Obesity and survival after radical prostatectomy: A 10-year prospective cohort study

Authors


Abstract

BACKGROUND.

Obesity and prostate cancer are among the most common health problems affecting American men today. The authors' goal was to assess the impact of obesity on clinical and pathologic features of prostate cancer and long-term outcomes.

METHODS.

The authors performed a prospective cohort study on 5313 men who underwent radical prostatectomy between 1990 and 1999. Patient height and weight were measured at the time of surgery to calculate the body mass index (BMI). The patients were separated into 3 BMI groups: BMI <25, 25–29.9, and ≥30 kg/m2. The associations between BMI and age, prostate-specific antigen (PSA) level, and Gleason score were assessed with the Spearman rank correlation test. The associations between BMI and pathologic features were assessed with the Mantel–Haenszel χ2 test. Fifteen-year biochemical progression-free survival, systemic progression-free survival, cancer-specific survival, and overall survival were estimated using the Kaplan–Meier method and evaluated using Cox models.

RESULTS.

The median length of follow-up for the entire cohort was 10.1 years. Clinical and pathologic features appear worse in patients with a higher BMI. On univariate and multivariate analyses, it was found that BMI had no impact on biochemical progression, systemic progression, prostate cancer survival, or overall survival.

CONCLUSIONS.

Obese patients appear to have worse pathologic features at the time of prostatectomy. Despite these features, long-term oncologic outcomes, including cancer-specific survival, remain the same regardless of BMI. BMI appears to influence prostate cancer outcomes at the time of prostatectomy, as evidenced by more aggressive pathologic features. However, after prostatectomy, BMI does not appear to be an independent predictor of recurrence or survival. Cancer 2006. © 2006 American Cancer Society.

Prostate cancer and obesity are among the most common health problems affecting American men today. Over the past 25 years, the number of obese men in the U.S. has increased from 15% to 30%.1 Although obesity has been undeniably linked to an increased risk of developing, and dying from, cancer, the epidemiologic data regarding prostate cancer incidence and mortality are conflicting.2–5 The results of recent radical prostatectomy series have also been mixed, with some groups suggesting that obese prostate cancer patients may have worse pathologic features and higher biochemical recurrence rates after radical prostatectomy.6–11 All these surgical studies share 2 major limitations: 1) their median follow-up is very short (≤5 years) and 2) they rely exclusively on a surrogate endpoint—biochemical (PSA) recurrence—as their measure of outcome. These 2 criticisms are of particular importance, as it has been previously established that biochemical recurrence can occur late in the postoperative course and that the presence of biochemical recurrence alone is not always a good predictor of prostate-cancer–specific death.12–15

The goal of the current study was to analyze of the impact of obesity, as measured by the body mass index, on the clinicopathologic features of prostate cancer and on the long-term mortality outcomes of radical prostatectomy. We present data from a large prospective cohort study with a median follow-up of 10 years that demonstrates that obesity is associated with a more aggressive pathologic tumor profile but that it does not affect prostate cancer outcomes after radical prostatectomy.

MATERIALS AND METHODS

Study Cohort

With approval from the Mayo Clinic Institutional Review Board, patients who had undergone radical retropubic prostatectomy (RRP) for prostate cancer between 1990 and 1999 were identified from our Mayo Clinic Prostatectomy Registry. The clinicopathologic and follow-up information in this registry is collected prospectively and updated annually. A minority of patients receive follow-up elsewhere and have their records updated by correspondence, leaving only 4% of participants for whom follow-up is incomplete. After exclusion of patients who received neoadjuvant therapies prior to surgery (n = 707), who refused research authorization (n = 54), or who did not have their body mass index (BMI) recorded (n = 1807), a total of 5313 men remained and formed the current study cohort. Radical retropubic prostatectomy and pelvic lymphadenectomy were performed by a number of different surgeons, using standard operative techniques. Pathologic evaluation was done using a limited sampling technique on frozen tissue sections at the time of surgery, with subsequent examination of paraffin-embedded sections the following day.16, 17 Tumor stage and grade were assigned using the 1997 UICC-AJCC TNM system and the Gleason system respectively.18, 19 Tumor DNA ploidy was assessed by flow cytometry.

Body Mass Index

The BMI was calculated using patient height and weight as measured on the day of surgery. BMI was categorized as per the National Institutes for Health classifications, with individuals with a BMI <25 kg/m2 considered normal, those with a BMI of 25–29.9 kg/m2 considered overweight, and those with a BMI ≥30 kg/m2 considered obese.20

Clinical Endpoints and Definitions

Postoperative follow-up was carried out quarterly to semiannually for the first 2 years and annually thereafter by clinical assessment, measurement of serum PSA, and other investigations as indicated. Adjuvant treatment was defined as any form of androgen deprivation or radiation therapy within 90 days of surgery. Biochemical failure was defined as a follow-up PSA of 0.4 ng/mL or greater and occurred in 1687 (32%) patients.21 Systemic progression, defined as demonstrable metastatic disease on imaging (radionuclide bone scintigraphy or plain film) or pathologic evidence of prostate cancer in any postoperative tissue biopsy, occurred in 290 (5%) patients. For patients who died during the course of follow-up, the cause of death was verified by death certificates or correspondence to their treating physician. A total of 151 (3%) patients died of prostate cancer and 967 (18%) died of other causes.

Statistical Methods

Statistical analyses were performed using the SAS software package, version 8.2 (SAS Institute, Cary, NC). Statistical analyses and modeling procedures used BMI as a continuous variable. The calculation of odds ratios (ORs) and their confidence intervals (95% CI) are the only exception to this rule because patients were categorized by WHO BMI groupings for these comparisons. The associations between BMI and age, serum PSA level, and Gleason score were assessed with Spearman's rank correlation test. The associations between BMI and stage, margin status, seminal vesicle invasion, and tumor ploidy were assessed with the Mantel–Haenszel χ2 test. Biochemical recurrence-free survival (bRFS), systemic progression-free survival (sPFS), prostate-cancer–specific survival (CSS), and overall survival (OS) were estimated using the Kaplan–Meier method. Cox proportional hazards modeling was used to assess the impact of baseline BMI on time-dependent outcomes while adjusting for other known predictors of prostate cancer events in our patient population: prostatectomy Gleason score, preoperative serum PSA, surgical margin status, seminal vesicle invasion, and the use of adjuvant treatments.22 All P values are two-sided and considered statistically significant if less than 0.05.

RESULTS

BMI and Clinicopathologic Features

As summarized in Table 1, clinical and pathologic features appear generally worse in patients with a higher BMI. Obese patients had prostate cancer treatment at a younger mean age than did those with a BMI <25 kg/m2 (63.8 vs. 65.4 years, P < 0.001). Gleason scores 8–10 were slightly more common in the biopsy specimens (6.4% vs. 5%, P = 0.010) and radical prostatectomy specimens (6.6% vs. 4.1%, P < 0.001) of obese patients than in those of normal patients. Although greater number of obese than normal patients clinically had organ-confined prostate cancer (94.2% vs. 91.1%, P < 0.001), pathologic staging revealed no significant difference in organ-confinement between these 2 groups (65.6% vs. 76.5%, P = 0.234). Despite this, the surgical margins were more frequently positive in obese than in normal patients (46.3% vs. 32.6%, P < 0.001) and the seminal vesicles were more frequently invaded (15.5 vs. 12.8%, P = 0.036). We found no significant differences in the incidence of abnormal tumor ploidy (26.5% vs. 28.4%, P = 0.424) or in the median preoperative serum PSA levels (7.1 vs. 7.2 ng/mL, P = 0.970) between obese and normal patients.

Table 1. Clinical and Pathologic Characteristics Stratified by BMI
 BMI, kg/m2P
<25 (N = 1163)25–29.9 (N = 2889)30+ (N = 1261)
  • BMI indicates body mass index.

  • *

    Values in parentheses are percentages.

Age at surgery (rounded)<0.0001
 Mean ± SD65.4 ± 6.9664.6 ± 6.7363.8 ± 6.63
 Median66.065.065.0
 Range(41.0–83.0)(39.0–83.0)(42.0–82.0)
Preop PSA0.97
 Median7.27.27.1
 Q1, Q34.7, 11.84.8, 11.34.9, 11.7
 Range(0.5–388.0)(0.1–616.0)(0.2–321.0)
Clinical stage, 1997 TNM revision<0.0001
 Missing186
 T1ab18 (1.5)*67 (2.3)28 (2.2)
 T1c295 (25.4)851 (29.5)415 (33.1)
 T2a586 (50.4)1362 (47.3)557 (44.4)
 T2b160 (13.8)400 (13.9)182 (14.5)
 T34103 (8.9)201 (7)73 (5.8)
Biopsy Gleason score<0.0001
 <6334 (40)806 (38.2)330 (33.5)
 6280 (33.5)761 (36)362 (36.8)
 7179 (21.4)436 (20.7)230 (23.4)
 8+42 (5)108 (5.1)63 (6.4)
Pathologic Gleason score<0.0001
 <6458 (39.5)1036 (36.1)367 (29.2)
 6347 (29.9)916 (31.9)384 (30.5)
 7307 (26.5)767 (26.7)424 (33.7)
 8+48 (4.1)153 (5.3)83 (6.6)
Pathologic stage,1997 TNM revision0.137
 T2aN0307 (26.4)705 (24.5)292 (23.2)
 T2bN0478 (41.1)1249 (43.4)534 (42.4)
 T34N0306 (26.3)763 (26.5)364 (28.9)
 TxN+72 (6.2)161 (5.6)69 (5.5)
Margin positive379 (32.6)1117 (38.7)584 (46.3)<0.0001
Seminal vesicle involvement149 (12.8)371 (12.8)195 (15.5)0.048
Ploidy
 Diploid796 (71.6)2045 (74)882 (73.6)0.444
 Tetraploid238 (21.4)565 (20.4)218 (18.2)
 Aneuploid78 (7)155 (5.6)99 (8.3)

After removing patients with positive lymph nodes, we examined the relationship between the clinical stage (determined preoperatively by digital rectal examination (DRE) by the treating urologist) and pathologic stage (determined postoperatively by the pathologist) across BMI categories and observed an interesting finding (Table 2). Patients who were obese were more likely to be clinically understaged than were normal patients (27% vs. 24%, P = .028). This implies that the DRE tended to miss larger cancers in obese patients.

Table 2. Clinical Staging Error Stratified by BMI Group
BMI, kg/m2Clinical understagingAccurate clinical stagingClinical overstagingP*
  • BMI indicates body mass index.

  • *

    Spearman's rank correlation test.

  • Values in parentheses are % row total.

<25258 (24)798 (73)33 (3).028
25–30666 (25)1987 (73)58 (2)
>30322 (27)844 (71)19 (2)

Oncologic Outcomes

The median length of follow-up for the entire cohort was 10.1 years (interquartile range [IQR]: 7.9–12.3 years). Median follow-up for BMI <25, 25–29.9, and >30 kg/m2 was 10.7 years (IQR: 8.0-12.8 years), 10.2 years (IQR: 8.0–12.4 years), and 9.4 years (IQR: 7.7–11.8 years), respectively. These differences were statistically significant (P < 0.001). Obese patients were treated more frequently with adjuvant treatment than were normal patients (OR = 1.40, 95% CI: 1.13–1.72, P < 0.001). The estimated 15-year bRFS for normal, overweight, and obese patients was 63% (95% CI: 59%–67%), 60% (95% CI: 57%–63%), and 60% (95% CI: 56%–64%), respectively (Fig. 1). The estimated 15-year sPFS for normal, overweight, and obese was 92% (95% CI: 89%–96%), 91% (95% CI: 90%–93%), and 90% (95% CI: 87%–93%), respectively (Fig. 2). The estimated 15-year CSS for normal, overweight, and obese patients was 89% (95% CI: 78%–100%), 93% (95% CI: 91%–96%), and 90% (95% CI: 85%–96%), respectively (Fig. 3). The estimated 15-year OS for these same categories was 59% (95% CI: 50%–70%), 62% (95% CI: 56%–68%), and 58% (95% CI: 48%–70%), respectively (Fig. 4). The univariate survival endpoints are outlined in Table 3.

Figure 1.

Fifteen-year Kaplan–Meier estimate of biochemical progression-free survival based on patient BMI after radical prostatectomy.

Figure 2.

Fifteen-year Kaplan–Meier estimate of systemic progression-free survival based on patient BMI after radical prostatectomy.

Figure 3.

Fifteen-year Kaplan–Meier estimate of cancer specific survival based on patient BMI after radical prostatectomy.

Figure 4.

Fifteen-year Kaplan–Meier estimate of overall survival based on patient BMI after radical prostatectomy.

Table 3. Univariate Survival Endpoints Stratified by BMI Group
BMI, kg/m2Event rate*15-year survival estimate, %P
  • BMI indicates body mass index.

  • *

    Number at risk: overall = 5313, normal = 1162, overweight = 2889, obese = 1262.

  • Log rank test.

  • Values in parentheses are percentages.

  • §

    Values in square brackets are 95% CI.

Biochemical failure1687 (32) 0.135
 <25358 (31)63 [59, 67]§
 25–30908 (31)60 [57, 63]
 >30421 (33)60 [56, 64]
Systemic progression290 (5) 0.121
 <2554 (5)92 [89, 96]
 25–30159 (6)91 [90, 93]
 >3077 (6)90 [87, 93]
Prostate cancer death151 (3) 0.213
 <2531 (3)89 [89, 96]
 25–3081 (3)93 [90, 93]
 >3039 (3)90 [87, 93]
Any death967 (18) 0.465
 <25225 (9)59 [89, 96]
 25–30511 (8)62 [90, 93]
 >30231 (8)58 [87, 93]

After adjusting for other important covariables, the multivariate Cox proportional hazards model showed that BMI did not have an independent impact on bRFS, as outlined in Table 4 (Hazard ratio (HR) = 1.00, 95% CI: 0.99–1.01, P = 0.933); CSS, as outlined in Table 5 (HR = 1.02, 95% CI: 0.97–1.07, P = 0.372); or overall survival, as outlined in Table 6 (HR = 0.99, 95% CI: 0.98–1.02, P = 0.830). Because adjuvant treatments were more common in patients with higher BMIs, we further examined the possibility that this factor could be confounding the relationship between BMI and our survival endpoints, by stratifying the Cox regression models by adjuvant therapy status. In patients who did not receive adjuvant therapy, BMI did not appear to affect bRFS (HR = 1.00, 95% CI: 0.98–1.01, P = 0.842), cancer-specific survival (HR = 1.05, 95% CI: 0.98–1.12, P = 0.139), or overall survival (HR = 1.00, 95% CI: 0.98–1.02, P = 0.987). Similarly, in patients who were treated with adjuvant therapies, BMI did not impact bRFS (HR = 1.00, 95% CI: 0.97–1.03, P = 0.915) or cancer-specific survival (HR = 0.99, 95% CI: 0.92–1.07, P = 0.815, or overall survival (HR = 0.98, 95% CI: 0.94–1.02, P = 0.328).

Table 4. Multivariate Associations of BMI at Treatment and Pathological Variables With the Risk of Biochemical Progression
VariableHazard ratio95% confidence intervalP
  1. BMI indicates body mass index.

BMI1.000.99–1.010.864
Pathologic Gleason score1.371.31–1.43<0.0001
Preop PSA (log 2)1.271.22–1.33<0.0001
Seminal vesicle (+)1.931.69–2.20<0.0001
Margin (+)1.591.43–1.77<0.0001
Adjuvant treatment0.390.34–0.45<0.0001
Table 5. Multivariate Associations of BMI at Treatment and Pathological Variables With Cancer-Specific Survival
VariableHazard ratio95% confidence intervalP
BMI1.020.97–1.070.372
Pathologic Gleason score1.991.72–2.30<0.0001
Preop PSA (log 2)1.060.93–1.210.376
Seminal vesicle (+)3.762.60–5.43<0.0001
Margin (+)1.230.85–1.780.278
Adjuvant treatment0.850.58–1.250.420
Table 6. Multivariate Associations of BMI at Treatment and Pathological Variables With Overall Survival
VariableHazard ratio95% confidence intervalP
BMI1.000.98–1.020.830
Pathologic Gleason score1.141.07–1.21<0.0001
Preop PSA (log 2)1.071.01–1.30.019
Seminal vesicle (+)1.621.37–1.92<0.0001
Margin (+)1.130.98–1.310.099
Adjuvant treatment0.920.77–1.090.329

DISCUSSION

Obesity is an established risk factor for developing, and dying from, cancer.3, 4 Whether or not the risk of specifically developing prostate cancer is increased in overweight and obese individuals is a matter of considerable debate.3–5 This controversy is, in fact, somewhat surprising because prostate cancer is relatively prevalent and hundreds of thousands of men have been studied while attempting to answer this question. One would expect that if obesity played an unequivocal role in the pathogenesis of prostate cancer, then the large population studies would generally lead to similar conclusions; but this has not been the case. The Health Professionals Follow-Up Study tracked 47,757 men prospectively over 14 years and showed that obese men had a relative risk of 0.52 of developing prostate cancer when compared with men with a normal BMI.2 Contrarily, a study on 951,466 Norwegian men showed that the incidence of prostate cancer was increased by 10% when compared with normal men.23 When 67,477 men were followed for 10 years in the Austrian VHM and PP Study, however, no increased risk of prostate cancer was detected in obese men at all.3 Similarly, an 18-year prospective study on 135,006 Swedish construction workers failed to find an association between BMI and prostate cancer incidence.5 A reasonable conclusion is that we do not know whether obesity is associated with an increased incidence of prostate cancer.

The Swedish construction worker study did show, however, that the risk of prostate cancer death was increased by ∼40% in obese men when compared with that in normal men.5 Similar results were shown in the Cancer Prevention Study II, wherein after following 404,576 men for 16 years, an increase in prostate cancer mortality of roughly 30% was noted in obese men.4 Unfortunately, these 2 studies did not assess the impact of what is arguably the single most important confounder that affects prostate cancer mortality: treatment status. Randomized controlled trials have shown not only that radical prostatectomy reduces the risk of dying from prostate cancer by nearly half, but that this benefit can be improved further by the appropriate use of adjuvant radiotherapy, adjuvant androgen deprivation, and salvage chemotherapy.24–28 One might therefore question whether these population-based studies are demonstrating biologically worse prostate cancers occurring in obese patients or simply inferior prostate cancer diagnosis and treatment in these individuals.

To help clarify this issue, we conducted a prospective cohort study on 5313 consecutive men with prostate cancer who were treated by radical prostatectomy and compared by their perioperative BMI for prostate-cancer–specific outcomes. When compared with other similar studies, ours is novel in 4 regards: First, our median duration of follow-up—over 10 years—is much longer than that of other studies. This ensures that we are missing a minimum of patients who may present with late recurrences and that we are observing an important proportion of disease mortality endpoints.12, 14 Second, by evaluating prostate-cancer-specific survival and overall survival, we are not relying exclusively on a potentially inaccurate surrogate endpoint—PSA failure—to draw conclusions.15 Third, a higher proportion of the patients in the current study have high-risk prostate cancers. If obesity is associated with increased prostate cancer aggressiveness, this should be evidenced by an increase in high-risk obese patients. Fourth, we provide evidence that obese men are understaged clinically more frequently than are normal men, suggesting impaired diagnostic precision. On the basis of the findings of the current study, we propose that obese men may actually have cancers biologically similar to those of nonobese men but that other physician-related factors, such as impaired prostate cancer diagnostic tests (i.e. serum PSA and DRE) and bias in the physician–patient relationship, are leading to changes in the presentation of prostate cancer in obese individuals.

In the current series, obese patients presented with higher pathologic Gleason scores, higher positive margin rates, and more seminal vesicle involvement than did patients with a normal BMI. These findings are generally consistent with those from recently published series.6–11, 29 The greater surgical margin positivity rate observed in our obese patients is likely due to increased technical difficulty of the surgery. Freedland et al. have shown that obese men are over twice as likely to have an inadvertent capsular incision at surgery than are normal men.30 The higher pathologic Gleason scores and the increased incidence of seminal vesicle invasion that were noted in obese patients are more difficult to explain. Although there are biologic and hormonal changes associated with obesity that may potentially be involved in oncogenesis, good prospective studies evaluating their role specific in prostate cancer are lacking.31 An alternative explanation would be that obese patients are treated relatively later in the course of their disease because of diagnostic difficulties and problems with the physician–patient interaction. This is supported by studies of active surveillance with delayed intervention as a treatment for prostate cancer, which have shown that roughly 5%–10% of men who are rebiopsied after 1.5–2 years of observation will show an increase in Gleason score by at least 2 points.32

Since prostate biopsies are usually precipitated by a serum PSA elevation or an abnormal DRE, it stands to reason that factors affecting these 2 diagnostic modalities could impact on prostate cancer diagnosis. Baillergon et al. have shown that serum PSA levels are lower in obese men than in normal men—presumably because of lower circulating androgen levels—and have suggested that this may mask prostate cancer diagnosis.33 Our finding that DRE understages obese patients more frequently than normal patients suggests that the DRE is also less sensitive in obese patients. Other groups have similarly shown that the pelvic examination is impaired in obese patients.34 There is also much evidence to suggest that there is bias in the way physicians diagnose and treat obese patients.35, 36 This includes differences in the way physicians interview and examine patients,37, 38 differences in the perceived priorities of medical problems,39 differences in the way they screen for cancers,40, 41 and differences in how they actually perceive the patient as a person.35, 37, 38 All of these factors could potentially affect whether or not an obese man is screened for prostate cancer.

Similar to the study of Mallah et al.,10 the current study failed to identify a meaningful association between BMI and radical prostatectomy outcomes. In their study, BMI was very weakly associated with outcome and not considered a useful predictor of disease progression.10 Why did the current study fail to find an association between BMI and radical prostatectomy outcome when several other studies have implied otherwise?6–9 Part of the answer may be explained by confounding variables such as race and adjuvant treatment status. Other factors including length of follow-up and hormonal mechanisms may also play a role. Our study has a median follow-up of roughly 10 years—a period that is 2–5 times longer than those in the previous studies—and could be considered moderate for prostate cancer. This increased follow-up enables the capture of late biochemical recurrences and also allows the use of ‘hard’ survival endpoints. Biochemical recurrence should be considered a surrogate endpoint that requires confirmation by actual survival data.15 The current study uses these hard endpoints and shows that there is no difference in radical prostatectomy survival outcomes with increasing BMI. Obese patients in the current study did, however, receive more adjuvant treatment than did nonobese patients and it could therefore be argued that adjuvant treatment may have confounded the relationship between BMI and outcome. The results of our multivariate Cox modeling process (which included adjuvant treatment status as a covariable) and our multivariate models stratified by treatment status argue very strongly against this. One other major difference between our study and other series is the racial homogeneity of our cohort. Both Freedland et al. and Amling et al. reported the proportion of African American patients in their study populations to be 26% and 21% respectively.6, 7 These proportions are clearly much higher than the 1% incidence of African American race in our cohort. The importance of this point is that African American race is a potential confounder because it is associated with both the exposure (increased BMI) and the outcome (prostate cancer recurrence).42, 43 Interestingly, in the analysis of Amling et al.,7 black race—not BMI—remained a significant predictor of prostate cancer recurrence after adjustment for race in their multivariate model. Finally, from a hormonal standpoint, we recognize that obesity is associated with decreased free and total testosterone levels.44, 45 Increased abdominal adiposity promotes peripheral aromatization of testosterone to estrogen, decreasing the ratio of free testosterone to free estradiol.46 In a suppressed androgen state, postprostatectomy patients may derive some protection from disease recurrence. It is conceivable that alongside increased follow-up, adjuvant treatment use, and racial homogeneity, inherent hormonal protection may partly explain the improved outcomes in our cohort. We recognize another important limitation to our study. The equivalent overall survival for all BMI groups suggests that ‘healthy’ obese patients were selected for surgery. Patients were not stratified based on a Charlson comorbidity score or ASA index. It is conceivable that overall survival outcomes may vary if patients were adjusted to preoperative morbidity status.

In this study we demonstrate that despite having worse pathologic features in their prostate cancers at the time of prostatectomy, obese patients experience similar long-term cancer endpoints as do nonobese patients. Obese patients can be counseled that although their increased weight may be associated with certain characteristics of more aggressive prostate cancers, it does not impart any additional long-term risk of biochemical recurrence or prostate cancer death when treated by radical prostatectomy.

Ancillary