were classified as either overweight (39 percent) or obese (24 percent) compared with those who were neither overweight nor obese (37 percent). For highest weight reported, 40 percent of respondents were overweight, 36 percent were obese and 24 percent were neither overweight nor obese. About 31 percent of respondents did not have any of the six co-morbid conditions, 49 percent reported having one or two co-morbid conditions and the remaining 20 percent reported three or more co-morbid conditions. The prevalence of each of the six co-morbid conditions in the study populations was reported as follows: 17 percent with heart disease, 12 percent with diabetes, 38 percent with hypertension, 35 percent with high blood cholesterol, 30 percent with arthritis and 8 percent with asthma. There were no significant differences in any of these health status variables at the county level, indicating that the health status of the study population was similar in each of Delaware’s three counties. However, differences were observed in health status among the four cancer types. Fewer lung cancer patients reported good to excellent health before diagnosis (81 percent) when compared with patients with prostate, breast or colorectal cancer (90 percent or more for each cancer). More breast cancer patients reported a healthy weight prior to diagnosis (45 percent) compared with colorectal (32 percent), lung (34 percent) and prostate (33 percent) cancer patients. A similar pattern was seen for highest reported weight. More lung cancer patients reported three or more co-morbid conditions (27 percent), and fewer lung cancer patients reported not having any co-morbid conditions (19 percent) compared with the other cancer patients. This is consistent with the finding that lung cancer patients were less likely to report good to excellent health prior to diagnosis compared with the other cancer patients. Differences were also observed in the prevalence of each condition separately. Patients with lung and prostate cancer were more likely to have high blood cholesterol (40 percent and 39 percent, respectively) compared with patients with breast cancer or colorectal cancer (32 percent and 27 percent, respectively). Lung cancer patients were more likely to have asthma (16 percent) compared with patients with breast cancer, colorectal cancer or prostate cancer (all less than 10 percent). 3.1.6. Diet Diet questions were designed to measure consumption of fruit and vegetables and dietary fat intake at the time of diagnosis. Respondents were asked how often they usually ate certain foods. These questions were taken from the BRFSS and grouped to form two indices: • Fruit and vegetable intake—This index reflects the daily number of servings of fruit and vegetables the respondent eats. The index categorizes respondents into four groups: those who ate fruit and vegetables less than once per day, once to less than three times per day, three to less than five times per day and five or more times per day. The index was derived from the SAS code provided by BRFSS.12 • Dietary fat intake—The 13-item dietary fat screener was developed by Block (1989) to identify groups of individuals with high or low fat intake by estimating intakes of fat and saturated fat in the few foods believed to contribute the most to total fat intake.13 These questions were used in the BRFSS from 1990–94. The fat index was derived using the same algorithm as for the fruit and vegetable index and categorized respondents into the same four categories. About 20 percent of the study population met the current recommendation for eating five or more servings of fruits and vegetables per day. More than 60 percent of respondents reported eating foods high in fat three or more times per day. There were no differences in fruit and vegetable intake or dietary fat intake at the county level. Significant differences were found for fruit and vegetable consumption for each cancer type. Breast cancer patients ate the most servings of fruits and vegetables per day, while lung cancer patients ate the fewest servings of fruits and vegetables per day. Differences in fat consumption were borderline significant. The data suggest that lung cancer and colorectal cancer patients tend to eat high-fat foods more often than breast cancer and prostate cancer patients. 3.1.7. Alcohol Consumption Respondents were asked a series of three questions regarding alcohol consumption. These questions were taken from the BRFSS survey and designed to measure usual consumption of alcohol, heavy drinking and binge drinking. Respondents were asked whether they had at least one drink of alcohol in the 30 days prior to diagnosis. A drink of alcohol was defined as one 12 oz. can or bottle of beer, one glass of wine, one 12 oz. can or bottle of wine cooler, one cocktail or one shot of liquor. Respondents were asked how many drinks they drank per day on average and how many times in the 30 days prior to diagnosis they had five or more drinks. Heavy drinkers were defined as males who consumed on average three or more drinks per day or females who consumed on average two or more drinks per day. Binge drinkers were defined as respondents who reported having a drink in the past 30 days and who drank five or more drinks on one or more occasions in the past month. About 60 percent of the study population reported having at least one drink in the 30 days prior to diagnosis, 8 percent were classified as heavy drinkers and 13 percent were classified as binge drinkers. Differences were seen for all three alcohol indicators by county. Respondents living in New Castle County at time of diagnosis were more likely to have had a drink in the 30 days prior to diagnosis and were more likely to be both heavy drinkers and binge drinkers compared with respondents living in either Kent County or Sussex County. Alcohol consumption also differed by cancer type. A higher proportion of prostate cancer patients (68 percent) reported having a drink in the 30 days prior to diagnosis compared with the lung (57 percent), breast (57 percent) and colorectal (45 percent) cancer patients. More lung and prostate cancer patients were classified as heavy drinkers and binge drinkers than respondents with breast cancer or colorectal cancer. 3.1.8. Tobacco Indicators The questions regarding tobacco use were taken from the BRFSS. The BRFSS definitions for smoking status were used to categorize respondents as current smokers, former smokers or never smoked based on their smoking history prior to diagnosis. Current smokers were defined as individuals who had smoked at least 100 cigarettes in their lifetime and smoked every day or some days at the time of diagnosis. Former smokers were defined as individuals who had smoked at least 100 cigarettes in their lifetime and did not smoke at time of diagnosis. Individuals who had not smoked at least 100 cigarettes in their life at the time of diagnosis were classified as never having smoked. Age of smoking initiation was categorized into two groups: individuals who started smoking before age 18 and those who started smoking at age 18 or older. Respondents were also asked questions regarding passive tobacco exposure. These questions were taken from the BRFSS and used to determine the probability of passive exposure to tobacco smoke based on the reported smoking rules in the home and at work. Respondents were categorized into four categories: possibly exposed at home, possibly exposed at work, possibly exposed at both home and work, and probably not exposed. Exposure was defined based on the reported smoking rules. If respondents reported that smoking was allowed either anywhere or in some places at home, or in public or work areas at their job, then they were classified as possibly exposed. Respondents who reported that smoking was not allowed anywhere at home, or in public or work areas at their job, were considered probably not exposed. The passive tobacco exposure status of respondents who reported that there were no rules about smoking at home or in public or work areas at their job was considered unknown. Overall, 19 percent of study respondents reported smoking at the time of their diagnosis, and 43 percent reported formerly smoking. Sixty-seven percent of study participants who had ever smoked started smoking before the age of 18. No differences in smoking status or age first smoked were observed at the county level. Forty-two percent of study participants reported that smoking was not allowed at either home or work, and only five percent reported that smoking was allowed at both home and work. Twenty-six percent reported that smoking was allowed at home only, and 27 percent reported that smoking was allowed at work only. There were no reported differences by county regarding smoking rules at home or work. As would be expected, a much higher proportion of lung cancer patients were smokers at the time of diagnosis or formerly smoked (93 percent) compared with patients with breast cancer (47 percent), colorectal cancer (57 percent) or prostate cancer (68 percent). Eighty-two percent of respondents with lung cancer who had ever smoked reported that they started smoking before the age of 18, compared with 73 percent of respondents with prostate cancer, 57 percent of breast cancer respondents and 55 percent of respondents with colorectal cancer. Lung cancer patients were also more likely to report that smoking was allowed at either home or work or both (86 percent). 3.1.9. Physical Activity Questions about physical activity were designed to determine the respondents’ level of physical activity in their leisure time and while at work. These questions were taken from the BRFSS, and SAS code from the BRFSS was used to calculate the length and duration of leisure-time activity levels. Respondents who worked at the time of their diagnosis were asked about the level of activity that their job required. They were categorized as working either at jobs where they were mostly sitting or standing or working at jobs that required walking or more physically demanding work. Three variables were created for leisure-time activity levels based on BRFSS definitions. The first variable measured whether the respondent reported participating in any type of leisure- time activities or exercise such as running, calisthenics, golf, gardening or walking for exercise. Respondents were also asked about moderate and vigorous activity levels. For moderate activity level, respondents were classified based on whether they met the requirement for doing moderate activities for at least 30 minutes, on five or more days per week. For vigorous activity level, respondents were classified based on whether they met the requirement for doing vigorous activity for at least 20 minutes, on three or more days per week. Overall, 68 percent of study participants who worked at the time of their diagnosis reported sitting or standing at work, and 32 percent reported working at jobs that were more physically demanding. Seventy-seven percent of the study population reported doing some leisure-time physical activity or exercise. About half of study participants met the requirement for moderate activity level, and 26 percent met the requirement for vigorous activity level. The only observed difference in activity levels at the county level was for leisure-time activity. More respondents living in Sussex County at time of diagnosis (81 percent) reported participating in leisure-time activities or exercise than respondents living in New Castle County (78 percent) or Kent County (69 percent). There were no observed differences in reported work activity levels among the four cancer types. However, the data suggest that lung cancer respondents worked in more physically demanding jobs compared with respondents with other cancers. Prostate cancer patients were the most physically active during their leisure time and were more likely to meet the recommendations for moderate and vigorous activity levels as compared with lung, breast and colorectal cancer patients. 3.1.10. Health Care Access The questions about health care access were taken from the BRFSS. Respondents were asked about availability of health insurance, their usual source of care, health care utilization and satisfaction with care. Those without insurance or who had other barriers to care were asked their reasons. Usual source of care was categorized as seeking non-emergency care in a doctor’s office, public health clinic or community health clinic; seeking care in a hospital or urgent care center; and other or no usual source of care. Health care utilization was categorized based on the frequency of seeking care; the three categories were: seeing a medical professional at least every six months, seeing one every seven months to a year, or seeing one less than once a year. Overall, 98 percent of survey respondents had health insurance at time of diagnosis, so there were too few without health insurance to explore reasons for that lack. Additionally, 95 percent of the study participants (850 respondents) reported having at least one personal doctor. The 42 remaining respondents were asked their reasons for not having a personal doctor. They were read a list of reasons and given the option to answer yes to any that applied. The most common answers given were: • Was healthy so didn’t need one (35 respondents or 83 percent), • Cost (15 respondents or 36 percent), • No health insurance (10 respondents or 24 percent), • Didn’t believe it would help (six respondents or 14 percent), and • Hadn’t found a doctor they liked (five respondents or 12 percent). The analysis of health care coverage by county revealed fewer respondents in Kent County (95 percent) had health insurance compared with respondents in New Castle County (99 percent) and Sussex County (97 percent). Overall 97 percent of the study population reported their usual source of care was in a doctor’s office or public health or community health clinic, and 90 percent of respondents reported seeing a doctor at least once a year. There were no differences in these variables at the county level. There were significant differences in health care access and utilization among the cancer types. Lung cancer patients were less likely to have health insurance compared with the other cancer patients. A higher percentage of colorectal cancer patients saw a doctor less than once a year (18 percent), compared with those with prostate cancer (12 percent), lung cancer (10 percent) or breast cancer (5 percent). Breast and prostate cancer patients reported non-emergency care as their usual source of care (97 and 96 percent, respectively) more often than colorectal or lung cancer patients (93 and 92 percent, respectively). 3.1.11. Comparison of Risk Factors With the General Delaware population In this section, the health risk factors previously described are compared for each of the four cancer types with the results from the 2002 BRFSS survey of the general Delaware population. We were not able to compare all the risk factors previously described because some were not asked on the 2002 BRFSS survey. These include four of the health conditions (heart disease, hypertension, high blood cholesterol and arthritis) and dietary fat intake. The odds ratios presented for each risk factor are for the at-risk group compared with the not-at-risk group for individuals with cancer compared with the general Delaware population. The at-risk group for each risk factor is defined as follows: • Health status: respondents who report having fair or poor health, • Body mass index: respondents who are overweight and/or obese, • Diabetes: respondents who have been told by a doctor, nurse or health professional that they had diabetes, • Asthma: respondents who have been told by a doctor, nurse or health professional that they had asthma, • Fruit and vegetable consumption: respondents who report they never consume fruits and vegetables or consume fewer than five servings per day, • Alcohol consumption: respondents who report drinking alcohol in the past 30 days, • Heavy drinkers: male respondents who report having more than two drinks per day, or female respondents who report having more than one drink per day, • Binge drinkers: respondents who report they did drink in the past 30 days and had five or more drinks on one or more occasions in the past month, • Smokers: respondents who report having smoked at least 100 cigarettes in their lifetime and smoked at time of diagnosis, • Former smokers: respondents who report having smoked at least 100 cigarettes in their lifetime and did not smoke at time of diagnosis, • Smoking initiation: respondents who report they smoked their first cigarette before the age of 18, • Work activity: respondents who report mostly sitting or standing at their job, • Physical activity: respondents who report doing no moderate or vigorous physical activity or exercise, • Moderate activity: respondents who report doing insufficient moderate or vigorous physical activity to meet recommendations, or respondents who report doing no moderate or vigorous physical activity, and • Vigorous activity: respondents who report doing insufficient vigorous physical activity to meet recommendations, or respondents who report doing no vigorous physical activity. For each of these risk factors, the odds ratio for the at-risk group compared with the not-at-risk group is reported for individuals with cancer compared with those without cancer. The odds ratios are adjusted for five-year age group, sex, race and county of residence. An odds ratio significantly greater than one indicates respondents with cancer are more likely to be at risk compared with the general Delaware population, and an odds ratio significantly less than one indicates that respondents with cancer are less likely to be at risk. If the value of one falls outside the confidence interval, the odds ratio is significant. Breast, colorectal and prostate cancer patients were all less likely to report being in fair or poor health prior to diagnosis compared with the Delaware population. There was no difference in reported health for lung cancer patients. For BMI, there were no differences between any of the four cancer types and the Delaware population. With the exception of colorectal cancer patients, the data suggest cancer patients are somewhat less likely to be overweight and/or obese compared with the Delaware population. There were no differences between the cancer population and the Delaware population regarding diabetes. The data do suggest cancer patients are somewhat less likely to have diabetes than the general Delaware population. Breast cancer patients were almost half as likely to have asthma when compared with the Delaware population. For colorectal, prostate and lung cancer, there were no differences, although the data suggest lung cancer patients were somewhat more likely to have asthma than the general Delaware population. The consumption of fruits and vegetables less than five times a day does not appear to be a significant risk factor for the cancer population, although the data show cancer patients are slightly more likely to eat less than five servings a day compared with the Delaware population. Several differences were observed for alcohol consumption. Breast cancer patients were one and a half times more likely and prostate cancer patients were almost two times more likely to report having had a drink in the past 30 days compared with the Delaware population. Lung cancer patients were three times more likely to be heavy drinkers, and prostate cancer patients were two times more likely to be heavy drinkers compared with the Delaware population. Breast cancer patients were 2.7 times more likely, lung cancer patients were 2.4 times more likely and prostate cancer patients were twice as likely to report binge drinking compared with the Delaware population. The risk for binge drinking for colorectal cancer patients was elevated but not significant. Lung cancer patients were 27 times more likely to smoke at the time of their diagnosis and more than seven times more likely to formerly smoke compared with the Delaware population. They were almost twice as likely to begin smoking before the age of 18. For the other cancers, current or former smoking status was not a significant risk factor. Breast and colorectal cancer patients were about half as likely to report that they began smoking before the age of 18, compared with the Delaware population. In general, the activity levels of the cancer population before diagnosis are higher than that of the general Delaware population, with the exception of work activity levels. Work activity level is not significantly different for any of the four cancers, but the data suggest colorectal and prostate cancer patients are somewhat more likely to report mostly sitting and standing at work compared with the Delaware population, and breast and lung cancer patients are somewhat less likely to report sitting or standing at work. Breast, colorectal and prostate cancer patients are about three times more likely to report being physically active or exercising compared with the Delaware population and are also more likely to meet the requirements for moderate physical activity levels. Breast and prostate cancer respondents are also more likely to meet the requirements for vigorous activity levels. Lung cancer patients show no differences for any physical activity or moderate physical activity but are more than twice as likely to meet the requirements for vigorous activity levels. 3.2. Cancer-Specific Attributes The previous sections describe the results of general questions all study respondents were asked. The sections below present the results of the cancer site-specific questions. In particular, results are presented from questions about specific cancer risk factors and cancer-specific screening usage. 3.2.1. Breast Cancer This section of the report outlines the questions included in the breast cancer module and displays the summarized information. There were 335 respondents with breast cancer; for some of the variables the total will be less due to missing values from respondents not knowing the answer to a question or choosing not to answer it. Exposure to estrogen is associated with breast cancer risk. Therefore, many reproductive events in a woman’s life have been shown to affect her risk of breast cancer. A reproductive risk variable was created that counts the number of reproductive risk factors a woman has.14 This variable included the following risk factors: • Never pregnant or a first pregnancy after age 35, • Never breast fed, • Took hormone replacement therapy for more than five years, • Reached menarche before age 12, and • Reached menopause after age 55. A second variable of other non-reproductive risk factors for breast cancer was created. This variable counted the number of other breast cancer risk factors a respondent had: • Took diethylstilbestrol, • Received radiation to the chest, • History of abnormal breast biopsies, and • Previous diagnosis of breast cancer. Family history of breast cancer was examined by adding up the number of close family relatives who were diagnosed with breast cancer. Close family relatives include grandmothers, mother, aunts, sisters and daughters. Overall, 21 percent of respondents did not have any reproductive risk factors for breast cancer. The data suggest more women in Sussex County have reproductive risk factors than in Kent County and New Castle County. A majority of the participants had one or two reproductive risk factors; none of the participants had all five. The most common risk factors were never having breast fed (149 women) and taking hormone replacement therapy for more than five years (91 women). The non-reproductive risk factors were rarely present; 75 percent of the survey population had none of these risk factors. This prevalence did not differ by county of residence. Of the risk factors included in this variable, the most common risk factor was a history of an abnormal breast biopsy (62 women). Overall, 66 percent of the study population did not have an immediate family member with breast cancer, and there was no evidence of a difference by county of residence. Mothers and aunts were the most commonly reported family member with breast cancer (40 and 45 women, respectively). Respondents with breast cancer were asked about their history of screening for breast cancer by mammography, clinical breast exams and breast self-exams. For each of the three tests the following was determined: • The percentage of women who ever received or performed the test, and • The percentage of women who followed the current recommendations regarding the age to begin testing and the frequency of testing.15 For mammography, a woman was considered to have followed recommendations if she received her first mammogram before age 50 and reported receiving at least a yearly mammogram afterward. Currently, women are recommended to have a mammogram beginning at age 40, but many of the study respondents were in their 40s before this recommendation was enacted, so we used the previous recommendation of receiving a first mammogram beginning at age 50. For clinical breast exams, the recommendation is to receive annual exams, and for self-exams, the recommendation is for monthly exams. Overall, the usage of breast cancer screening tests is very high. Of the respondents in this study, 93 percent have received a mammogram, 90 percent have received a clinical breast exam and 85 percent have performed a breast self-exam. There were 25 women who had never received a mammogram. Of these, nine women were younger than 40 when they were diagnosed with breast cancer, so they were not eligible for mammography. A majority of women (58 percent) began receiving mammography at the recommended age and are receiving the tests yearly. For the remaining 42 percent of women who did not meet this criterion, the majority did not meet the age-at-first-mammogram criterion. The question about the average frequency of mammography was compared with the data from the question regarding the last screening mammogram before diagnosis. Of the 284 women who reported they generally have a mammogram every year or more than once a year, 161 reported that their last screening mammogram was less than 12 months ago, 101 women reported it was more than one year but less than two years before diagnosis and the remainder reported it had been more than two years. Annual clinical breast exams were reported by 85 percent of the population. Again, the frequency of screening was compared with last screening before diagnosis, and for clinical breast exams 279 respondents reported receiving a clinical breast exam at least once a year, 186 reported a clinical breast exam in the 12 months before diagnosis, 77 reported a test more than one year but less than two years before diagnosis and the remaining respondents reported it had been more than two years since their last clinical breast exam before diagnosis. Monthly breast exams were reported by 56 percent of the population. There were no county differences in mammography usage or performance of self-exams, but the data suggest lower receipt of annual clinical breast exams in Kent County compared with New Castle County or Sussex County. The women in our study population were compared with the general Delaware population using data from the BRFSS. The BRFSS does not collect information about breast self-exams. Therefore, the table below examines the percentages of women who reported having a mammogram or a clinical breast exam within the past two years. The past two years is defined based on the date of interview for the BRFSS population, and on the date of diagnosis in our cancer population. Our screening definitions were modified from above to allow comparisons with the BRFSS data. The data on mammograms are restricted to women aged 40 and older. Even with this restriction, the age distribution of the general population and the cancer population differ substantially, so the age-adjusted prevalence in our population and the general population are presented. In addition, an odds ratio and 95-percent confidence interval are displayed. The odds ratios have been adjusted for five-year age groups, race and county of residence. For each test we are comparing the percentage of women in the cancer population who have not received the screening test within the past two years with the percentage in the general population who have not received the test in the past two years. The reference group for the odds ratio is women who received a screening test. Therefore, odds ratios greater than one indicate women in the cancer population were less likely to have received a screening test than women in the general population, while odds ratios less than one indicate women in the cancer population were more likely to receive a screening test than women in the general population. There were no differences in the percentage of women with breast cancer compared with the general female Delaware population regarding the likelihood of receiving a mammogram. However, women with breast cancer were 16 percent more likely to have not received a clinical breast exam in the past two years compared with the general Delaware population. 3.2.2. Colorectal Cancer Survey respondents with colorectal cancer were asked questions about specific risk factors for colorectal cancer, family history of colorectal cancer and usage of colorectal cancer screening tests. There were 163 respondents with colorectal cancer; some variables may have fewer numbers due to missing values. The non-behavioral risk factors for colorectal cancer were added together to create a summary variable. The factors were: • Previous diagnosis of colorectal cancer, • History of intestinal polyps, and • History of chronic inflammatory bowel disease. A family history of colorectal cancer includes a diagnosis of colorectal cancer in the respondent’s grandparents, parents, siblings and children. The colorectal cancer-specific risk factors were rare in this population, with 79 percent of the total survey respondents having none of the risk factors. None of the respondents had all three risk factors. The most common risk factor was intestinal polyps (20 respondents). Overall, 79 percent of the study respondents did not have any immediate family members who were diagnosed with colorectal cancer. Parents were the most commonly reported family members with colorectal cancer (19 respondents). The data suggest respondents in Sussex County were least likely to have a colorectal cancer risk factor but most likely to have a family history of colorectal cancer compared with respondents in other counties. Survey respondents were asked about two screening tests for colorectal cancer: • Sigmoidoscopy/colonoscopy. • Fecal occult blood test, and For each test, the proportion of respondents who had never received the test was calculated, and the proportion of respondents who followed the recommendation regarding age to begin screening and frequency of screening was calculated.16 Respondents were considered to have met the criteria for following the recommendations for a fecal occult blood test if they indicated they received their first test by age 55 and were screened every year afterward. For sigmoidoscopy and colonoscopy we will consider respondents to have met the screening requirements if they received one of these procedures beginning before age 55 and at least every five years afterward. Overall, the use of colorectal cancer screening tests is lower than breast cancer, with 36 percent of the population having ever received a fecal occult blood test and 54 percent having ever received a sigmoidoscopy or colonoscopy. Only 19 of the respondents with colorectal cancer were diagnosed before age 50, so the low rate of screening is not due to ineligibility. Overall, 10 percent of the study respondents followed guidelines for the fecal occult blood test and 18 percent for sigmoidoscopy. Study respondents in Sussex County were less likely to have received a sigmoidoscopy. The data suggest residents of Sussex County were least likely to have ever received a fecal occult blood test; however, among those who have ever received the test, Sussex County residents may be most likely to receive annual screening compared with respondents in the other counties. The screening history of the study respondents with colorectal cancer was compared with the general population using the BRFSS data. To compare our data with the BRFSS data, the percentage of respondents who had received a fecal occult blood stool test in the past two years and the percentage who had ever received a sigmoidoscopy or colonoscopy were examined. The BRFSS asks respondents only aged 50 and older whether they received screening for colorectal cancer. Therefore, to be able to compare our data on cancer patients with the BRFSS data we restricted our data to colorectal patients who were at least 50 years old at diagnosis. The percentages reported are age-adjusted to address the differences in age in the cancer and general populations. The odds ratios are adjusted for five-year age group, sex, race and county of residence and examine the odds of not being screened in the cancer population as compared with the general population. The reference group for the odds ratio is respondents who received a screening test. Therefore, odds ratios greater than one indicate respondents in the cancer population were less likely to have received a screening test than respondents in the general population, while odds ratios less than one indicate respondents in the cancer population were more likely to receive a screening test than respondents in the general population. Respondents with colorectal cancer were two and a half times less likely to have received a fecal occult blood stool test in the two years before diagnosis, compared with the general Delaware population. There is no difference in receipt of a sigmoidoscopy or colonoscopy. 3.2.3. Prostate Cancer There were no risk factors that are specific for prostate cancer, so survey respondents with prostate cancer were asked about family history and usage of screening tests. There were 307 men with prostate cancer in the survey; variables with less than this number are due to missing values. A family history of prostate cancer was determined based on the number of immediate family members who were diagnosed with prostate cancer. This includes grandfathers, father, uncles and brothers. Overall, 71 percent of participants did not have any family members who had been diagnosed with prostate cancer. The data suggest respondents in Sussex County were less likely to report a family history than respondents in Kent or New Castle counties. Of the men who reported a family history, the most commonly reported relative with prostate cancer was a father (43 men). There are two screening tests available for prostate cancer: the prostate-specific antigen test and digital rectal exams. The efficacy of these screening tests for prostate cancer has not been definitively established.17 For each of the two tests, the proportion of men who had ever received the test and the proportion who reported an annual test beginning before age 55 were calculated. Study respondents commonly received both prostate-specific antigen tests and digital rectal exams. Before their prostate cancer diagnosis, 82 percent of men had received a prostate-specific antigen test, and 93 percent of men had received a digital rectal exam. However, a smaller proportion of men reported receiving annual prostate-specific antigen tests and digital rectal exams (32 and 42 percent, respectively). There is no evidence that the use of prostate cancer screening tests differs by county of residence. The use of prostate cancer screening tests in the study population was compared with the general population. Men who indicated during the BRFSS interview that they had been diagnosed with prostate cancer were excluded from this analysis. In addition, the BRFSS asks only men aged 40 and older the prostate cancer screening questions, so our study population was restricted to men aged 40 and older. To compare our data with the BRFSS data, the percentages of men who had received a prostate-specific antigen test or digital rectal exam in the past two years were calculated. The odds ratios are adjusted for five-year age group, race and county of residence and examine the odds of not being screened in the cancer population as compared with the general population. The reference group for the odds ratio is men who received a screening test. Odds ratios greater than one indicate respondents in the cancer population were less likely to have received a screening test than respondents in the general population, while odds ratios less than one indicate respondents in the cancer population were more likely to receive a screening test than respondents in the general population. There was no difference in ever having received a prostate-specific antigen test comparing the study population with the general Delaware population; however, after adjusting for confounders, the study respondents with prostate cancer were more than twice as likely to have received a digital rectal exam than the general population (odds ratio 0.42). 3.2.4. Lung Cancer This section of the report outlines the questions included in the lung cancer module and displays the summarized information. There were 95 respondents with lung cancer; for some variables, the total will be less due to missing values. Currently, there is no population-based screening test for lung cancer, so screening usage was not included in the lung cancer module. The number of non-behavioral lung cancer-specific risk factors was calculated for each respondent with lung cancer. The risk factors included in this variable are: • Prior diagnosis of lung cancer, • Prior diagnosis of tuberculosis, • History of radiation to the chest, and • Lived in a home with high radon levels. A family history of lung cancer was determined by adding the number of immediate relatives who had been diagnosed with lung cancer. Eligible relatives were grandparents, parents and siblings. Among the respondents with lung cancer, 88 percent reported no non-behavioral lung cancer risk factors. There was no difference in the prevalence of lung cancer risk factors by county of residence. The most commonly reported risk factor was radiation to the chest (seven respondents). None of the respondents who have tested their homes for radon (19 respondents) reported high levels of radon. In this population, 75 percent of the respondents had no family history of lung cancer, and the maximum number of relatives with lung cancer was two. The most commonly reported relative with lung cancer was a father (10 respondents). There is evidence that occupational exposure to carcinogens may be a risk factor for lung cancer. Respondents were asked whether they had any occupational exposure to a list of 20 occupational chemicals, which included: solvents; asbestos; mineral or mining dust; silica/sand and mineral dust; gasoline, diesel fuel and engine exhaust fumes; welding fumes; electroplating fumes; ether (chloromethyl and/or bischloromethyl); pesticides; arsenic; nickel; cadmium; radon; plutonium; uranium; vinyl chloride; nickel chromates; coal; mustard gas; and formaldehyde. The proportion of respondents who were exposed to any of the chemicals was calculated, as well as the number of chemicals to which a respondent was exposed. Overall, 60 percent of respondents with lung cancer were exposed to at least one of the potential lung carcinogens while they were employed. The data suggest respondents in Sussex County were most likely to be exposed, and respondents in Kent County were least likely to be exposed. The top three most commonly reported exposures were solvents, asbestos and gasoline, and diesel and engine exhaust fumes (33, 24 and 24 respondents, respectively). None of the respondents were exposed to radon, plutonium, nickel chromates or mustard gas. 3.2.5. Diagnosis and Treatment This section displays the results from the data provided by the respondents regarding their cancer diagnosis process and their decision process regarding treatment and the treatment they received. These questions were used in the Maryland Department of Health and Mental Hygiene’s Look Back Study of Invasive Cervical Cancer & Late Stage Breast Cancer and were adapted for this interview. 3.2.5.1. Cancer Diagnosis Respondents were asked to report how their cancer was detected: by a screening test; by experiencing a symptom, including a lump that caused them to seek medical attention; or by a physician during an unrelated visit or procedure; or the respondent was allowed to provide another response. For the most part, the responses given were recoded into the available categories. A number of respondents indicated their cancer was found through a diagnostic test; however, from the information they provided, the indication for the test could not be determined, so they were coded as unknown. Respondents were asked the specialty of the physician who diagnosed their cancer. This variable was categorized as: 1) a primary care physician, which included gynecologists, 2) oncology specialists, which included surgeons and radiologists and 3) non-oncology specialists, which included urologists, gastroenterologists, proctologists and pulmonologists. The time from detection to diagnosis was measured from the time that respondents experienced a symptom, had a positive screening test or had a suspicious clinical finding until they were told they had cancer. The time from detection to diagnosis was categorized as less than one month, one to three months and more than three months.18 All three of the variables that examined cancer detection differed by cancer type. The majority of breast cancer was found by a screening test, but 41 percent of respondents reported a symptom. Of the women who reported that their cancer was found by a symptom, 67 percent reported having a normal mammogram within a year of diagnosis, and 23 percent reported having a mammogram within two years of diagnosis. Half of the respondents with colorectal cancer reported that a symptom led to their cancer diagnosis. The majority of prostate cancers were detected by a screening test. Lung cancer was more likely to have been found by a physician than the other cancers. Currently, the data do not support reducing lung cancer mortality by population-based screening. However, both chest x-rays and sputum cytology can be used to detect lung cancer in asymptomatic patients, which is why 28 percent of the lung cancer patients reported that their cancer was detected by a screening test.19 Primary care physicians were most likely to diagnose lung cancer, oncologists were most likely to diagnose breast cancer and non-oncology specialists were most likely to diagnose colorectal and prostate cancer (gastroenterologists and urologists, respectively). Overall, about 60 percent of the respondents had less than one month between detecting their cancer and receiving a diagnosis. The data suggest that on average women with breast cancer had the shortest time from detection to diagnosis. Of the respondents with less than one-month between cancer detection and diagnosis, 75 percent were respondents whose cancer was detected by a physician or a screening test. Of the respondents with more than three months between cancer detection and diagnosis, 69 percent were respondents whose cancer was detected by a symptom. 3.2.5.2. Cancer Treatment Cancer treatment was examined by asking the respondents to recall whether specific treatment modalities had been recommended and received. The treatment modalities were surgery, radiation therapy, chemotherapy and hormone therapy for breast and prostate cancer respondents. Time from diagnosis to treatment was measured from the time respondents were told they had cancer until receipt of their first cancer treatment. It was categorized as less than one month, one to three months and more than three months. Recommendations and receipt of each of the treatment modalities differed by cancer site. Respondents with breast cancer were most likely to have surgery recommended, while the data suggest men with prostate cancer were the least likely to have surgery recommended. The majority of breast, colorectal and lung cancer patients received surgery, while about half of prostate cancer patients received surgery. Radiation therapy was recommended for the majority of respondents with breast and prostate cancer but less frequently for respondents with colorectal and lung cancer. The pattern for receipt of radiation therapy across cancers mirrored the recommendations. Recommendations of chemotherapy differed by cancer site, from 16 percent of respondents with prostate cancer to 42 percent of respondents with breast cancer. Respondents with breast or prostate cancer were asked about recommendations and receipt of hormone therapy. Respondents with breast cancer were more likely to be recommended and receive hormone therapy than respondents with prostate cancer. Respondents who chose not to follow all of their physician’s recommendations were asked to provide their reasons. This was asked as an open-ended question, but the responses followed some general themes. There were 68 respondents in total who answered this question. For 31 respondents, it was a personal choice to not receive the recommended treatment. Twenty-two respondents were given recommendations for multiple treatment plans that were mutually exclusive, 11 respondents received a second option and four respondents had a co-morbidity that prevented them from receiving the recommended treatments. More than 75 percent of respondents with breast, colorectal and lung cancer began their cancer treatment within one month of diagnosis, and more than 90 percent began treatment within three months. In contrast, only about one-third of men with prostate cancer began treatment within one month of diagnosis, and 22 percent began treatment more than three months after diagnosis. 4. DISCUSSION AND SUMMARY OF FINDINGS This study was designed to gain a better understanding of factors affecting cancer incidence and mortality in Delaware residents. The following sections discuss and summarize the limitations of the study and the study findings. 4.1. Limitations This study gathered retrospective information regarding risk factors and the diagnosis and treatment experience of individuals diagnosed with the four most common cancers in Delaware. Because of the consent process and the nature of the study data, the population eligible for this study consisted only of patients who were alive at the beginning of the study period. Therefore, the study results cannot be generalized to the full cancer population, but only to cancer survivors. An additional 4,246 individuals were diagnosed in Delaware between 1999 and 2003 with one of the four cancers but died before the study began, and another 117 were discovered to be deceased during the study period. Since deceased individuals were more likely to be diagnosed with distant disease (40 percent versus three percent of survivors), the study findings for factors related to stage at diagnosis, such as screening practices and diagnosis and treatment outcomes, can be interpreted only within the context of living cancer patients. The selection of survivors for this study is less likely to impact the findings for lifestyle risk factors, since the likelihood of systematic differences with respect to these factors in those diagnosed with early-stage versus late-stage disease is not as great. Due to the nature of the recruitment process, the low response rate and the self-selection of study participants, there is a likelihood of selection bias. Selection bias exists if individuals who participate differ systematically from those who do not. Many studies have shown that those volunteering to participate are healthier and more likely to comply with medical recommendations than those who do not volunteer. Several of the comments made by individuals contacted for this study who refused or were unable to participate indicated this to be true. If the study participants were different in these respects compared with those who did not participate, then prevalence estimates for variables related to health, screening or medical treatments would reflect higher rates of healthy living practices and higher rates of compliance with screening and treatment recommendations than are actually present for the full study population. In addition, these inflated rates would reduce the measured effect of the variables for cancer patients compared with the general Delaware population if the same selection bias were not present for participants in the BRFSS. Since the BRFSS sample of the general Delaware population is randomly selected, it is unlikely this type of selection bias is present. Through the use of weighting, we were able to adjust for selection differences in basic demographic variables such as age, race and gender, and we controlled for these differences in comparisons of the study population with the general Delaware population. This study attempted to determine the prevalence of lifestyle risk factors, health conditions, health care access and screening practices for cancer patients prior to their cancer diagnosis. Because cancer patients were asked to recall these practices prior to diagnosis—and for some this may have been as long as six years—their responses may not be as accurate as those provided by participants in the BRFSS. The possibility of recall bias exists if systematic differences in the respondents’ ability to recall their past behaviors occurred between the two groups. A related type of bias is called reporting bias, in which a participant may be reluctant to report an exposure due to perceived beliefs and attitudes. If such underreporting exists more among cancer patients than among the BRFSS participants, a bias may result. For example, cancer patients may be more likely to deny certain exposures related to lifestyle, such as smoking or drinking, because they seek to show that the disease is not their fault. The term “wish bias” was coined by Wynder et al. to describe this type of reporting bias.20 Since the interviewers who conducted the interviews for the cancer patients were not the same as those who conducted the interviews for the BRFSS, interviewer bias is another type of information bias that may be present in this study. To minimize this bias, a very structured interview format was used, and interviewers were trained to not deviate from the standard format. In addition, interviewers who had prior experience conducting BRFSS interviews were selected to conduct the interviews for this study. Due to these factors and the nature of the BRFSS questions, the likelihood of interviewer bias is considered to be extremely small. 4.2. General Attributes The prevalence of demographic, lifestyle and health care access factors in the study population were analyzed by county. Differences were observed in race, education and income levels, length of residency, occupation, alcohol consumption, leisure-time physical activity levels and health care coverage. A higher proportion of residents of Sussex County at time of diagnosis were Caucasian compared with residents of Kent County or New Castle County. Education and income levels for residents of New Castle County at time of diagnosis were higher than those reported by residents of Kent County or Sussex County. These results are consistent with the racial and socioeconomic distributions of the general Delaware population in these counties as reported in the 2000 U.S. Census.21 Residents of New Castle County at time of diagnosis had lived in their county and in Delaware longer than residents of Kent County or Sussex County. This result is also consistent with U.S. Census estimates of population change, which show a lower rate of population change in New Castle County (4.5 percent) during the period 1995–2000 compared with that in either Kent County (13.6 percent) or Sussex County (12.7 percent).22 Although some county-level differences were found for occupational categories, these associations were based on very small numbers and may not be reliable. These differences may be explained by the concentration of industries in each county that require more of specific occupational categories—such as construction or production—than others. Similarly, county- level differences in participants that worked in high-risk industries can be explained by the geographic locations of these industries in each of the three counties. Residents of New Castle County at the time of diagnosis were more likely to be heavy drinkers and to binge drink compared with residents of Kent County or Sussex County. These county-level patterns of alcohol use were also observed in the general Delaware population.23 However, county-level differences in leisure-time physical activity levels for study participants differed from those in the general Delaware population. Leisure-time activity levels for study participants were highest in residents of Sussex County, but these were lowest in residents in Sussex County in the general Delaware population.24 This inconsistency may be explained by a selection bias of healthier, more physically active study participants in Sussex County. For each cancer, the prevalence of health status variables and behavioral risk factors were compared with that reported in the general Delaware population to determine which factors are associated with cancer incidence in Delaware. The general trend suggested by the data for health status variables and health conditions was that our study population appeared to be healthier before diagnosis than the general Delaware population. This phenomenon may be explained by a selection bias, as discussed previously, in which the healthiest individuals are those who are most likely to volunteer to participate in such a study. Study participants with breast, colorectal and prostate cancer were three to four times more likely to report being in good or better health before diagnosis compared with the general Delaware population. This may be due to a combination of selection bias and the perception of cancer patients that their health was much better before they were diagnosed compared with after diagnosis. Lung cancer patients reported being less healthy and having more co-morbid conditions compared with the other cancer patients. Although higher BMIs have been associated with increased risk for colorectal cancer,25, 26 the odds ratio of 1.12 (CI 0.76–1.65), although elevated, was not significant in our study. Selection bias may have weakened the effect of this association. The data for fruit and vegetable consumption suggest study participants with lung, prostate and colorectal cancers were less likely to eat the currently recommended amounts of five or more servings per day compared with the general Delaware population; however, these associations were not significant. Studies of the association between fruit and vegetable consumption and colorectal, breast and prostate cancers have not conclusively supported the protective effect of higher fruit and vegetable intake; however, a consistent protective effect has been noted for lung cancer in case-control and cohort studies in many different countries.27 About 23 percent of study respondents reported eating foods high in fat five or more times per day and another 40 percent reported eating foods high in fat at least three times but less than five times per day. Fat intake is the aspect of diet that has been singled out to be most importantly related to cancer.28,29 Scientific evidence currently supports an association between high fat intake and breast, colorectal and prostate cancers. It is no surprise that the strongest risk factor association detected in this study was for smoking and lung cancer. This has been known since the early 1950s when case-control studies documented the association between lung cancer and smoking.30, 31 Ninety-four percent of lung cancer patients were either current or former smokers. Lung cancer patients were 27 times more likely to be current smokers and seven times more likely to be former smokers compared with the general Delaware population. The associations between smoking and breast, prostate or colorectal cancers were not significant, and the scientific evidence regarding the association between smoking and these cancers is not conclusive. Alcohol use was positively associated with the risk for lung, prostate and breast cancer. Breast cancer patients were more likely to have reported having a drink in the 30 days prior to diagnosis (OR=1.58, CI 1.20–2.08) and to binge drink (OR=2.68, CI 1.54–4.69) than individuals in the general Delaware population. The association between alcohol consumption and breast cancer risk has been evaluated in more than 100 investigations that now clearly support a causal relation.32 Prostate cancer patients were twice as likely to report binge drinking, heavy drinking and having a drink in the 30 days prior to diagnosis when compared with individuals in the general Delaware population. Recent prospective studies have supported a modest direct association between alcohol consumption and prostate cancer.33,34 Lung cancer patients were two and a half times more likely to report heavy drinking and three times more likely to report binge drinking compared with members of the general Delaware population. The possibility of confounding due to the known positive association between smoking and drinking may be responsible for these elevated risks. A recent meta-analysis of both cohort and case-control studies reported that smoking was responsible for the elevated risks in studies of alcoholics and lung cancer.35 Colorectal cancer patients had elevated odds ratios for heavy drinking (OR=1.45, CI 0.73–2.90) and binge drinking (OR=1.38, CI 0.69–2.77), but these associations were not significant. The association between alcohol consumption and colorectal cancer has been controversial and not conclusively supported by the literature; however, the weight of evidence suggests that high intake of alcohol increases the risk of colorectal cancer.36 The results for the relationship between leisure-time physical activity levels and the risk for cancer were surprising. For all cancers, a significant positive effect between higher activity levels and the risk for cancer was shown. These findings are contrary to the current scientific evidence for colorectal and breast cancer that indicates an inverse relationship between physical activity and these cancers. The current evidence for lung and prostate cancer do not support an association with physical activity levels. These results may be explained by a selection bias of healthier patients and may possibly reflect a systematic difference in the perceived intensity of physical activity levels in cancer patients. Since cancer patients may have experienced reduced activity levels after their diagnosis due to surgery and other treatments, there is the likelihood they might over-report activity levels prior to diagnosis if they used their post-diagnosis activity levels as a benchmark. The wish bias effect may also explain this difference, with cancer patients more likely to intentionally over-report physical activity levels so as not to be held responsible for their disease. 4.3. Cancer-Specific Attributes Overall, almost 80 percent of the women with breast cancer in the study had at least one reproductive risk factor, while risk factors that relate to former medical events and a family history were more rarely observed in this population. The literature suggests reproductive risk factors, which are common, are moderate risk factors (OR between 1 and 2) while other risk factors, although more rare, are stronger risk factors (OR greater than 2).37 Unfortunately, reproductive, medical history and genetic risk factors, while important, are less amenable to intervention. Use of breast cancer screening tests was high among respondents with breast cancer. About 83 percent received a mammogram and a clinical breast exam in the past two years. This is higher than the current Healthy People 2010 objective that 70 percent of all women receive a mammogram in the past two years.38 There is about a 50-percent overlap in the women who did not receive a mammogram and the women who did not receive a clinical breast exam, suggesting a group of vulnerable women who are not receiving any breast cancer screening. Colorectal cancer-specific risk factors were observed in 22 percent of participants with colorectal cancer, and a similar proportion had a family history of colorectal cancer. The use of colorectal screening tests was less common. Among study respondents aged 50 and older, 18 percent of the respondents had a fecal occult blood test in the past two years, and 55 percent had a sigmoidoscopy. The Healthy People 2010 objective is that 50 percent of all adults aged 50 and older will have received a fecal occult blood test in the past two years and will have ever received a sigmoidoscopy.39 Despite the fact that screening for prostate cancer has not been proven to decrease prostate cancer mortality, 55 percent of the men with prostate cancer aged 40 and older received a prostate-specific antigen test in the past two years, and 58 percent received a digital rectal exam. Among our respondents with lung cancer, 60 percent had occupational exposure to at least one of the potential lung cancer carcinogens. Among those with occupational exposure, 93 percent were ever smokers; these two risk factors may act synergistically. The way in which cancer was first detected differed by cancer type. Respondents with breast or prostate cancer were most likely to have their cancer detected by a screening test. Population-based screening programs are available for both breast and colorectal cancers, but as the colorectal screening is more invasive and involved, it has lower uptake than breast or prostate cancer screening, which may be why fewer colorectal cancers are detected by screening. More than 40 percent of study respondents had their cancer detected as a result of experiencing a symptom. The time from when cancer was first detected until diagnosis differed by cancer type. The time ranged from eight percent of breast cancer patients to 22 percent of colorectal cancer patients reporting three or months between initial detection and diagnostic confirmation. Studies have shown that longer times from cancer detection to diagnosis are associated with a poorer prognosis.40, 41 For all four treatment modalities, the percentage of respondents who received the treatment differed by cancer type. Almost all breast and colorectal cancer patients (99 and 96 percent, respectively), most lung cancer patients (88 percent) and half of prostate cancer patients were treated with surgery. Radiation therapy was received by two-thirds of breast cancer patients, half of prostate cancer patients, about 30 percent of lung cancer patients and 13 percent of colorectal cancer patients. Between one-third and one-half of breast, colorectal and lung cancer patients were treated with chemotherapy, which was rarely received by prostate cancer patients. Forty-two percent of breast cancer patients and 30 percent of prostate cancer patients received hormone therapy. These differences reflect the current knowledge regarding appropriate treatment for each cancer type, especially considering the majority of our study respondents had early-stage cancer.42 The cancer treatment received does not always match the treatment that was recommended. This may be due to respondents having better recall of the treatment they received than what was recommended, choosing not to follow their physician’s recommendation or requesting a treatment that was not the physician’s initial recommendation. The majority of breast, colorectal and lung cancer patients received their first treatment for cancer within one month of diagnosis while the majority of men with prostate cancer began treatment within three months of diagnosis. The longer time to first treatment seen among men with prostate cancer may reflect the fact that for early stage prostate cancer “watchful waiting” is an accepted treatment plan.42 4.4. Future Directions The findings from this study have highlighted several lifestyle behaviors that are associated with increased risk for cancer in Delaware residents. These results can be used to develop cancer-control plans targeted to specific lifestyle factors that may reduce the incidence of cancer in the future. It should be noted that lifestyle risk factors were evaluated individually, and other lifestyle factors were not controlled for in this analysis. There is a strong possibility many of these lifestyle factors are interrelated. A future analysis of the combined effects of these risk factors will help gain a better understanding of their role in the etiology of these cancers in Delaware. In addition, the identification of high-risk sub-populations for these lifestyle risk factors will allow a more targeted approach for cancer control programs to reach those individuals at greatest risk. A second area that merits further investigation is the method by which cancer was detected and the effect of recent cancer screening on time-to-diagnosis in respondents whose cancer was detected from experiencing a symptom. Although our study participants had high rates of cancer screening, about 40 percent reported that their cancer was diagnosed when they experienced a symptom. A more detailed analysis of the tumor and demographic characteristics for these individuals, their screening history, and the time from detection to diagnosis may provide answers that will help to develop recommendations for increased awareness of cancer symptoms regardless of screening history. REFERENCES 1 Delaware Cancer Consortium. Turning Commitment into Action. http://www.delawarecancerconsortium.org. 2 Delaware Cancer Registry and the National Program of Cancer Registries Cancer Surveillance System, January 2005 data submission. 3 Delaware Department of Health and Social Services, Division of Public Health, Delaware Cancer Registry. Unpublished report. 4 American Cancer Society. 2006. Cancer Facts and Figures 2006. Atlanta: American Cancer Society. 5 Ries LAG, Harkins D, Krapcho M, Mariotto A, Miller BA, Feuer EJ, Clegg L, Eisner MP, Horner MJ, Howlader N, Hayat M, Hankey BF, Edwards BK (Eds). 2006. SEER Cancer Statistics Review, 1975–2003, National Cancer Institute. http://seer.cancer.gov/csr/1975_2003/, based on November 2005 SEER data submission, posted to the SEER Web site, table I-4. 6 Day, Jennifer Cheeseman. Population Projections of the United States by Age, Sex, Race, and Hispanic Origin: 1995 to 2050. 1996. U.S. Bureau of the Census, Current Population Reports, P25-1130. Washington, DC: U.S. Government Printing Office. http://www.census.gov/prod/1/pop/p25-1130/p251130.pdf. 7 Fay MP, Feuer EJ. 1997. Confidence intervals for directly standardized rates: A method based on the Gamma distribution. Statistics in Medicine, 16:791–801. 8 Peduzzi P, Concato J, Kemper E, Holford TR, Feinstein AR. 1996. A simulation study of the number of events per variable in logistic regression analysis. Journal of Clinical Epidemiology, 49(12):1373–1379. 9 DeNavas-Walt C, Proctor BD, Lee CH. Income, Poverty, and Health Insurance Coverage in the United States: 2004. http://www.census.gov/prod/2005pubs/p60-229.pdf. U.S. Census Bureau. 10 Dickens C, Chamblee R, Washington LA, Minino AM, Walker JT, Schumacher PK. Instruction Manual Part 19: Industry and Occupation Coding for Death Certificates, effective 2003. http://www.cdc.gov/nchs/data/dvs/pt19manB1.pdf . 2002. 11 Department of Health and Human Services: Centers for Disease Control and Prevention. Overweight and Obesity: Defining Overweight and Obesity. http://www.cdc.gov/nccdphp/dnpa/obesity/defining.htm. 12 National Center for Chronic Disease Prevention and Health Promotion. Calculated Variables in Data Files. http://w