Obesity and oncological outcome after radical prostatectomy: impact of prostate-specific antigen-based prostate cancer screening: results from the Shared Equal Access Regional Cancer Hospital and Duke Prostate Center Databases
Stephen J. Freedland, Division of Urologic Surgery and the Duke Prostate Center, and Departments of Surgery and Pathology, Box 2626, Duke University Medical Center, Durham, NC 27710, USA. e-mail: email@example.com
To indirectly test the hypothesis that prostate-specific antigen (PSA)-based screening is biased against obese men due to haemodilution of PSA, and thus results in delayed diagnosis and poorer outcome beyond the biological link between obesity and aggressive prostate cancer.
PATIENTS AND METHODS
We sought to examine the association between body mass index (BMI) and the outcome of radical prostatectomy (RP) separately for men with PSA-detected cancers (cT1c) or with abnormal digital rectal examination (DRE) findings (cT2/T3), and stratified by year of treatment, using two large databases. We conducted a retrospective cohort study of 1375 and 2014 men treated by RP between 1988 and 2007 using the Shared Equal Access Regional Cancer Hospital (SEARCH) and Duke Prostate Center (DPC) databases. We evaluated the association between BMI and adverse pathological features and biochemical progression, using logistic regression and Cox proportional hazards models, adjusting for several clinical characteristics, respectively. Data were examined as a whole and as stratified by clinical stage (cT1c vs cT2/T3) and year of surgery (≥2000 vs <2000).
In both cohorts a higher BMI was associated with high-grade disease (P ≤ 0.02) and positive surgical margins (P < 0.001) and these results did not vary by clinical stage. A higher BMI was significantly associated with biochemical progression (P ≤ 0.03) in both cohorts. When stratified by clinical stage, obesity was significantly related to progression in both cohorts among men with T1c cancers (P ≤ 0.004) but not in men with cT2/T3 cancers (P > 0.3). Among men with T1c disease, the association between BMI and biochemical progression was limited to men treated in 2000 or later (P ≤ 0.002) and was not apparent in men treated before 2000 (P > 0.4).
Obese men with PSA-detected cancers and treated with RP since 2000 were at significantly greater risk of biochemical progression, while obese men treated before 2000 or diagnosed with an abnormal DRE were not at significantly greater risk of progression. These findings support the hypothesis that current PSA-based screening is less effective at finding cancers in obese men, leading to more aggressive tumours at diagnosis. Lowering the PSA threshold for biopsy among obese men might help to improve outcomes among this high-risk group.
body mass index
Shared Equal Access Regional Cancer Hospital
Duke Prostate Center.
Many prospective cohort studies have reported that obese men were 20–30% more likely to die from prostate cancer than were men of normal weight [1,2]. This increased risk of death from prostate cancer was noted in a large study from the 1960s, before PSA screening and effective local treatment, suggesting that obesity is associated with a more aggressive biology .
Among men with early-stage disease, we and others found that obese men were at greater risk of progression after radical prostatectomy (RP) [3–9]. Obese men have lower PSA levels [10–13], possibly related to haemodilution from a larger plasma volume . Lower PSA values mean that fewer obese men have an elevated PSA level and fewer will have a biopsy, possibly leading to a delay in diagnosis and more aggressive disease at diagnosis. If this is true, obese men with PSA-detected cancers would have dramatically worse outcomes than non-obese men. By contrast, obese men with DRE-detected disease might have only slightly worse outcomes, in line with historical data from the era before PSA testing . Moreover, in men whose initial PSA level was below the biopsy threshold but were aggressively and serially screened, the subtle differences of a 20% lower PSA level would have a greater impact on the delay in diagnosis than in men receiving their first PSA test, where many men have elevated PSA levels regardless of haemodilution. Indeed, a previous study of men with predominantly PSA-detected cancers treated with RP found that obesity predicted significantly poorer outcomes among more recently treated men, but not early in the PSA era .
Thus we sought to indirectly test the hypothesis that PSA-based screening is biased against obese men by examining the association between body mass index (BMI) and the outcome of RP separately for men with PSA-detected cancers (cT1c) and men with abnormal DREs (cT2/T3), and as stratified by year of treatment, using two large databases, i.e. the multicentre, multiethnic Shared Equal Access Regional Cancer Hospital (SEARCH) and the Duke Prostate Center (DPC) database. We previously found that obesity was significantly associated with progression within SEARCH when not stratified by clinical stage .
PATIENTS AND METHODS
After obtaining Institutional Review Board approval from each institution for data abstraction, data from patients treated with RP between 1988 and 2007 at the Veterans Affairs Medical Centers in West Los Angeles, Palo Alto, California, and Augusta, Georgia, and Durham, North Carolina were combined into SEARCH. Patients treated with preoperative androgen deprivation or radiotherapy were excluded. Of 1747 patients within SEARCH, we excluded 50 diagnosed from a TURP (clinical stage T1a/T1b), 207 with missing BMI data, and 25, 31, 53, and one due to missing data for PSA level, biopsy Gleason score, clinical stage and race, respectively, resulting in a study population of 1380 men. The RP specimens were sectioned per each institution’s protocol. Biochemical progression was defined as one PSA level of >0.2 ng/mL, two of 0.2 ng/mL, or secondary treatment for a high PSA level after RP.
After obtaining Institutional Review Board approval, we identified 4495 men who had RP at Duke University from 1988 to 2006. We excluded 985, 387, 842, 120 and 10 who had missing data for BMI, preoperative PSA level, clinical stage, biopsy Gleason score, and race, respectively, and 12 with clinical stage T1a/T1b, resulting in a study population of 2014 men.
The BMI was examined as a categorical variable of <25.0, 25.0–29.9, 30.0–34.9 and ≥35.0 kg/m2. The distribution of clinicopathological characteristics across BMI strata were compared using the chi-square test for categorical variables and anova for continuous variables. For all analyses, data were examined as a whole, stratified by clinical stage (T1c vs cT2/T3), and stratified by surgery year among men with T1c disease. We used logistic regression to estimate the odds ratios for BMI groups of the following binary pathological variables: pathological Gleason sum (≥7), positive surgical margins, extracapsular extension, and seminal vesicle invasion; few men had lymph node metastases. We adjusted for preoperative PSA level (continuous after logarithmic transformation), biopsy Gleason score (2–6, 3 + 4, and ≥4 + 3), age at surgery (continuous), race (black, white, or other), surgery year (continuous), clinical stage (cT2/T3 vs T1c), and centre (categorical; SEARCH only). We tested for trend by entering the median BMI of each category as a continuous term into the model and evaluating the coefficient using the Wald test. We tested for interactions between clinical stage and BMI by including both main effect terms and a cross-product term in the model, and evaluated the coefficient using the Wald test. We also tested for interactions with surgery year, i.e. before 2000 vs 2000 or later. The rationale for choosing this threshold was that within SEARCH, since 2000 the percentage of men with clinical T1c has changed only slightly (61% in 2000 vs 68% in 2006), suggesting similar screening practices during this period. Also, within DPC, PSA values have remained unchanged over time since 2000 (P = 0.14), further supporting that there was no stage/PSA migration during this more recent period. We performed sensitivity analyses using alternative thresholds from 1994 to 2002.
To estimate the relative risk of progression associated with BMI, indicator variables for BMI categories were entered into a Cox proportional hazards regression model adjusting for the preoperative characteristics, as described above. In separate analyses, we further controlled for pathological features, including Gleason sum, margin status, extraprostatic extension, seminal vesicle invasion, lymph node metastasis, and prostate weight (continuous after logarithmic transformation). As described above, we examined for interactions between BMI, clinical stage and surgery year.
The clinicopathological characteristics of both cohorts are shown in Table 1; in both cohorts a higher BMI was associated with younger age (P < 0.001), treatment in more recent years (P ≤ 0.001), and lower PSA level (P ≤ 0.008). In SEARCH (P = 0.03), but not DPC (P = 0.33), a higher BMI was associated with a higher biopsy Gleason grade. There were trends for black men to be more likely to be obese, although this only reached significance in DPC (P = 0.001) and not SEARCH (P = 0.10). In both cohorts, there was no significant association between clinical stage and BMI (all P > 0.40). When stratified by clinical stage, these associations in general remained unchanged, except for the association between more recent surgery and a higher BMI in the DPC cohort, which was only significant for men with clinical cT2/T3 (P interaction = 0.001).
Preoperative clinical and pathological characteristics of men undergoing RP in the SEARCH and DPC Databases
|Mean (sd) or (, median):|
|BMI, kg/m2|| 28.2 (4.9, 27.7)|| 28.2 (4.4, 27.6)|
|Age, year|| 61.0 (6.4)|| 61.9 (7.4)|
|PSA level, ng/mL|| 9.5 (8.4, 7.1)|| 8.3 (8.9, 6.0)|
| T1c|| 763 (55)||1563 (78)|
| T2|| 616 (45)|| 428 (21)|
| T3|| 1 (<1)|| 23 (1)|
|Biopsy Gleason score|
| 2–6|| 857 (62)||1447 (72)|
| 3 + 4|| 292 (21)||323 (16)|
| ≥4 + 3|| 231 (17)||244 (12)|
|Pathological Gleason sum|
| 2–6|| 562 (41)||905 (45)|
| 3 + 4|| 503 (37)||670 (33)|
| ≥4 + 3|| 301 (22)||438 (22)|
|Positive surgical margins|| 603 (45)||645 (34)|
|Extraprostatic extension|| 324 (24)||597 (30)|
|Seminal vesicle invasion|| 126 (9)||192 (10)|
In SEARCH, after adjusting for several clinical characteristics, an increased BMI was significantly associated with high-grade pathological disease (P = 0.002) and positive surgical margins (P < 0.001) among all patients (Table 2). The trends were similar when stratified by clinical stage, except the association between BMI and high-grade disease was not statistically significant for T1c cancers (P interaction = 0.08). Among men with T1c disease, there were no significant interactions between BMI and surgery year for predicting any pathological findings (all P interaction >0.10).
Table 2. Odds ratio (95% CI) of adverse pathological findings at the time of RP by BMI (relative to normal weight, BMI < 25 kg/m2) in the SEARCH and DPC databases, adjusted for age, race, biopsy Gleason sum, clinical stage, preoperative PSA level, year of surgery, and centre (SEARCH only)
|Pathological Gleason sum ≥7||0.002||0.11||0.002||0.08|
| 25–29.9||1.15 (0.84–1.57)||1.12 (0.73–1.71)||1.18 (0.74–1.86)|| |
| 30.0–34.9||1.72 (1.18–2.51)||1.40 (0.85–2.30)||2.41 (1.34–4.32)|| |
| ≥35.0||1.76 (1.06–2.94)||1.56 (0.79–3.06)||2.47 (1.07–5.70)|| |
|Positive surgical margins:||<0.001||0.01||0.003||0.69|
| 25–29.9||1.25 (0.93–1.66)||1.29 (0.86–1.91)||1.20 (0.78–1.83)|| |
| 30.0–34.9||1.52 (1.08–2.13)||1.29 (0.82–2.05)||2.05 (1.22–3.44)|| |
| ≥35.0||2.23 (1.41–3.54||2.48 (1.34–4.59)||2.15 (1.05–4.40)|| |
| 25–29.9||1.25 (0.88–1.77)||0.97 (0.59–1.59)||1.64 (1.00–2.70)|| |
| 30.0–34.9||0.96 (0.63–1.46)||0.77 (0.43–1.40)||1.19 (0.64–2.19)|| |
| ≥35.0||1.57 (0.93–2.64)||1.33 (0.65–2.68)||1.88 (0.86–4.11)|| |
|Seminal vesicle invasion:||0.23||0.13||0.68||0.44|
| 25–29.9||1.19 (0.72–1.98)||3.60 (1.27–10.20)||0.73 (0.39–1.36)|| |
| 30.0–34.9||1.07 (0.58–1.99)||2.44 (0.77–7.78)||0.80 (0.36–1.75)|| |
| ≥35.0 kg/m2||1.85 (0.85–4.02)||4.10 (0.96–17.57)||1.46 (0.55–3.87)|| |
|Pathological Gleason sum ≥7:||0.02||0.046||0.27||0.90|
| 25–29.9||1.11 (0.87–1.45)||1.01 (0.75–1.35)||1.68 (0.96–2.96)|| |
| 30.0–34.9||1.42 (1.04–1.94)||1.50 (1.06–2.11)||1.09 (0.55–2.17)|| |
| ≥35.0||1.40 (0.88–2.22)||1.19 (0.70–2.03)||2.43 (0.92–6.41)|| |
|Positive surgical margins:||<0.001||<0.001||0.52||0.13|
| 25–29.9||1.23 (0.94–1.60)||1.15 (0.86–1.56)||1.41 (0.78–2.54)|| |
| 30.0–34.9||1.47 (1.07–2.00)||1.54 (1.09–2.18)||1.01 (0.49–2.10)|| |
| ≥35.0||2.80 (1.80–4.36)||3.11 (1.88–5.13)||1.82 (0.69–4.81)|| |
| 25–29.9||0.97 (0.74–1.26)||0.96 (0.71–1.31)||0.95 (0.55–1.64)|| |
| 30.0–34.9||1.06 (0.77–1.47)||1.13 (0.78–1.63)||0.80 (0.40–1.59)|| |
| ≥35.0||1.55 (0.99–2.43)||1.33 (0.78–2.25)||2.14 (0.86–5.32)|| |
|Seminal vesicle invasion:||0.32||0.26||0.78||0.61|
| 25–29.9||1.29 (0.85–1.96)||1.62 (0.94–2.80)||0.89 (0.45–1.73)|| |
| 30.0–34.9||1.11 (0.66–1.86)||1.71 (0.90–3.25)||0.50 (0.19–1.30)|| |
| ≥35.0||1.60 (0.82–3.14)||1.44 (0.57–3.61)||1.78 (0.64–4.95)|| |
In DPC, after adjusting for several clinical characteristics, the overall findings were similar, with a higher BMI associated with high-grade pathological disease (P = 0.02) and positive margins (P < 0.001) among all patients (Table 2). Although when stratified by clinical stage these associations were only significant for T1c patients, there were no significant interactions between BMI and clinical stage (all P interaction ≥0.13). Among men with T1c disease, a higher BMI was associated with extracapsular extension (P = 0.003) among men treated in 2000 or later but not among men treated before 2000 (P = 0.17; P interaction = 0.005). The association between BMI and positive margins was stronger among men treated in 2000 or later (P < 0.001) than among men treated before 2000 (P = 0.24; P interaction = 0.07).
The mean (sd, median) follow-up was 4.8 (3.4, 4.2) years within SEARCH and 4.5 (3.2, 3.9) years within DPC. During this period 434 patients (32%) and 483 (24%) progressed within SEARCH and DPC, respectively.
A high BMI was significantly associated with a greater risk of recurrence in both cohorts (Table 3). When stratified by clinical stage, the association between obesity and poor outcome was only apparent in men with clinical stage T1c disease in both cohorts, although the interaction was not significant in either cohort. When men with T1c disease were further stratified by surgery year, the positive association between a higher BMI and worse outcome was apparent only for men treated since 2000 (P interaction; SEARCH = 0.01, DPC = 0.11). Among men with cT2/T3 disease, the BMI was not significantly related to recurrence among all men (Table 3) or after stratification by surgery year (data not shown) in either cohort. Overall, the results were little changed after further controlling for pathological findings (data not shown), except that BMI was less strongly linked with a poor outcome within DPC, although it remained significantly associated with an increased risk of recurrence among men with T1c disease treated since 2000 (P trend = 0.046).
Table 3. Relative risk (95% CI) of time to biochemical progression after RP by BMI (relative to normal weight, BMI < 25 kg/m2) adjusted for only preoperative clinical factors
|All patients*||0.004|| ||0.03|| |
| 25–29.9||1.25 (0.97–1.59)|| ||1.16 (0.92–1.46)|| |
| 30.0–34.9||1.46 (1.10–1.95)|| ||1.43 (1.08–1.88)|| |
| ≥35.0||1.59 (1.16–2.27)|| ||1.29 (0.87–1.92)|| |
|Stratified by clinical stage|
| 25–29.9||1.25 (0.86–1.84)|| ||1.13 (0.86–1.50)|| |
| 30.0–34.9||1.59 (1.04–2.43)|| ||1.63 (1.18–2.23)|| |
| ≥35.0||2.18 (1.24–3.83)|| ||1.46 (0.92–2.32)|| |
|T2 only||0.32|| ||0.81|| |
| 25–29.9||1.22 (0.88–1.69)|| ||1.19 (0.76–1.85)|| |
| 30.0–34.9||1.31 (0.88–1.93)|| ||0.87 (0.47–1.61)|| |
| ≥35.0||1.16 (0.65–2.10)|| ||1.05 (0.49–2.28)|| |
|Stratified by year of surgery among men with T1c disease|
| 25–29.9||1.10 (0.62–1.95)|| ||0.96 (0.67–1.36)|| |
| 30.0–34.9||0.87 (0.44–1.75)|| ||1.43 (0.95–2.15)|| |
| ≥35.0||1.27 (0.42–3.88)|| ||0.87 (0.45–1.70)|| |
|2000 or later||<0.001|| ||0.002|| |
| 25–29.9||1.55 (0.89–2.72)|| ||1.51 (0.93–2.46)|| |
| 30.0–34.9||2.49 (1.37–4.53)|| ||2.01 (1.18–3.41)|| |
| ≥35.0||2.95 (1.44–6.01)|| ||2.50 (1.26–4.98)|| |
In sensitivity analyses among men with T1c disease, BMI was significantly linked with a poor outcome in more recent men in both cohorts, where ‘more recent’ was defined using all possible thresholds from 1994 to 2002, whereas the BMI was not significantly associated with a poor outcome in any earlier time point (data not shown).
Obesity is associated with a greater risk of death from prostate cancer [1,2], and a greater risk of progression after RP [3–9]. We hypothesized that lower PSA values among obese men [10,11,14] make it more difficult to detect prostate cancer, resulting in delayed diagnosis, a later stage and more aggressive behaviour at diagnosis, contributing to worse outcomes. In the current study of two cohorts totalling nearly 3400 men treated with RP, obese men with PSA-detected cancers had a >40% excess risk of biochemical progression relative to men of normal weight. Among PSA-detected cancers, obesity was most strongly associated with a poor outcome in more recently treated patients. By contrast, obese men with an abnormal DRE had similar outcomes as men of normal weight, regardless of surgery year. The current findings lend further support to the hypothesis that PSA-based screening is biased against obese men. Consideration should be given to lowering the PSA threshold recommended for biopsy among obese men.
Many cohort studies, following nearly a million men prospectively from the 1960s and 1980s, found that obese men were 20–30% more likely to die from prostate cancer [1,2]. More recent series in the USA from the PSA era suggest that obese men are 45–55% more likely to develop metastasis or die from prostate cancer than are men of normal weight [16,17]. Thus, the strength of the association between obesity and prostate cancer mortality appears to be strengthening over time.
The 1960s represented an era before serum-based cancer screening, early detection, and aggressive curative therapy for prostate cancer. As such, these data are the best available to biologically link obesity with a modest but significant increase in disease aggressiveness . Added to this underlying biology, in the PSA era, diagnosis and treatment is biased against obese men. Specifically, many studies found that obese men have lower PSA levels, possibly due to haemodilution from a larger plasma volume [10–14]. We hypothesized that this represents a barrier to early detection and appropriate therapy. If true, men with PSA-detected cancers would have two reasons for a poor outcome, i.e. aggressive underlying biology and delayed diagnosis. By contrast, men with cancers detected due to factors unrelated to a lower PSA level would only have the biological explanation.
To indirectly test this hypothesis, we examined data from two well-described surgical cohorts and found in both series that obesity was significantly linked with a poor outcome, but only among more recently treated men with clinical stage T1c disease. Among all men, obesity resulted in a >40% excess risk of PSA recurrence, similar to other surgical series [3–8] and the increased risk of death from prostate cancer in the PSA era [16,17]. However, when only T1c patients treated since 2000 were examined, obesity was associated with more than a doubling of the risk of PSA recurrence, far exceeding observations from other series not stratified by stage or surgery year. In agreement with the current results, in data from Johns Hopkins obesity was not significantly associated with the outcome in the early PSA era, but there was a 90% increased risk of recurrence among obese men treated since 1995 . The present data combined with the Johns Hopkins data support the hypothesis that PSA-based screening is biased against obese men, contributing to the poor outcomes among obese men beyond that ascribable to the biological link between obesity and aggressive cancer.
An alternative explanation for the present results is that due to technical challenges in performing the DRE, more obese men classified as having T1c disease actually had T2 disease, and thus were at higher risk than their clinical stage suggested. In a previous SEARCH database study, the excess risk associated with cT2a, cT2b and cT2c disease relative to T1c was −8%, 81% and 89%, respectively . Thus, unless all obese men had undetected T2b/T2c or greater disease, this alone cannot explain the excess risk among obese men with T1c disease. Moreover, this cannot explain why the excess risk was only apparent in more recent patients. Finally, if true, it would be expected that obese men would be more likely to have T1c disease, which was not apparent in the present study. Thus, while there might be selective understaging of obese men, it is unlikely to completely explain the current findings.
In the present study many men with cT2/T3 disease might have been detected due to an abnormal PSA level with the DRE only performed afterwards. The degree to which this occurred or how this might have affected our results cannot be determined in this retrospective study. Whether the year 2000 is the optimal threshold to separate earlier vs later PSA eras is unknown, although the sensitivity analyses suggested that, regardless of the threshold chosen, more recently treated obese patients had worse outcomes and thus our conclusions were independent of the threshold. The technical difficulty of performing an adequate resection among obese men  might contribute to poor outcomes. Indeed, in the present study there were significantly higher rates of positive margins in obese men, possibly related to ‘iatrogenic’ margins. However, even after adjusting for pathological characteristics, including margin status, more recently treated obese men with T1c but not T2/T3 disease had worse outcomes. Our primary end-point was biochemical progression. Although early recurrence is associated with an increased risk of death from prostate cancer , future studies are needed using more distant end-points. We used BMI as a measure of obesity; alternative measures of adiposity might have given different results. The present study included only men treated with RP and thus selection bias in biopsy, referral and treatment patterns might have influenced our results. Future studies in different populations of patients are needed to confirm these results. Finally, although the current results support the hypothesis that PSA-based screening is biased against obese men, given that all men in our study had prostate cancer, we cannot address this directly.
In conclusion, in a study of nearly 3400 men undergoing RP, obese men with PSA-detected cancers treated since 2000 were significantly more likely to have a PSA recurrence, while among men treated before 2000 or those with cT2/T3 disease, obesity was not significantly associated with recurrence. These findings support the hypothesis that PSA-based screening is biased against finding cancers in obese men. Lowering the PSA threshold recommended for biopsy among obese men might help to improve outcomes among this high-risk population.
CONFLICT OF INTEREST