Coffee Consumption and Prostate Cancer Risk and Progression in the Health Professionals Follow-up Study

Kathryn M. Wilson; Julie L. Kasperzyk; Jennifer R. Rider; Stacey Kenfield; Rob M. van Dam; Meir J. Stampfer; Edward Giovannucci; Lorelei A. Mucci


J Natl Cancer Inst. 2011;103(11):876-884. 

In This Article


The Health Professionals Follow-up Study is a prospective cohort study of 51 529 male health professionals in the United States aged 40–75 years at baseline in 1986. The men are followed through biennial questionnaires to update information on lifestyle and health outcomes, and usual diet has been assessed every 4 years.

Men who completed the baseline food frequency questionnaire (FFQ) in 1986 form the study population for this analysis (N = 49 911). We excluded men who had an implausible energy intake (<800 or >4200 kcal/day) or who left more than 70 food items blank on the baseline FFQ. We also excluded men who reported a diagnosis of cancer (except nonmelanoma skin cancer) before baseline (N = 2000). This left a total of 47 911 men who were followed prospectively for cancer incidence until 2006 and for metastases and mortality outcomes until 2008. The Health Professionals Follow-up Study is approved by the Human Subjects Committee at the Harvard School of Public Health.

Assessment of Coffee Intake

Updated dietary data, including coffee consumption, was available from FFQs, which reported on intake of over 130 food items at baseline in 1986, and again in 1990, 1994, 1998, and 2002. Participants were asked how frequently they had consumed a specified portion size of each item over the previous year, with nine possible responses ranging from "never or less than once a month" to "six or more times per day." The FFQ included questions concerning cups of decaffeinated and regular coffee intake. A validation study in this cohort found a high correlation (r = 0.93) between participants' reports of coffee intake on the FFQ compared with two week-long diet records.[33]

Ascertainment and Classification of Subjects Who Developed Prostate Cancer

Prostate cancer diagnoses were initially identified by self-reports from the participants or their next of kin on the biennial questionnaires and then confirmed by review of medical records and pathology reports. Deaths in the cohort were ascertained through reports from family members and searches of the National Death Index. Underlying cause of death was assigned by an endpoints committee based on all available data including medical history, medical records, registry information, and death certificates. Approximately 90% of prostate cancer patients were documented by medical records; the remaining 10% of men with prostate cancer, based on self-reports or death certificates, were included because the reporting of prostate cancer was highly accurate (>98%) among men with available medical records. We followed men with prostate cancer starting in 2000 with an additional prostate cancer–specific questionnaire separate from the regular Health Professionals Follow-up Study questionnaire every year to ascertain disease progression and diagnosis of metastases.

We studied total prostate cancer incidence excluding stage T1a cancers, which are discovered incidentally during treatment for benign prostatic hypertrophy. Because of the considerable heterogeneity in the biological potential of prostate cancer, we also examined the data for men with advanced, lethal, or nonadvanced cancers separately to distinguish those patients in whom the cancer was likely to progress clinically. Advanced cancers were those that had spread beyond the prostate, including to the seminal vesicle, lymph nodes, or bone. This category included men with stage T3b, T4, N1, or M1 prostate cancer at diagnosis, men who developed lymph node or distant metastases, and men who died of prostate cancer before the end of follow-up. Lethal cancers, a subset of advanced cancers, were those that caused death or metastasis to bone before the end of follow-up. Nonadvanced cancers were stage T1 or T2 and N0 and M0 at diagnosis and did not progress to lymph node or distant metastases or death during the follow-up period. (Some cancers that were diagnosed near the end of the follow-up period will be misclassified as nonadvanced because they had less time to progress before the end of follow-up). Cancers were also categorized as high grade (Gleason sum at diagnosis 8–10), grade 7, or low grade (Gleason sum 2–6) at diagnosis based on prostatectomy or biopsy pathology reports; Gleason grade was not available for all men with prostate cancer, particularly for those who were diagnosed earlier in the follow-up period.

Statistical Analysis

Each participant contributed person-time from the date on which he returned the baseline questionnaire in 1986 until prostate cancer diagnosis, death, or the end of the study period, January 31, 2006. Participants were followed for prostate cancer incidence until January 31, 2006, and for death and metastases until January 31, 2008. Participants' data were divided according to levels of total (regular and decaffeinated) coffee intake, and relative risks of prostate cancer were calculated as the incidence rate in a given category of intake divided by the rate in the lowest category, adjusted for age and calendar time.

Because coffee intake may affect carcinogenesis over an extended period, we used the cumulative average intake of coffee to represent long-term dietary intake as our primary measure of exposure. That is, the coffee intake reported by particpants in 1986 was used to compute exposure for the 1986–1990 follow-up period, the average of the intakes reported in 1986 and 1990 was used for the 1990–1994 follow-up period, the average of intakes reported in 1986, 1990, and 1994 was used for the 1994–1998 follow-up period, and so on. In a secondary analysis, we used baseline (1986) coffee intake only. In addition, we used our repeated measures to analyze the effect of latency time (time from exposure to cancer diagnosis) by relating each measure of coffee intake to prostate cancer incidence during specific time periods: 0–4, 4–8, 8–12, and 12–16 years after exposure. Finally, to assess the potential for symptoms of subclinical advanced disease to affect coffee intake (reverse causation), we conducted a secondary analysis using cumulative average intake with a lag of 4 years to avoid using data on coffee consumption from FFQs completed immediately before diagnosis.

We used Cox proportional hazards regression to adjust for potential confounding by prostate cancer risk factors previously identified in this cohort and in other studies. Scaled Schoenfeld residuals were used to test the proportional hazards assumption. Multivariable models were adjusted for race (White, African American, Asian American, other), height (quartiles), body mass index at age 21 (<20, 20 to <22.5, 22.5 to <25, ≥25 kg/m2), current body mass index (<21, 21 to <23, 23 to <25, 25 to <27.5, 27.5 to <30, ≥30 kg/m2), vigorous physical activity (quintiles, metabolic equivalents-hours/week), smoking (never, former quit >10 years ago, former quit <10 years ago, current), diabetes (type I or II, yes or no), family history of prostate cancer in father or brother (yes or no), multivitamin use (yes or no), history of prostate-specific antigen (PSA) testing (yes or no, lagged by one period to avoid counting diagnostic PSA tests as screening; collected from 1994 onwards), and intakes of processed meat, tomato sauce, calcium, alpha linolenic acid, supplemental vitamin E, alcohol intake (all quintiles), and energy intake (continuous). All covariates except race, height, and body mass index at age 21 were updated in each questionnaire cycle. To test for a linear trend across categories of intake, we modeled coffee intake as a continuous variable using the median intake for each category.

Because we found associations for total coffee intake, we repeated our analyses for regular and decaffeinated coffee separately, and for caffeine intake to see if the observed associations were related to caffeine or other components of coffee. To investigate possible confounding due to differences in PSA testing, we stratified by time period to determine whether the association between coffee and prostate cancer risk differed in the pre-PSA (1986–1994) and PSA screening eras (1994–2006). All P values were two-sided, with a P value less than .05 considered to be statistically significant. Analyses were performed using SAS version 9.1 (SAS Institute, Inc; Cary, NC).


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: