Smoking and Prostate Cancer Survival and Recurrence

Context Studies of smoking in relation to prostate cancer mortality or recurrence in prostate cancer patients are limited, with few prostate cancer–specific outcomes.

Objective To assess the relation of cigarette smoking and smoking cessation with overall, prostate cancer–specific, and cardiovascular disease (CVD) mortality and biochemical recurrence among men with prostate cancer.

Design, Setting, and Participants Prospective observational study of 5366 men diagnosed with prostate cancer between 1986 and 2006 in the Health Professionals Follow-Up Study.

Main Outcome Measures Hazard ratios (HRs) for overall, prostate cancer–specific, and CVD mortality, and biochemical recurrence, defined by an increase in prostate-specific antigen (PSA) levels.

Results There were 1630 deaths, 524 (32%) due to prostate cancer and 416 (26%) to CVD, and 878 biochemical recurrences. Absolute crude rates for prostate cancer–specific death for never vs current smokers were 9.6 vs 15.3 per 1000 person-years; for all-cause mortality, the corresponding rates were 27.3 and 53.0 per 1000 person-years. In multivariable analysis, current vs never smokers had an increased risk of prostate cancer mortality (HR, 1.61; 95% confidence interval [CI], 1.11-2.32), as did current smokers with clinical stage T1 through T3 (HR, 1.80; 95% CI, 1.04-3.12). Current smokers also had increased risk of biochemical recurrence (HR, 1.61; 95% CI, 1.16-2.22), total mortality (HR, 2.28; 95% CI, 1.87-2.80), and CVD mortality (HR, 2.13; 95% CI, 1.39-3.26). After adjusting for clinical stage and grade (likely intermediates of the relation of smoking with prostate cancer recurrence and survival), current smokers had increased risk of prostate cancer mortality (HR, 1.38; 95% CI, 0.94-2.03), as did current smokers with clinical stage T1 through T3 (HR, 1.41; 95% CI, 0.80-2.49); they also had an increased risk of biochemical recurrence (HR, 1.47; 95% CI, 1.06-2.04). Greater number of pack-years was associated with significantly increased risk of prostate cancer mortality but not biochemical recurrence. Current smokers of 40 or more pack-years vs never smokers had increased prostate cancer mortality (HR, 1.82; 95% CI, 1.03-3.20) and biochemical recurrence (HR, 1.48; 95% CI, 0.88-2.48). Compared with current smokers, those who had quit smoking for 10 or more years (HR, 0.60; 95% CI, 0.42-0.87) or who have quit for less than 10 years but smoked less than 20 pack-years (HR, 0.64; 95% CI, 0.28-1.45) had prostate cancer mortality risks similar to never smokers (HR, 0.61; 95% CI, 0.42-0.88).

Conclusions Smoking at the time of prostate cancer diagnosis is associated with increased overall and CVD mortality and prostate cancer–specific mortality and recurrence. Men who have quit for at least 10 years have prostate cancer–specific mortality risks similar to those who have never smoked.

Accumulating evidence suggests that smoking may increase risk of aggressive prostate cancer and prostate cancer mortality. The latest review by the US surgeon general found the evidence “probable” that smoking contributes to a higher prostate cancer mortality rate,¹ in agreement with a review of the literature in which we reported an approximately 30% increase in risk of fatal prostate cancer when comparing current with never smokers.² Several studies reported that smoking is associated with more aggressive disease at diagnosis, defined as a higher stage or tumor grade^3-5 and the relation between smoking and disease progression after diagnosis, defined as biochemical recurrence,^5-7 metastasis,⁸ and hormone-refractory prostate cancer,⁹ is suggestively positive. However, these studies had few prostate cancer–specific deaths and either observed no clear association with prostate cancer mortality^5,8 or did not examine this outcome.^6,9 Three studies reported a positive association between smoking and prostate cancer mortality but were based on 57 and 54 prostate cancer deaths^3,10 or an unspecified number of deaths in a single-institution study of 214 patients.¹¹ Moreover, concern remains that some or all of the observed associations may be due to delayed diagnosis and treatment among smokers. With 8 years of follow-up in the Health Professionals Follow-Up Study, we previously provided preliminary data on smoking status in relation to prostate cancer mortality.¹² With 22 years of follow-up and a large number of outcomes, we now can examine in detail the relation of current and former smoking to overall and prostate cancer–specific mortality and recurrence in a nationwide cohort of prostate cancer patients.

The Health Professionals Follow-Up Study is a prospective cohort study of 51 529 US male health professionals who enrolled in 1986 by completing a mailed questionnaire. Participants provided information about medical history and risk factors for chronic diseases, including cancer. Participants complete biennial follow-up questionnaires to collect information on new medical diagnoses and to update information on lifestyle factors (response rate 96%). This study was approved by the institutional review board of the Harvard School of Public Health; participants provided implied consent by virtue of returning their questionnaires and written informed consent for review of medical records.

Current smoking status (cigarettes smoked per day) was assessed every 2 years, beginning in 1986. At baseline, we also inquired about past smoking, time since quitting, and the average number of cigarettes smoked per day before age 15 years, ages 15 through 19, 20 through 29, 30 through 39, 40 through 49, 50 through 59, and 60 years and older. Pack-years was calculated as years of smoking multiplied by the average number of packs smoked per day. A pack contains 20 cigarettes. All smoking variables were updated every 2 years until the questionnaire just before the report of prostate cancer diagnosis. We focused on smoking prior to prostate cancer diagnosis because the majority of ever smokers at diagnosis were former smokers, with only 5.2% of men reporting current smoking.

After a participant reported a diagnosis of prostate cancer, medical records and pathology reports were sought from treating physicians and hospitals to confirm the diagnosis and obtain information on pathology, treatments, prostate-specific antigen (PSA) values at diagnosis, increases in PSA after treatment (PSA or biochemical recurrence), and metastasis. Participants completed biennial follow-up questionnaires to update data on treatments, PSA, and clinical progression. The primary outcomes for this analysis were prostate cancer mortality and biochemical recurrence. Other outcomes of interest were total and cardiovascular (CVD) mortality. Biochemical recurrence was defined from medical records and physician questionnaires based on the primary treatment, using standard definitions: for radical prostatectomy,^13,14 PSA higher than 0.2 ng/mL after surgery and for at least 2 consecutive measures; for radiation,¹⁵ an increase of 2 or more ng/mL higher than the nadir PSA; for brachytherapy,¹⁶ hormones, or other treatments, an increase of 1 or more ng/mL higher than the nadir PSA for at least 2 consecutive measures; for watchful waiting, a postdiagnosis PSA increase of 1 or more ng/mL for at least 2 consecutive measures. We also used patient-reported PSA increase from participants' questionnaires that comprised 37% of all recorded increases in PSA values. Date of failure was the date of first increase. Men for whom we could not ascertain a PSA recurrence but who reported metastasis or died of prostate cancer were assigned a date of biochemical recurrence as the earliest date for any of these events. Using reports of deaths from families and the National Death Index for nonrespondents, we ascertained more than 98% of deaths.¹⁷ Causes of death were centrally adjudicated by study physicians who reviewed medical records and death certificates without knowledge of participants' smoking status.

We included men who in 1986 were free of a cancer diagnosis, except nonmelanoma skin cancer, and who had provided information on their smoking status before diagnosis (n = 5366). For the mortality analyses, we included all participants regardless of their stage at diagnosis. We also performed an additional analysis among men with clinical stage T1 through T3. For the analyses of prostate cancer recurrence, we excluded men with metastatic disease at diagnosis (n = 220) and those who reported either a PSA increase or metastasis after diagnosis but did not provide a date for this event (n = 5). Additionally, 1508 participants without data on recurrence or progression after diagnosis were excluded.

We used multinomial logistic regression models, adjusting for age at diagnosis in 5-year age categories, to test whether there were differences in clinical stage and clinical Gleason score by smoking status at diagnosis. We used Cox proportional hazards regression models to calculate hazard ratios (HRs) of death from any cause, prostate cancer death, CVD death, and biochemical recurrence. In the main analysis for prostate cancer and CVD mortality, deaths from other causes were censored, and Cox survival analysis was used. Person-years were calculated beginning at diagnosis until death or end of follow-up (January 1, 2008), whichever came first, for mortality analyses, and until the earliest of the following events: progression or mortality or end of follow-up, for the recurrence analysis. We used the participant's most recent smoking status reported prior to his diagnosis and classified current smokers by categories of pack-years (<40 and ≥40), in addition to cigarettes smoked per day, and past smokers by a combination of pack-years (<20 and ≥20 pack-years), and time since quitting (<10 and ≥10 years).

Our final models for prostate cancer mortality and recurrence included age at diagnosis (years), previous PSA screening history (yes, no, unknown), prediagnosis vigorous activity (<1, 1-3, and ≥3 h/wk), body mass index (BMI, calculated as weight in kilograms divided by height in meters squared, categories <25, 25-<30, ≥30), energy (quartiles), and coffee intake (quartiles). Coffee intake was added to the models based on recent findings from our study reporting a lower risk of advanced prostate cancer with increased coffee consumption.¹⁸ We used months since diagnosis as the time scale and stratified by calendar time in 2-year intervals. We considered models adjusted for race; height; family history of prostate cancer; parental history of myocardial infarction (MI) at age 60 years or younger; diabetes; elevated cholesterol; elevated blood pressure; and intakes of calcium, saturated fat, cholesterol, red meat, tomato sauce, fish, and α-linolenic acid because most of these factors were previously associated with prostate cancer incidence or progression in our study^19,20 or could be potential confounders. We considered similar models for overall and CVD mortality. There was little evidence of confounding by these factors, so they were not included in our final models. We hypothesized that smoking may affect prostate cancer mortality risk by promoting more aggressive tumors, characterized by poorer differentiation (higher Gleason score) or more advanced stage at diagnosis. Because we considered stage and grade as likely intermediates of the relation of smoking with biochemical recurrence and prostate cancer–specific mortality, we did not initially adjust for these factors in our primary analysis. However, we considered these factors in a secondary analysis.

We conducted 2 sensitivity analyses to evaluate potential bias from any difference in screening behavior between smokers and nonsmokers. The first analysis included only men who reported having a PSA screening history in the cycle prior to diagnosis, starting with men diagnosed from 1988. The second analysis included only men diagnosed from 1994, after PSA screening had become well established. In that analysis, we further adjusted for screening intensity as reflected in the proportion of 2-year periods in which a participant reported at least 1 PSA screen, dichotomizing at 50%.

We assessed the interaction between smoking and vitamin E supplement use and BMI by entering cross products of smoking status with those variables in models. We chose these 2 covariates a priori because previous studies have reported a potential interaction of smoking and vitamin E supplements for prostate cancer incidence,^21,22 and higher prediagnostic BMI is related to increased prostate cancer mortality.²³ We also evaluated the effect of smoking among men treated by radical prostatectomy and radiation as primary treatments. All P values were 2-sided; P < .05 was considered statistically significant. All analyses were conducted using SAS version 9.1 (SAS Institute Inc, Cary, North Carolina).

Among 5366 men diagnosed with prostate cancer, we documented 1630 deaths, 524 (32%) due to prostate cancer and 416 (26%) to CVD, and 878 biochemical recurrences. The other causes of death included infectious disease (0.5%), other cancer (19.5%), metabolic or endocrine disorders (1%), neurological disease (6%), respiratory disease (5%), digestive tract disease (1%), kidney failure (1%), other causes (5%), and unknown (2%). In the mortality analysis, which included all cases including those advanced at diagnosis, the median duration of follow-up from diagnosis to censoring (either until death or the end of follow-up in January 2008) was 8.1 years (25th and 75th percentiles, 5.0 and 12.2 years) for survivors and 6.5 years (25th and 75th percentiles, 3.5 and 10.2 years) for men who died. In the recurrence analysis, which excluded men with metastatic disease at diagnosis, the median duration of follow-up from diagnosis to censoring (recurrence, death, or end of follow-up) was 3.8 years (25th and 75th percentiles, 2.0 and 6.6 years) for those with recurrence and 8.1 years (25th and 75th percentiles, 4.8 and 12.2 years) for those without. Age-standardized characteristics after diagnosis are presented in Table 1. Compared with never smokers, current smokers had higher clinical stage and grade (P < .001 for both). For example, 14.7% of current smokers had stage T3 or higher at diagnosis compared with 8.3% of never smokers, and 16.0% of current smokers had a Gleason score of 7 or more compared with 10.7% of never smokers. Current smokers exercised less, drank more coffee, and had a higher intake of saturated fat and lower intake of calcium compared with never and former smokers. Former and current smokers consumed more alcohol but had similar BMIs compared with never smokers. Current smokers tended to have less intense PSA testing, but the rates among past smokers of 10 years were similar to never smokers.

Proportions of men free of prostate cancer cancer–specific death at 5 years (Figure 1) were 94.8% for never smokers, 94.4% for former smokers who had quit for more than 10 years, 91.6% for former smokers quitting less than 10 years, and 91.7% for current smokers (P = .02, log-rank test). At 10 years, the proportions free of prostate cancer–specific death were 89.8% for never smokers, 89.1% for those who had quit for more than 10 years, 85.7% for those who quit for less than 10 years, and 82.7% for current smokers. The absolute crude rates per 1000 person-years for prostate cancer–specific death were 9.6 for never smokers, 10. 3 for former smokers who had quit for more than 10 years, 13.8 for former smokers who had quit for less than 10 years, and 15.3 for current smokers (P < .001, χ² test). A similar pattern was observed for prostate cancer recurrence: 26.4 for never smokers, 28.1 for former smokers who had quit for more than 10 years, 34.6 for former smokers who had quit for less than 10 years, and 38.2 for current smokers per 1000 person-years (P = .006, χ² test). We observed a statistically significant difference in overall survival across smoking status (P < .001, Figure 2). Proportions of men alive at 5 years were 89.7% for never smokers, 86.2% for former smokers, and 78.8% for current smokers; at 10 years, the proportions alive were 74.8% for never smokers, 68.2% for former smokers, and 54.8% for current smokers. Absolute crude rates per 1000 person-years for all-cause mortality were 27.3 for never smokers, 35.2 for former smokers, and 53.0 for current smokers (P < .001, χ² test).

Compared with never smokers, current smokers had an increased risk of dying from prostate cancer, CVD, and all-cause mortality and an increased risk of biochemical recurrence. In multivariable models, the HRs for current smoking for these outcomes were 1.61 (95% confidence interval [CI], 1.11-2.32) for prostate cancer mortality and 1.80 (95% CI, 1.04-3.12) for men with clinical stage T1 to T3, 1.61 (95% CI, 1.16-2.22) for biochemical recurrence, 2.28 (95% CI, 1.87-2.80) for total mortality, and 2.13 (95% CI, 1.39-3.26) for CVD mortality (Table 2). When limiting the biochemical recurrence analysis to men treated with radical prostatectomy, external beam radiation, or brachytherapy, current smokers had an HR for biochemical recurrence of 1.63 (95% CI, 1.15-2.31). Because smoking likely acts to increase prostate cancer mortality through changes in grade and stage, we did not adjust for those mediating factors in our primary analysis. In secondary analyses, we adjusted for stage and grade to address the effect of smoking apart from its effect on those mediating factors; the HRs for current smoking were 1.38 (95% CI, 0.94-2.03) for prostate cancer mortality, 1.41 (95% CI, 0.80-2.49) for men with clinical stage T1 to T3, 1.47 (95% CI, 1.06-2.04) for biochemical recurrence, 2.01 (95% CI, 1.64-2.47) for total mortality, and 1.98 (95% CI, 1.29-3.04) for CVD mortality. Among men treated with radical prostatectomy, external beam radiation, or brachytherapy, current smokers had an HR for biochemical recurrence of 1.48 (95% CI, 1.04-2.10). A greater number of pack-years was associated with an increased risk of prostate cancer mortality, CVD mortality, and total mortality but not biochemical recurrence (Table 2). Among current smokers, we found no association for smoking dose (cigarettes per day) for prostate cancer mortality (P for trend = .85) and biochemical recurrence (P for trend = .71). The HRs for prostate cancer mortality were 1.69 (95% CI, 0.99-2.91) for those who smoked 1 to 14 cigarettes a day, 1.32 (95% CI, 0.73-2.37) for between 15 and 24 cigarettes a day, and 1.67 (95% CI, 0.78-3.61) for 25 or more cigarettes a day. For biochemical recurrence, the HRs were 1.73 (95% CI, 1.09-2.76) for those who smoked 1 to 14 cigarettes a day, 1.78 (95% CI, 1.03-3.06) for between 15 and 24 cigarettes a day, and 1.48 (95% CI, 0.77-2.85) for 25 or more cigarettes a day.

Adjustment for PSA screening attenuated the estimates for current smoking and prostate cancer mortality and minimally attenuated prostate cancer recurrence (Table 2). In the first sensitivity analysis limited to men who had reported PSA screening in the cycle prior to being diagnosed with prostate cancer, the estimates for current smoking became even stronger, although the CIs were wider due to a smaller number of outcomes. The HRs were 2.13 (95% CI, 0.97-4.68) for prostate cancer mortality and 2.06 (95% CI, 1.25-3.42) for biochemical recurrence. In the second sensitivity analysis that included cases only diagnosed from 1994, with adjustment for PSA screening intensity, the HRs were similar to the first sensitivity analysis: 2.12 (95% CI, 1.18-3.79) for prostate cancer mortality and 2.02 (95% CI, 1.30-3.13) for biochemical recurrence.

We observed no statistically significant interaction between smoking status and vitamin E supplementation (P = .16) or BMI (P = .26) for prostate cancer mortality nor for smoking status and vitamin E supplementation (P = .53) or BMI (P = .60) for biochemical recurrence. Lastly, we also assessed the relation between current smoking and prostate cancer mortality and biochemical recurrence among men having radical prostatectomy or radiation as their primary treatment. After adjusting for stage and grade, the HRs for prostate cancer mortality were 1.30 (95% CI, 0.55-3.06) for patients with radical prostatectomy and 2.07 (95% CI, 0.60-7.08) for patients who had radiation. The HRs for recurrence were 1.68 (95% CI, 1.02-2.76) for patients who had radical prostatectomy and 1.33 (95% CI, 0.70-2.52) for patients who had radiation.

Former smokers overall, which includes past smokers who have quit for less than 10 years before diagnosis, had an HR for total mortality of 1.23 (95% CI, 1.10-1.37). Former smokers overall were not at significantly increased risk of the other end points (Table 2). Compared with current smokers, men who had quit smoking for 10 or more years or who had quit for less than 10 years but smoked less than 20 pack-years had prostate cancer mortality risks similar to those who had never smoked (Figure 3). The HRs were 0.60 (95% CI, 0.42-0.87) for those who had quit for 10 or more years; 0.60 (95% CI, 0.40-0.88) for those who had quit for 10 or more years and smoked less than 20 pack-years, 0.61 (95% CI, 0.40-0.92) for those who had quit for 10 or more years and smoked more than 20 pack-years, 0.64 (95% CI, 0.28-1.45) for those who had quit for less than 10 years and smoked fewer than 20 pack-years, and 0.61 (95% CI, 0.42-0.88) for those who had never smoked. Those who had quit for less than 10 years and smoked 20 or more pack-years had risks similar to the reference group of current smokers (HR, 1.05; 95% CI, 0.64-1.73). The estimates for recurrence were similar for former smokers quitting more than 10 years before diagnosis, but we observed no reduction in risk among former smokers quitting less than 10 years before diagnosis (Figure 3).

We observed an elevated risk of prostate cancer mortality, CVD mortality, total mortality, and biochemical recurrence among men who were current smokers at the time of their prostate cancer diagnosis. Former smokers who had quit 10 or more years prior to diagnosis had risks for prostate cancer mortality and recurrence similar to those who had never smoked. Those who had quit for less than 10 years prior to diagnosis and with less than 20 pack-years had risks similar to current smokers for prostate cancer recurrence (an early progression event) but similar to never smokers for prostate cancer mortality. In our assessment of the impact of smoking beyond its effect on stage and grade, which we considered intermediates of the relation between smoking and biochemical recurrence and prostate cancer–specific mortality, these associations remained elevated but were attenuated, as expected, providing evidence that the effect of smoking is mediated by these factors. Nevertheless, even after adjustment for stage and grade, the estimate for biochemical recurrence remained significant.

Smokers tend to have less PSA testing²⁴ and may be diagnosed at a more advanced stage; hence, smokers may be at increased risk of dying of prostate cancer or of having a biochemical recurrence due to the differences in screening behavior rather than the smoking per se. However, differential PSA screening across strata of smoking status was unlikely to fully account for our results, because the percentage of men who had at least 1 PSA test before the cycle of their diagnosis varied little among never smokers (86%), former smokers (85%), and current smokers (75%) when considering cases diagnosed after 1994, when screening became widely used (Table 1). We adjusted for PSA screening history in the multivariable analyses, which did not materially affect the estimates. We also conducted 2 sensitivity analyses, one limited to men who had reported PSA testing in a cycle prior to their diagnosis and the other adjusted for PSA screening intensity. If the association between smoking and fatal prostate cancer and recurrence resulted from delayed diagnosis, we would have expected to see an attenuation of the association. Instead, we observed stronger associations. These analyses were limited to cases diagnosed after introduction of PSA screening, so they included fewer cases and fewer end points, leading to wider CIs. However, the relative risk estimates were larger than in the main analysis, suggesting that our primary analyses were not largely affected by this potential bias.

Several studies in specific treatment populations report that outcomes in men treated with external beam radiation therapy,^5,8 androgen deprivation therapy,⁹ and radical prostatectomy^6,7 are poorer among current smokers, in univariate^5-9 as well as multivariable^5,7-9 models adjusted for other clinical factors. One other study involving patients who had undergone brachytherapy noted a nonsignificant trend for poorer outcome in current smokers vs never or former smokers (P = .13).²⁵ We found that the association of smoking with prostate cancer recurrence and mortality may differ in men treated with radical prostatectomy or radiation, the 2 most common treatments in our cohort, after adjustment for stage and grade, but the CIs were wide, so these findings should be interpreted cautiously.

A direct effect of smoking on prostate cancer progression is biologically plausible. The main hypotheses proposed include: (1) tumor promotion through carcinogens from tobacco smoke,²⁶ with suggestive prostate cancer–specific data for nitrosamines²⁷ and cadmium^28-32 and one study reporting that gene variants involved in detoxification may influence risk of aggressive prostate cancer³³; (2) increased plasma levels of total and free testosterone, an androgen involved in the development and progression of prostate cancer, in current smokers,^34-38 with some studies reporting a dose-dependent association^36,37; (3) epigenetic effects, including aberrant methylation profiles among current smokers, which correlate with aggressive disease³⁹; and (4) nicotine-induced angiogenesis, capillary growth, and tumor growth and proliferation.^40-42

In summary, smoking at the time of diagnosis was associated with substantially increased overall mortality and prostate cancer mortality and recurrence. Ten-year quitters had risks similar to never smokers. These results provide further support that smoking may increase risk of death from prostate cancer.

Corresponding Author: Stacey A. Kenfield, ScD, Channing Laboratory, 181 Longwood Ave, Room 452, Boston, MA 02115 (skenfiel@hsph.harvard.edu).

Author Contributions: Drs Kenfield and Giovannucci had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Kenfield, Stampfer, Chan, Giovannucci.

Acquisition of data: Kenfield, Stampfer, Chan, Giovannucci.

Analysis and interpretation of data: Kenfield, Stampfer, Chan, Giovannucci.

Drafting of the manuscript: Kenfield.

Critical revision of the manuscript for important intellectual content: Kenfield, Stampfer, Chan, Giovannucci.

Statistical analysis: Kenfield, Giovannucci.

Obtained funding: Stampfer, Giovannucci.

Administrative, technical, or material support: Stampfer, Chan, Giovannucci.

Study supervision: Stampfer, Chan, Giovannucci.

Funding/Support: This work was supported by grants CA055075 and CA141298 from the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Role of the Sponsor: The funding source had no role in the design or conduct of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, or approval of the manuscript.

Additional Contributions: We thank the participants and staff of the Health Professionals Follow-Up Study for their valuable contributions. We thank the following state cancer registries for their help: Alabama, Arizona, Arkansas, California, Colorado, Connecticut, Delaware, Florida, Georgia, Idaho, Illinois, Indiana, Iowa, Kentucky, Louisiana, Maine, Maryland, Massachusetts, Minnesota, Nebraska, New Hampshire, New Jersey, New York, North Carolina, North Dakota, Ohio, Oklahoma, Oregon, Pennsylvania, Rhode Island, South Carolina, Tennessee, Texas, Virginia, Washington, Wyoming.