A Prospective Study of Chronic Inflammation in Benign Prostate Tissue and Risk of Prostate Cancer: Linked PCPT and SELECT Cohorts

This article is featured in Highlights of This Issue, p. 1475

Conditions that cause continued bouts of epithelial cell injury and cell death, such as chronic infections and chronic inflammatory conditions, result in a sustained abnormal increase in epithelial cell proliferation that can enhance carcinoma development (1–3). These scenarios are established for gastric carcinomas induced by Helicobacter pylori, colitis-associated colorectal carcinomas, and viral and other chronic hepatitis-induced hepatocellular carcinomas (2, 4). Such a link between inflammation and prostate cancer has been proposed, but not yet established (5).

Currently, we do not have evidence-based diet, lifestyle, or FDA-approved chemoprevention approaches to prevent prostate cancer, and do not have tools to readily identify men who would most benefit from preventive interventions. If a link can be established between inflammation and prostate cancer, then the development of novel interventions to abrogate intraprostatic inflammation (or its causes) could be a viable primary prevention strategy.

Previously in the Prostate Cancer Prevention Trial (PCPT) placebo arm, we assessed inflammation in benign areas of prostate biopsy cores from 191 cases and 209 controls (6). Men with inflammation in at least one of three biopsy cores assessed had 1.8 times the odds of prostate cancer and 2.4 times the odds of Gleason sum ≥7 disease compared with men for whom none of their cores had inflammation. The odds increased with proportion of cores with inflammation. An association with Gleason sum <7 disease was also present, but weaker. Prevalence of inflammation was higher in the finasteride arm (92.4%; ref. 7) than in the placebo arm (78.2%; ref. 6), and in contrast to the placebo arm, inflammation was not associated with prostate cancer in the finasteride arm (7). Because of a design feature of the PCPT—prostate biopsies were performed at the end of the trial including on men without indication—our studies had clear strengths: (i) the likelihood of occult prostate cancer was reduced in controls, minimizing attenuation of associations resulting from admixing cases with controls (8); and (ii) inflammation in controls was less likely to be distorted by a common indication for biopsy—elevated PSA—which is higher in men with intraprostatic inflammation, including in the PCPT placebo arm (6, 9).

For inflammation to be a cause, it must temporally precede disease. While our prior results suggest that inflammation influences prostate cancer (in the absence of finasteride), we could not determine whether inflammation preceded prostate cancer because we assessed inflammation in the biopsies used to rule in or out a prostate cancer diagnosis. Thus, the most critical step in establishing the inflammation–prostate cancer association is a prospective study.

Therefore, in this study we linked the PCPT cohort and the Selenium and Vitamin E Cancer Prevention Trial (SELECT) cohort to form a cohort of men who were negative for prostate cancer on PCPT end-of-study biopsies and were eligible, specifically having PSA ≤4 and a normal DRE, and participated in SELECT. We used PCPT end-of-study biopsies as “baseline” biopsies in the linked cohort. Linking these cohorts is the only epidemiologically sound approach to prospectively test the inflammation–prostate cancer association in men without an indication for biopsy now and in the foreseeable future. We hypothesized that men with more inflammation in their baseline biopsies would have a greater prostate cancer risk. We also sought to determine if we could confirm, in a prospective study, our prior observation (7) that inflammation is not associated with risk in men previously treated with finasteride.

Men who participated in both of two randomized, multisite chemoprevention trials conducted by Southwest Oncology Group—PCPT and SELECT—were eligible for this study (Fig. 1).

Starting in 1993, PCPT randomized 18,882 men ≥55 years old with a normal DRE and PSA ≤3.0 ng/mL to 5 mg/day finasteride (5α-reductase type II inhibitor) or placebo for 7 years; 3.8% were African American. Because finasteride reduces serum PSA, values in the finasteride arm were adjusted upward in post-baseline samples to keep the proportion of biopsies triggered by PSA screening comparable between arms. All men not diagnosed with prostate cancer during the trial were recommended to undergo biopsy at the end, even if PSA and DRE results were normal, to determine prostate cancer status. If during the trial a man had an elevated PSA, was biopsied, and cancer was not found, then a subsequent PSA had to have an increase of ≥50% the prior level or be >10 ng/mL to be considered as an indication for biopsy, including at the end of the trial. In total, 7,761 men without clinical indication underwent an end-of-study biopsy within 7 years ± 90 days of randomization, and no cancer was found. PCPT found that finasteride reduced the period prevalence of prostate cancer by 25% (10).

Starting in 2001, SELECT randomized 35,533 men ages ≥50 years old, if African American, or ≥55 years old otherwise, with a PSA ≤4.0 ng/mL, normal DRE, and no other cancer diagnoses within the preceding 5 years to receive supplements containing selenium, vitamin E, selenium and vitamin E, or placebo for 7 years. PCPT participants were eligible to subsequently join SELECT if they met the eligibility criteria. Twelve percent were African American. Participants were followed by the local study site for prostate cancer for a median of 5.5 years using community standards for PSA and DRE to recommend prostate biopsy. Diagnosis and Gleason sum were centrally reviewed for 85% of cases. SELECT was stopped early based on interim results showing that selenium and vitamin E supplementation did not decrease prostate cancer risk (11). With 18 months of extended follow-up, risk was 17% higher in the vitamin E arm compared with the placebo arm (12).

The linked cohort consists of men negative for prostate cancer on PCPT end-of-study biopsies and who subsequently enrolled in SELECT. We used PCPT end-of-study biopsies to estimate the prevalence and extent of inflammation at entry into SELECT. By 2011, when we designed this inflammation study, 100 biopsy-confirmed incident prostate cancer cases were diagnosed in 1,780 men in the linked cohort (51% from the finasteride arm). Given SELECT's eligibility criteria, all men in the linked cohort had no clinical indication for biopsy at the start of follow-up. To form a nested case–control study, we selected all 100 incident cases, and sampled 200 controls frequency matched on age at PCPT end-of-study biopsy (</≥ median of 69.3 years), race (white/nonwhite), PCPT arm (finasteride/placebo), and SELECT arm (vitamin E, selenium, both, placebo). Most cases were early stage. For this inflammation study, we classified cases as Gleason sum <4+3 (grade groups 1–2 in the Epstein and colleagues prostate cancer grading system; ref. 13) or ≥4+3 (grade groups 3–5). Although at entry into SELECT none of the men had an indication for biopsy, five had a PSA >4 ng/mL at the PCPT end-of-study biopsy. For this reason, we excluded one control and two cases among men previously in the PCPT placebo arm and two cases among men previously in the PCPT finasteride arm.

The institutional review boards (IRBs) at the participating sites approved PCPT and SELECT. The Johns Hopkins Bloomberg School of Public Health and Department of Defense IRBs approved this linked cohort and research on inflammation and prostate cancer.

PCPT end-of-study biopsies were stored at the University of Colorado Prostate Biorepository. Tissue was located for 92 cases (96%) and 193 controls (97%). Joint distribution of PCPT and SELECT treatment arms among cases and controls is shown in Supplementary Table S1; 81% of men had 6 cores and 13.5% had 7–10 cores taken during the prostate biopsy. As we did previously (6), for feasibility we sampled two slides per man (96%), Slides had multiple cores mounted, yielding a mean of 4.6 cores in controls and 4.9 cores in cases.

Images were visually assessed using the Aperio ImageScope Viewer Software package to determine presence of total inflammatory cells, acute inflammatory cells (e.g., polymorphonuclear cells), and chronic inflammatory cells (e.g., cells with an appearance consistent with that of lymphocytes and macrophages), and quantify the percentage of the biopsy core tissue area per slide that had involvement of any inflammatory cells (6). The pathologist (I. Kulac) trained to score inflammation was blinded to case–control status.

We used data collected at SELECT entry to characterize cases and controls by risk and other factors, including family history, body mass index (BMI), smoking status, and energy intake. We used serum PSA concentration measured closest in time before the PCPT end-of-study biopsy. For men previously in the finasteride arm, we used data on continued use of the study drug at the end-of-study biopsy.

Based on our prior findings (6, 7), we performed analyses separately by previous PCPT treatment assignment. We compared participant characteristics, including measures of inflammation, by case and control status using linear and logistic regression. In controls, we compared participant characteristics by proportion of biopsy cores with inflammation (none, some, all). We estimated ORs and 95% confidence intervals (CIs) of total and Gleason sum <4+3 (groups 1 and 2) prostate cancer by prevalence (at least one biopsy core with inflammation), proportion of biopsy cores with inflammation (none, some, all), and mean percentage of tissue area with inflammation (0%, tertiles of the distribution >0% with cutoff points based on controls) using logistic regression. We could not evaluate associations for Gleason sum ≥4+3 (grade groups 3–5) disease due to small number (N = 9). We adjusted for age at PCPT end-of-study biopsy and race in the main analysis, and confirmed that additional adjustment for family history, height, BMI, smoking, and energy intake did not notably change the results. To test for trend, we used an ordinal term for proportion of cores with inflammation [none (0), some (1), all (2) cores] or mean percentage of tissue area with inflammation [none (0), tertile 1 (1), tertile 2 (2), tertile 3 (3)] and evaluated the coefficient by the Wald test. To determine whether associations differed by previous PCPT treatment arm, we used the likelihood ratio test to compare models with and without an interaction term between treatment arm and inflammation. Although vitamin E and selenium did not reduce prostate cancer risk in SELECT (11, 12) and inflammation was assessed in tissue collected before the men took SELECT supplements, we confirmed that trial arm was not an effect modifier using this same method.

To test whether we could confirm the link between inflammation and PSA that we previously observed in the PCPT placebo arm (6, 9) and the lack of a link in the finasteride arm (7), we calculated mean serum PSA concentration closest in time before the PCPT end-of-study biopsy by prevalence of inflammation and by proportion of biopsy cores with inflammation separately in controls and cases. We used linear regression and adjusted for age at end-of-study biopsy and race. We conducted analyses using SAS version 9.4 and report P values from two-sided tests.

This analysis included 41 prostate cancer cases (31 cases <4+3 = grade groups 1 and 2; 5 cases ≥4+3 (grade groups 3–5); 5 missing grade) and 85 controls. Cases and frequency-matched controls did not differ on baseline characteristics except possibly for family history (Table 1). Control characteristics did not differ across proportion of biopsy cores with inflammation (Table 2). 78.1% of cases (84.1% of cases with grade <4+3) and 68.2% of controls had at least one baseline biopsy core (of ∼5 reviewed) with inflammation (Table 3). In controls, on average, 32.1% of cores per man had inflammation of any grade. Of controls with at least one biopsy core with inflammation, on average, 4.4% of tissue area had inflammation. Most inflammation was chronic.

>Compared with men whose mean percentage of tissue area with inflammation was 0% (no biopsy core had inflammation), the age and race-adjusted OR of Gleason sum <4+3 prostate cancer, statistically significantly increased with increasing mean percentage of biopsy core tissue area with inflammation (vs. 0%, >0 to <1.8% OR = 1.70, 1.8% to <5.0% OR = 2.39, ≥5% OR = 3.31, P_trend = 0.047). The pattern was similar for total (Table 4), and for Gleason sum 3+4 (N = 11), and <7 (N = 20) prostate cancer. Although not statistically significant, ORs of total (OR = 1.66; 95% CI, 0.70–3.96) and Gleason sum <4+3 (OR = 2.44; 95% CI, 0.84–7.07) prostate cancer were in the positive direction for having at least one biopsy core with inflammation (Table 4). Also, ORs of total and Gleason sum <4+3 prostate cancer tended to increase across proportion of biopsy cores with inflammation (Table 4). Results were comparable after additional adjustment for prostate cancer risk factors.

In both controls and cases, PSA concentration at the PCPT end-of-study biopsy did not differ between men with at least one biopsy core with inflammation and men with no cores with inflammation, although men with all cores with inflammation tended to have higher PSA than man with none or some cores with inflammation (Table 5).

This analysis included 51 cases (34 cases <4+3 = grade groups 1 and 2; 4 cases ≥4+3 (grade groups 3–5); 13 missing grade) and 108 controls. In men previously assigned to finasteride, prostate cancer cases and frequency-matched controls did not significantly differ on baseline characteristics except possibly for family history and energy intake (Table 1). The proportion of cases and controls still taking finasteride at the PCPT end-of-study biopsy was similar. In controls, education, smoking status, and energy intake possibly differed across proportion of biopsy cores with inflammation (Table 2).

Controls previously in the finasteride arm appeared to have a slightly higher prevalence of inflammation than controls previously in the placebo arm (74.8% vs. 68.9%), but did not have a higher percentage of biopsy cores with inflammation or higher percentage of tissue area with inflammation (Table 3). However, Gleason sum <4+3 cases previously in the finasteride arm appeared to have a slightly lower prevalence (76.4% vs. 84.1%), smaller percentage of biopsy cores with inflammation, and smaller percentage of tissue area with inflammation than Gleason <4+3 cases in the placebo arm. Like in the placebo arm, most inflammation was chronic.

Unlike men previously in the PCPT placebo arm, among men previously in the finasteride arm having at least one biopsy core with inflammation was not associated with risk of total or Gleason sum <4+3 prostate cancer when adjusting for age and race (Table 4). We could not rule out a modest positive, non-statistically significant association among men who had all biopsy cores with inflammation or were in the top tertile of mean percentage of tissue area with inflammation (Table 4). Results were unchanged after further adjusting for potentially confounding factors. Results were unchanged after excluding men who were not still taking finasteride at the end-of-study biopsy. While the association between having at least one biopsy core with inflammation and total (P_interaction = 0.5) and Gleason sum <4+3 (P_interaction = 0.3) prostate cancer appeared to be different in the two PCPT treatment arms, we did not detect a statistical interaction.

In controls, PSA concentration at the PCPT end-of-study biopsy was higher in men with at least one biopsy core with inflammation and increased across proportion of biopsy cores with inflammation (Table 5). In cases, PSA concentration did not differ between men with and without at least one biopsy core with inflammation and did not change across proportion of biopsy cores with inflammation (Table 5).

We examined negative prostate biopsies for the presence of inflammation in men in the PCPT who were not biopsied for indication (e.g., had low PSA and negative DRE) and who later enrolled and were followed in SELECT to determine whether inflammation in the biopsy was associated with subsequent prostate cancer. We observed that the odds of detecting Gleason sum <4+3 [grade groups 1 and 2 (13) prostate cancer, the majority of cases] while on SELECT statistically significantly increased with percent tissue area with inflammation. We previously observed a positive association between inflammation and prostate cancer, including Gleason sum <7 disease (at least one core with inflammation: OR = 1.57, 95% CI, 0.83–3.00; ref. 6) in a different set of men from the PCPT in whom we assessed inflammation in the biopsies that were used to rule in or out cancer. In that prior study, we noted a stronger association for Gleason sum ≥7 prostate cancer (OR = 2.24, 95% CI, 1.06–4.71; across none, some, all cores with inflammation: P_trend = 0.01). In this study (prospective PCPT–SELECT), we could not evaluate the association for Gleason ≥4+3 disease because the number of cases was too small. Taken together, our current PCPT–SELECT findings and our prior PCPT findings (6) support the role of intraprostatic inflammation in prostate cancer development.

With respect to our second hypothesis, having at least one biopsy core with inflammation was not associated with prostate cancer in men previously assigned to finasteride in PCPT, consistent with what we previously observed (7). However in PCPT–SELECT unlike in PCPT, we could not rule out a modest, nonstatistically significant increased odds of Gleason sum <4+3 disease in men who had all cores with inflammation compared with none. In contrast to our study in the PCPT when inflammation and diagnosis of prostate cancer were concurrent with finasteride treatment assignment, in PCPT–SELECT, the inflammation we assessed was concurrent with finasteride treatment assignment (85% were still taking it), but during their follow-up, the men were no longer receiving finasteride. We speculated previously that finasteride-associated intraprostatic inflammation may not be pro-carcinogenic (7).

In PCPT–SELECT, we found that intraprostatic inflammation was common. We hypothesized that if inflammation is a cause of prostate cancer, and given prostate cancer's high incidence, then intraprostatic inflammation also should be common. Prior studies reported that inflammatory infiltrates are very frequently observed in biopsies performed for elevated PSA or abnormal DRE (14), radical prostatectomy specimens (15), and tissue resected for benign prostatic hyperplasia (16, 17). But these studies were conducted using tissue removed for clinical indications, which themselves are associated with or, a consequence of inflammation. By design, our evaluation of the inflammation–prostate cancer association was not biased by any link between clinical indication and inflammation. While the prevalence of inflammation (based on ∼5 cores reviewed per man) that we observed in controls in PCPT–SELECT was high, it was lower than what we previously reported in a different group of controls sampled from PCPT placebo (based on ∼3 cores reviewed per man—78.2%; ref. 6) and finasteride (92.4%; ref. 7) arms. We speculate that if inflammation is associated with prostate cancer risk, by sampling as controls men who reached the end of SELECT follow-up without having been diagnosed with prostate cancer, the sampled controls would have a reduced prevalence of inflammation compared with the source population for the cases and compared with controls sampled using risk set sampling (18). Despite the lower prevalence of inflammation in the current study, as we found before it was higher in those previously in the finasteride arm than placebo arm, which as we indicated (7) is consistent with the observation that androgen deprivation therapy elicits inflammation (19, 20).

Our PCPT–SELECT finding that men in the previous PCPT placebo arm with all cores with inflammation appeared to have higher PSA at the end-of-study biopsy compared with men with no cores with inflammation is compatible with our prior study (6, 9). These findings are supportive of the hypothesis that inflammation leading to higher PSA concentration could distort the association between inflammation and prostate cancer. This pattern was not observed in PCPT–SELECT cases previously in the PCPT placebo arm, or in our prior study (6, 9). Elevations in serum PSA in cases may be more likely to result from cancer than from inflammation, and indeed cases had, on average, a higher PSA concentration measured at the time of the PCPT end-of-study biopsy. Like in our prior study (7), in PCPT–SELECT cases previously in the finasteride arm PSA concentration did not differ by inflammation status. Unlike in our prior PCPT study (7), in PCPT–SELECT controls previously in the PCPT finasteride arm PSA concentration was higher in men with at least one biopsy core with inflammation and increased across proportion of biopsy cores with inflammation. The reason for this latter difference is unknown.

Studies in the Finnish prostate cancer screening trial (21) and REDUCE (22) reported that men who were biopsy negative for prostate cancer had a lower risk of prostate cancer subsequently if they had inflammation (or greater proportion of biopsy cores with inflammation) in their prior negative biopsy (21–23). Results from our current and prior (6) studies and results from the Finnish and REDUCE studies likely differ for many reasons, including that those other analyses required men to have had a negative biopsy following an elevated PSA on the first screen in the Finnish study or a negative biopsy prior to trial entry in REDUCE. In contrast, in PCPT–SELECT, the biopsies we used were from men who did not have an indication for biopsy and also had PSA ≤4 ng/mL at SELECT entry. Men with elevated PSA leading to a biopsy that was negative (criteria for the Finnish analysis (21) may indeed be more likely to be negative for cancer on a follow-up biopsy if their initial biopsies showed inflammation because the main determinant of the PSA rise in these men was the inflammation and not cancer. The same is possibly true in REDUCE for men whose negative biopsies before trial entry were performed for elevated PSA, although a sensitivity analysis among men with a PSA concentration of 2.5 to 4.0 ng/mL apparently supported the inverse association (22).

Several study features warrant discussion. (i) In this innovative study, we leveraged cohorts from two completed trials to prospectively test the inflammation–prostate cancer association in men without an indication for biopsy. (ii) We used Epstein and colleagues' new grading system (13) to classify cases for this analysis. (iii) We assessed intraprostatic inflammation 1.2 to 10.5 years (mean 5.9) before prostate cancer diagnosis. However, we do not know whether the inflammation we assessed was in the etiologically relevant window or correlated with it, or was of sufficient duration to influence prostate cancer development. Although the exact duration of chronic inflammation needed for a chronic inflammatory condition (e.g., gastritis-associated stomach cancer) to increase risk is unknown, at least at times, it may take decades. (iv) Because of the sample size, we could not evaluate this association in African American men, an important question given their substantially higher prostate cancer rates. (v) The number of biopsy cores we reviewed per man (∼5) was higher than in our prior study (∼3) because more cores were mounted per slide for the end-of-study biopsies than biopsies performed during the PCPT. This likely resulted from the increase over time nationally in number of cores sampled coupled with the PCPT protocol of embedding cores by region into a single block irrespective of number of cores taken from that region. (vi) we confirmed our previous observation for lower-grade prostate cancer, a disease with substantially lower morbidity and mortality. A limitation of our approach was that the sample size was fixed and small, and we could not evaluate important outcomes—lethal and Gleason ≥4+3 disease (grade groups 3–5). Finally, (vi) the approach we took allowed us to investigate the association between intraprostatic inflammation and prostate cancer risk in a setting that did not suffer from PSA-associated detection bias. The tradeoff for internal validity may be reduced generalizability in that the men in this analysis had to have participated in two long-term chemoprevention trials and to have been eligible for those trials including by having low PSA at entry into both.

In conclusion, this study—in which biopsy tissue was collected, on average, 6 years before prostate cancer diagnosis, and thus is temporally correct for drawing causal inferences, and that followed men who did not have a clinical indication for the biopsy used as the baseline for assessing inflammation – provides evidence supporting the hypothesis that inflammation influences prostate cancer development. Taken together with our prior PCPT study, our findings inform prostate cancer etiology with an eye toward identifying men at higher risk so that they can be targeted for preventive intervention.

No potential conflicts of interest were disclosed.

The content of this work is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Conception and design: E.A. Platz, C.E. Joshu, W.G. Nelson, S.M. Lippman, I.M. Thompson, P.J. Goodman, C.M. Tangen, A.M. De Marzo

Development of methodology: E.A. Platz, M.S. Lucia, C.M. Tangen, A.M. De Marzo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.A. Platz, M.S. Lucia, E.A. Klein, S.M. Lippman, I.M. Thompson, C.M. Tangen, A.M. De Marzo

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.A. Platz, I. Kulac, J.R. Barber, C.E. Joshu, S.M. Lippman, P.J. Goodman, A.M. De Marzo

Writing, review, and/or revision of the manuscript: E.A. Platz, I. Kulac, J.R. Barber, C.G. Drake, C.E. Joshu, W.G. Nelson, M.S. Lucia, E.A. Klein, S.M. Lippman, H.L. Parnes, I.M. Thompson, P.J. Goodman, C.M. Tangen, A.M. De Marzo

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Kulac, E.A. Platz, S.M. Lippman, I.M. Thompson, P.J. Goodman

Study supervision: E.A. Platz, S.M. Lippman, I.M. Thompson, A.M. De Marzo

This work was funded by W81XWH-12-1-0170 (to E.A. Platz), UM1 CA182883 (to C.M. Tangen and I.M. Thompson), U10 CA37429 (to C.D. Blanke), P50 CA58236 (to W.G. Nelson), P30 CA054174 (to I.M. Thompson), and P30 CA006973 (to W.G. Nelson). C.E. Joshu was supported by a young investigator award from the Prostate Cancer Foundation.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.