Depression and the Risk of Cancer
Depression and the Risk of Cancer
Details on the GAZEL cohort are available elsewhere. The target population consisted of 44,922 employees of the French national gas and electricity company Electricité de France-Gaz de France: 31,411 men aged 40–50 years and 13,511 women aged 35–50 years. The study was approved by the French authority for data confidentiality (Commission Nationale Informatique et Liberté) and by the Ethics Evaluation Committee of the Institut National de la Santé et de la Recherche Médicale (INSERM). In 1989, a total of 20,625 employees (45.9%) (15,011 men and 5,614 women) gave written informed consent to participate. Since 1989, volunteers have been followed through annual mailed questionnaires and administrative databases.
Clinician-based diagnoses of depression were extracted from records on medically certified sickness absences. All sickness absences exceeding 7 days in a 4-year window from January 1, 1990, to December 31, 1993, were considered. This window was chosen to ensure homogeneity of the period during which exposure was measured across participants. Diagnoses were coded by company physicians using an abridged version of the International Classification of Diseases, Ninth Revision. Episodes of sickness absence for depression (SAD) corresponded to the codes indicating major depression or dysthymic disorder.
Depressive mood was assessed in 1993, 1996, and 1999 with the CES-D. This 20-item questionnaire has been designed for use in community studies and has a high internal consistency (α = 0.8 to α = 0.9 across samples). The CES-D asks participants how often they have experienced specific symptoms during the previous week (e.g., "I felt depressed"; "I felt everything I did was an effort"; "My sleep was restless"). Responses range from 0 ("hardly ever") to 3 ("most of the time"). Based on the validation of the French version, a global score of ≥17 among men and ≥23 among women may signal clinically significant depression. Since sole reliance on a binary cutoff may be statistically unsafe, we also considered CES-D score in 1993 as a continuous measure of depressive symptoms.
In order to replicate the findings of Penninx et al., we defined chronic or recurrent depressive symptoms as having CES-D scores above the validated cutoff for the years 1993, 1996, and 1999. If a subject was missing a CES-D score in one of the 3 time periods, the 2 remaining CES-D scores were used, and both values had to be above the validated cutoff.
All participants were followed up for diagnoses of primary cancers from January 1, 1994, to December 31, 2009. Diagnoses made during the period of employment came from a registry kept by the medical department at Electricité de France-Gaz de France that has been validated for accuracy and completeness. Diagnoses made after retirement came from systematic validation of each self-reported primary cancer through a diagnosis validation survey that began in 2009. Each annual questionnaire asked participants to report whether or not they had been hospitalized or diagnosed with any of several conditions, including cancer, in the preceding 12 months. All participants who self-reported cancer at least once during follow-up were contacted (if alive) and asked to give consent for a detailed diagnostic investigation with their physician.
In a first set of analyses, we considered as cases all participants with a validated diagnosis, as well as participants who reported a diagnosis of primary cancer but died from cancer before the initiation of the diagnosis validation survey. Information on vital status and date of death was obtained annually for all participants from the company, because it pays out retirement benefits. Cause-of-death data were available from baseline (i.e., January 1, 1994) to December 31, 2009, and were coded by the French national cause-of-death registry (Centre d'Épidémiologie sur les Causes Médicales de Décès, INSERM) using the Ninth and Tenth Revisions of the International Classification of Diseases.
We planned to examine the 4 most frequent types of cancer in France, separately: prostate cancer in men, breast cancer in women, smoking-related cancers (as defined by the French National Institute of Cancer, i.e., cancer of the oral cavity and pharynx, esophagus, larynx, trachea, bronchi and lungs, and bladder), and colorectal cancer. We also planned to examine a fifth category encompassing all other cancers. Nonmelanoma skin cancers and in situ neoplasms were not considered as cancer cases.
Information on age, sex, and occupational grade (blue-collar worker or clerk; first-line supervisor or sales representative; management) was obtained from company human resources records at the beginning of follow-up. Alcohol consumption, smoking, fruit and vegetable consumption (<1, 1–2, or >2 times per week), height, weight, physical activity (at least 1 time per week, occasionally, or none), and perceived health status were self-reported at the beginning of follow-up. Alcohol consumption, assessed as number of drinks per week, was categorized as follows: nondrinker, occasional drinker (1–13 drinks/week for men, 1–6 drinks/week for women), and moderate or heavy drinker (≥14 drinks/week for men, ≥7 drinks/week for women). Smoking was categorized into 5 classes: never smoker, ex-smoker of <20 pack-years, current smoker of <20 pack-years, ex-smoker of ≥20 pack-years, and current smoker of ≥20 pack-years. Body mass index was calculated by dividing weight in kilograms by height in meters squared and was categorized as <18.5, 18.5–24.9, 25–29.9, or ≥30. Perceived health status was reported on an 8-point Likert scale ranging from 1 ("very bad") to 8 ("very good").
All statistical analyses were conducted with PASW 18.0.0 software (SPSS Inc., Chicago, Illinois). All P values were 2-sided. Associations between depression measures as well as covariates and cancer incidence were estimated with hazard ratios and 95% confidence intervals computed in Cox regression analyses. For SAD and depressive symptoms measured in 1993, the follow-up period ran from January 1, 1994, to the date of cancer diagnosis, death, refusal to receive any further questionnaires, or December 31, 2009, whichever came first. For depressive symptoms measured between 1993 and 1999, the follow-up period began on January 1, 2000. For participants who reported a diagnosis of cancer after retirement but died from cancer before the diagnosis validation survey, the estimated date of diagnosis was the date of the first self-report minus 180 days (i.e., the mean interval between 2 annual questionnaires). Discrete covariates were considered as nominal variables. The presence of at least 1 SAD exceeding 7 days from 1989 to 1993 and the presence of chronic or recurrent depressive symptoms from 1993 to 1999 were considered as binary variables. Depressive symptoms measured in 1993 were considered either as a continuous variable (i.e., CES-D score) or as a binary variable (i.e., whether the CES-D score met the validated sex-specific cutoff). Since all covariates were associated with cancer incidence at 1 or more sites, they were all simultaneously entered into multivariate models.
Materials and Methods
Participants
Details on the GAZEL cohort are available elsewhere. The target population consisted of 44,922 employees of the French national gas and electricity company Electricité de France-Gaz de France: 31,411 men aged 40–50 years and 13,511 women aged 35–50 years. The study was approved by the French authority for data confidentiality (Commission Nationale Informatique et Liberté) and by the Ethics Evaluation Committee of the Institut National de la Santé et de la Recherche Médicale (INSERM). In 1989, a total of 20,625 employees (45.9%) (15,011 men and 5,614 women) gave written informed consent to participate. Since 1989, volunteers have been followed through annual mailed questionnaires and administrative databases.
Sickness Absence for Depression
Clinician-based diagnoses of depression were extracted from records on medically certified sickness absences. All sickness absences exceeding 7 days in a 4-year window from January 1, 1990, to December 31, 1993, were considered. This window was chosen to ensure homogeneity of the period during which exposure was measured across participants. Diagnoses were coded by company physicians using an abridged version of the International Classification of Diseases, Ninth Revision. Episodes of sickness absence for depression (SAD) corresponded to the codes indicating major depression or dysthymic disorder.
The CES-D
Depressive mood was assessed in 1993, 1996, and 1999 with the CES-D. This 20-item questionnaire has been designed for use in community studies and has a high internal consistency (α = 0.8 to α = 0.9 across samples). The CES-D asks participants how often they have experienced specific symptoms during the previous week (e.g., "I felt depressed"; "I felt everything I did was an effort"; "My sleep was restless"). Responses range from 0 ("hardly ever") to 3 ("most of the time"). Based on the validation of the French version, a global score of ≥17 among men and ≥23 among women may signal clinically significant depression. Since sole reliance on a binary cutoff may be statistically unsafe, we also considered CES-D score in 1993 as a continuous measure of depressive symptoms.
In order to replicate the findings of Penninx et al., we defined chronic or recurrent depressive symptoms as having CES-D scores above the validated cutoff for the years 1993, 1996, and 1999. If a subject was missing a CES-D score in one of the 3 time periods, the 2 remaining CES-D scores were used, and both values had to be above the validated cutoff.
Cancer Cases
All participants were followed up for diagnoses of primary cancers from January 1, 1994, to December 31, 2009. Diagnoses made during the period of employment came from a registry kept by the medical department at Electricité de France-Gaz de France that has been validated for accuracy and completeness. Diagnoses made after retirement came from systematic validation of each self-reported primary cancer through a diagnosis validation survey that began in 2009. Each annual questionnaire asked participants to report whether or not they had been hospitalized or diagnosed with any of several conditions, including cancer, in the preceding 12 months. All participants who self-reported cancer at least once during follow-up were contacted (if alive) and asked to give consent for a detailed diagnostic investigation with their physician.
In a first set of analyses, we considered as cases all participants with a validated diagnosis, as well as participants who reported a diagnosis of primary cancer but died from cancer before the initiation of the diagnosis validation survey. Information on vital status and date of death was obtained annually for all participants from the company, because it pays out retirement benefits. Cause-of-death data were available from baseline (i.e., January 1, 1994) to December 31, 2009, and were coded by the French national cause-of-death registry (Centre d'Épidémiologie sur les Causes Médicales de Décès, INSERM) using the Ninth and Tenth Revisions of the International Classification of Diseases.
We planned to examine the 4 most frequent types of cancer in France, separately: prostate cancer in men, breast cancer in women, smoking-related cancers (as defined by the French National Institute of Cancer, i.e., cancer of the oral cavity and pharynx, esophagus, larynx, trachea, bronchi and lungs, and bladder), and colorectal cancer. We also planned to examine a fifth category encompassing all other cancers. Nonmelanoma skin cancers and in situ neoplasms were not considered as cancer cases.
Covariates
Information on age, sex, and occupational grade (blue-collar worker or clerk; first-line supervisor or sales representative; management) was obtained from company human resources records at the beginning of follow-up. Alcohol consumption, smoking, fruit and vegetable consumption (<1, 1–2, or >2 times per week), height, weight, physical activity (at least 1 time per week, occasionally, or none), and perceived health status were self-reported at the beginning of follow-up. Alcohol consumption, assessed as number of drinks per week, was categorized as follows: nondrinker, occasional drinker (1–13 drinks/week for men, 1–6 drinks/week for women), and moderate or heavy drinker (≥14 drinks/week for men, ≥7 drinks/week for women). Smoking was categorized into 5 classes: never smoker, ex-smoker of <20 pack-years, current smoker of <20 pack-years, ex-smoker of ≥20 pack-years, and current smoker of ≥20 pack-years. Body mass index was calculated by dividing weight in kilograms by height in meters squared and was categorized as <18.5, 18.5–24.9, 25–29.9, or ≥30. Perceived health status was reported on an 8-point Likert scale ranging from 1 ("very bad") to 8 ("very good").
Statistical Analyses
All statistical analyses were conducted with PASW 18.0.0 software (SPSS Inc., Chicago, Illinois). All P values were 2-sided. Associations between depression measures as well as covariates and cancer incidence were estimated with hazard ratios and 95% confidence intervals computed in Cox regression analyses. For SAD and depressive symptoms measured in 1993, the follow-up period ran from January 1, 1994, to the date of cancer diagnosis, death, refusal to receive any further questionnaires, or December 31, 2009, whichever came first. For depressive symptoms measured between 1993 and 1999, the follow-up period began on January 1, 2000. For participants who reported a diagnosis of cancer after retirement but died from cancer before the diagnosis validation survey, the estimated date of diagnosis was the date of the first self-report minus 180 days (i.e., the mean interval between 2 annual questionnaires). Discrete covariates were considered as nominal variables. The presence of at least 1 SAD exceeding 7 days from 1989 to 1993 and the presence of chronic or recurrent depressive symptoms from 1993 to 1999 were considered as binary variables. Depressive symptoms measured in 1993 were considered either as a continuous variable (i.e., CES-D score) or as a binary variable (i.e., whether the CES-D score met the validated sex-specific cutoff). Since all covariates were associated with cancer incidence at 1 or more sites, they were all simultaneously entered into multivariate models.
Source...