By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
What's known on the subject? and What does the study add?
Diabetes is known to be associated with a slightly lower risk of prostate cancer. Only a limited number of studies have examined the impact of diabetes on prostate cancer outcomes, with mixed results.
This study builds on our prior work showing that on the whole, diabetes is not a risk factor for progression to metastases after surgery. However, intriguingly we found a significant interaction with obesity for modifying the relationship between diabetes and progression. If confirmed in future studies, this suggests the mechanisms by which diabetes alters prostate cancer aggressiveness may differ in obese and non-obese men.
To examine the association between diabetes and metastasis risk after radical prostatectomy (RP) and to determine if race or obesity modifies this relationship.
Patients and Methods
Patients comprised 2058 US veterans with prostate cancer (PCa) enrolled in the Shared Equal-Access Regional Cancer Hospital (SEARCH) database and treated with RP between 1988 and 2010.
The association of diabetes with metastasis risk or secondary treatment rates was examined using Cox proportional hazards, adjusting for preoperative and, separately, clinical and postoperative findings.
The effect modification by race (black vs white) and obesity (body mass index [BMI] ≥30 vs <30 kg/m2) was tested via interaction terms.
Men with diabetes had higher BMIs and were more likely to be non-white (all P ≤ 0.001).
On multivariable analysis, diabetes was not associated with metastasis risk (P ≥ 0.45), but, among men with diabetes, longer diabetes duration was associated with higher metastasis risk (P ≤ 0.035).
When stratified by obesity, diabetes was linked with higher metastasis risk in obese but not in non-obese men (P-interaction ≤ 0.037), but there was no significant interaction with race (P-interaction ≥ 0.56).
Diabetes also predicted more aggressive secondary treatment among obese men but less aggressive treatment among non-obese men (hazard ratio 1.39 vs 0.63, P-interaction = 0.006).
Where applicable, results were similar for both pre- and postoperative models.
Diabetes was not associated with metastasis risk overall.
Stratification by obesity yielded significant differences, with diabetes linked to a fourfold higher metastasis risk in obese men, despite predicting more aggressive secondary treatment. Longer diabetes duration was also associated with increased metastasis risk.
Although diabetes affects nearly 30% of the US population >65 years, and prostate cancer (PCa) occurs most often in older men, the influence of diabetes on PCa progression has been poorly studied, despite the frequent co-existence of these disorders [1, 2]. Indeed, while the relationship between diabetes and risk of PCa diagnosis has been heavily examined, only eight studies have compared outcomes such as disease-specific mortality and recurrence in men with diabetes vs men without [3-10]. Among these, only four studies, including an earlier study from our group, had >600 men available for analysis [6, 8-10].
Within this limited literature, there is substantial disagreement: one study reported increased PCa-specific mortality in men with diabetes while others found no association [4-7]. To date, diabetes has not been shown to affect biochemical recurrence (BCR) after surgery or radiation [5, 8-10]. Interestingly, in a subgroup analysis from our previous study, we found that, among men undergoing radical prostatectomy (RP), obesity modified the effect of diabetes on PCa aggressiveness in white men (P-interaction = 0.006), leading to greater BCR risk in obese white men (hazard ratio [HR] 2.52, P = 0.002), but not in other patient groups .
Extending this work, we sought to determine if diabetes also influences time to metastasis, which is a more definitive clinical endpoint than BCR. To do this, we retrospectively examined the association between diabetes and PCa metastasis among men undergoing RP in the Shared Equal-Access Regional Cancer Hospital (SEARCH) database, using models adjusting for preoperative features and, separately, for clinical and postoperative findings. As our past work suggested that obesity and race might modify the relationship between diabetes and PCa aggressiveness, we further stratified by these characteristics and tested for interactions with diabetes . Finally, given that secondary treatments may influence progression to metastasis, we compared rates of postoperative radiation and androgen deprivation therapy (ADT) in men with diabetes vs men without diabetes.
Patients and Methods
After institutional review board approval, we combined data for 2892 men undergoing RP at Veterans Affairs Medical Centers (VAMC) in: West Los Angeles, CA; Palo Alto, CA; Augusta, GA; Durham, NC; and Asheville, NC, from 1988 to 2010 into the SEARCH database . Men receiving preoperative radiation or ADT were not enrolled into the SEARCH database. Information about diabetes status at surgery (yes vs no, and date diagnosed if yes) was obtained through medical chart review, while patients' weight and height, for the calculation of body mass index ([BMI] kg/m2), were taken through chart review as well but had been entered in the chart based upon direct measurement. In the SEARCH database, diabetes status was not known for 119 men, while an additional 434 men lacked data for calculating BMI; a further 281 men did not have complete data for at least one of the following covariates: age, preoperative PSA level, race, biopsy tumour grade, clinical stage, pathological tumour grade, margin status, extracapsular extension, seminal vesicle invasion, or lymph node involvement. These individuals were excluded from the study, leaving 2058 men ultimately available for analysis.
As a group, excluded patients were older than the analytical cohort and more likely to be white (all P < 0.001). Excluded patients also had higher preoperative PSA levels (median 7.3 vs 6.5 ng/mL), lower-grade tumours, more advanced clinical stage, higher rates of extracapsular extension and seminal vesicle invasion, and earlier year of RP (median 1995 vs 2003, all P ≤ 0.006). There were no significant differences in BMI, surgical margin status, or lymph node involvement (all P ≥ 0.21)
The patients received regular postoperative follow-up according to the attending physician's discretion. This typically involved PSA testing every 3 months in the first year after RP, every 4 months in the second year, every 6 months in the third year, and annually thereafter. Metastatic disease was identified most commonly by bone scan in response to rising PSA levels at the physician's discretion, with equivocal findings confirmed by additional imaging or biopsy. Patients without metastases were censored at last follow–up visit.
The association of diabetes with clinical and pathological characteristics was examined using Wilcoxon's rank-sum and chi-squared tests for continuous and categorical variables, respectively. The effect of diabetes on metastasis risk was tested using Cox proportional hazards. In the preoperative model, we adjusted for age (continuous), preoperative PSA (log-transformed, continuous), BMI (log-transformed, continuous), race (white, black, other), biopsy tumour grade (Gleason 2-6, 3+4, ≥4+3), clinical stage (cT1, cT2/T3), RP year (continuous), and centre. In the separate postoperative model, we adjusted for the same clinical features but replaced stage with operative findings including margin status, extracapsular extension, seminal vesicle invasion and lymph node status (negative, positive, not assessed) and replaced pathological tumour grade for biopsy grade. Finally, the relationship between diabetes duration and metastasis risk was explored by substituting months since diabetes diagnosis (log-transformed, continuous) for diabetes, yes vs no in our regression models.
Effect modification by race and obesity was explored by stratifying subjects into black vs white and obese vs non-obese (BMI ≥30 vs <30 kg/m2), respectively, and testing for interactions. This was done by including diabetes, the characteristic of interest, and their cross-product in each model. Non-white and non-black men were excluded from the racially stratified analysis owing to small sample size (n = 127 with one metastasis). The association of diabetes with secondary treatment (radiation or ADT) was examined using Cox proportional hazards adjusting for clinical and postoperative findings, and interaction with obesity was tested as before.
Additional sensitivity analyses were conducted by (i) excluding subjects receiving adjuvant therapy and (ii) examining all subjects with diabetes, including men diagnosed during follow-up. Finally, chart review of patients receiving supplemental insulin at the Durham VAMC showed that 97% had type 2 diabetes, which is consistent with the published prevalence of this disorder in the general population . Thus, given the rarity of type 1 diabetes in the present cohort, it was not possible to differentiate metastasis risk by diabetes type and all men with diabetes were grouped together. Statistical analyses were performed using Stata v11.2 (StataCorp, College Station, TX, USA).
Of the 2058 men in our analytical cohort, 379 (18%) had diabetes at RP. Men with diabetes had a higher median BMI (29.6 vs 27.3 kg/m2) and were more likely to be non-white (all P ≤ 0.001; Table 1). There was also a suggestion of lower mean PSA levels (6.0 vs 6.6 ng/mL) and slightly older mean age among men with diabetes (62 vs 61 years, all P ≤ 0.064; Table 1). Men with diabetes were more likely to have higher grade tumours at biopsy and on pathology (all P ≤ 0.002), but clinical stage and surgical findings, including margin status, extracapsular extension, seminal vesicle invasion and nodal involvement in patients where this was assessed, did not differ between groups (all P ≥ 0.14; Table 1). Finally, the median follow-up was similar for all non-recurrent men (n = 1398) at 49–52 months (P = 0.42; Table 1).
Table 1. Clinical characteristics of 2058 men from the SEARCH database.
Men without diabetes
Men with diabetes
n = 1679
n = 379
IQR, interquartile range. *For non-recurrent men (total n = 1398; men with diabetes n = 257; men without diabetes n = 1141).
Overall, diabetes at RP was not associated with metastasis risk (log rank P = 0.26; Fig. 1), and this remained true after adjusting for multiple characteristics in both pre- and postoperative models (all P ≥ 0.45; Table 2). On sensitivity analysis excluding men on adjuvant therapy (n = 62 with one metastasis), results were similar (data not shown). Lastly, among men with diabetes, longer duration of diabetes was related to increased metastasis risk in both pre- and postoperative models (all P ≤ 0.035).
Table 2. Multivariable-adjusted HRs and 95% CIs of metastasis risk associated with diabetes stratified by race and obesity status.
*Hazard ratios are for men with diabetes as compared with men without diabetes; For racially-stratified analyses, total metastasis do not equal 49 as one case of metastasis occurred in a non-white, non-black man, and was excluded. ‡,Adjusted for age, PSA, BMI, race, biopsy tumour grade, clinical stage, RP year and centre; §Adjusted for age, PSA, BMI, biopsy tumour grade, clinical stage, RP year, and centre; ¶Adjusted for age, PSA, BMI, race, RP tumour grade, margin status, extracapsular extension, seminal vesicle invasion, lymph node status, RP year, and centre; **Adjusted for age, PSA, BMI, RP tumour grade, margin status, extracapsular extension, seminal vesicle invasion, lymph node status, RP year and centre.
For all stratified analyses, results were similar for the pre- and postoperative models, with the former yielding more conservative effect size estimates. The more conservative estimates are presented below, while complete details are included in Table 2.
When stratified by race, the null association between diabetes and metastasis risk held for both black and white men (P-interaction = 0.88; Table 2); however, when stratified by obesity, diabetes was associated with higher metastasis risk in obese men (HR 3.98, P = 0.009) but not in non-obese men (HR 0.37, P = 0.19; Table 2). The formal test of interaction between diabetes and obesity was significant (P-interaction = 0.037; Table 2). When stratified by both race and obesity, diabetes was generally associated with higher metastasis risk in obese men regardless of race, but only the observation for obese white men was significant (HR 6.06, P = 0.008; Table 2).
In addition, we graphically explored outcomes by diabetes and obesity status. We found non-obese men with diabetes had the lowest metastasis risk, men without diabetes, whether obese or not, had intermediate risk, and men with diabetes and obesity together had the highest metastasis risk (log rank P = 0.007; Fig. 2). Multivariable analysis produced similar results (Table 3).
Table 3. Multivariable-adjusted HRs and 95% CIs of metastasis risk associated with obesity and diabetes status.
*Hazard ratios are for men with diabetes as compared with men without diabetes; †Determined by change in likelihood chi-squared ratio with 3 degrees of freedom; ‡Adjusted for age, PSA, race, biopsy tumour grade, clinical stage, RP year and centre; §Adjusted for age, PSA, race, RP tumour grade, margin status, extracapsular extension, seminal vesicle invasion, lymph node status, RP year and centre.Non-obese, BMI < 30 kg/m2; obese, BMI ≥ 30 kg/m2. Non-obese, BMI < 30 kg/m2; obese, BMI ≥ 30 kg/m2.
On further sensitivity analysis, when including all men diagnosed with diabetes during follow-up as having diabetes at baseline (additional n = 223 in the group with diabetes, with eight more metastases), the overall null association between diabetes and metastasis risk, significant interaction with obesity, and lack of significant interaction with race persisted for both pre- and postoperative models (data not shown).
Association of Diabetes with PCa Secondary Treatment
To explore if the higher metastasis risk observed in obese men with diabetes was attributable to less aggressive secondary treatment, we investigated if diabetes predicted receipt of radiation or ADT. As the decision to start secondary treatment depends heavily on pathological findings from surgery, we performed these analyses using the postoperative model exclusively. In obese men, we found a near-significant association between diabetes and higher secondary treatment rates (HR 1.39, P = 0.054) while, in non-obese men, diabetes was associated with lower treatment rates (HR 0.63, P = 0.006, P-interaction = 0.006; Table 4). After further excluding patients undergoing adjuvant radiation or ADT, these observations became stronger (HR 1.61 vs 0.59 for obese and non-obese men, respectively, all P ≤ 0.009, P-interaction = 0.001; Table 4).
Table 4. Multivariable-adjusted HRs and 95% CIs of PCa secondary treatment associated with diabetes stratified by obesity status.
*Hazard ratios are for men with diabetes as compared with men without diabetes; †Between diabetes status and obesity; ‡Adjusted for age, PSA, BMI, race, RP tumour grade, margin status, extracapsular extension, seminal vesicle invasion, lymph node status, RP year and centre. Non-obese, BMI < 30 kg/m2; obese, BMI ≥ 30 kg/m2.
Among the few studies examining diabetes and PCa outcomes, there remains significant disagreement, with one report noting higher PCa-specific mortality in men with diabetes while several others found no association [3-8]. Most recently, among men undergoing RP, we found that diabetes was associated with higher BCR risk in obese white men but not in other patient groups, suggesting race and obesity may modify the effect of diabetes on PCa progression .
In the present study, we found diabetes was not associated with metastasis risk after RP regardless of race. Although consistent with most previous studies concluding that diabetes does not influence BCR or PCa-specific mortality [5-10], our results differ from those of Hammarsten and Högstedt  who, in a cohort of 320 Swedish patients with T2-3 disease, found men with type 2 diabetes were more likely to die from PCa; however, this analysis did not control for age, PSA or BMI, despite the known relationship of these factors to both PCa progression and diabetes [4, 13-17]. Thus, it is unclear if the increased PCa-specific mortality was attributable to diabetes or a confounding factor.
Although diabetes was not associated with metastasis risk overall, when stratified by obesity, we found diabetes predicted a nearly fourfold higher metastasis risk in obese men (P = 0.009) while, in non-obese men, there was a suggestion that diabetes was protective (HR 0.37, P = 0.19). Moreover, the formal test of interaction between diabetes and obesity was significant (P-interaction = 0.037), suggesting synergy between these two disorders. Specifically, while we found that neither diabetes nor obesity independently raised metastasis risk, their co-existence predicted a higher likelihood of developing metastases (HR 2.80, P = 0.017). Interestingly, we found obese men with diabetes received more aggressive secondary treatment than obese men without diabetes; thus, inadequate therapy cannot explain the higher metastasis risk in obese men with diabetes. Conversely, among non-obese men, diabetes predicted less aggressive secondary treatment yet was linked with equal or lower metastasis risk.
These results support earlier findings from our group linking diabetes to increased BCR risk after RP, but only in obese white men (P = 0.002) . In our previous report, the formal test of interaction between diabetes and obesity was significant (P-interaction = 0.006), and there was also a suggestion that race may modify the effect of diabetes on BCR and PSA doubling time (all P-interaction ≤0.11) ; however, in the current analysis, we observed no significant interaction between race and diabetes. Our latest findings mostly agree with a recent study of 998 men undergoing biopsy at the Durham VAMC, a cohort partially overlapping that of the SEARCH database . In that study, while diabetes was linked with risk of high grade disease (Gleason > 7) overall, on additional subanalysis, this observation was significant only for obese white men (P = 0.038). In the same study, the interaction between diabetes and obesity approached significance (P-interaction = 0.088 in white men) whereas the interaction between diabetes and race did not (P-interaction = 0.60) . Finally, in a recent analysis from the REDUCE trial, we found a borderline significant interaction between diabetes and obesity for predicting high grade PCa (P-interaction = 0.053) . Taken together, these results suggest that obesity, but not race, modifies the effect of diabetes on PCa aggressiveness.
Possible biological mechanisms for our findings include changes in the hormonal milieu of obese men with diabetes, particularly decreased plasma concentrations of total and free testosterone, as well as elevated levels of insulin and IGF-1 [20-23]. While the role of androgens in PCa risk is controversial, with a recent large meta-analysis showing no association between plasma androgens and PCa risk, there are increasing data that among men with PCa, low androgen levels are associated with aggressive disease [24-29]. Whether this results from selection bias (i.e. only more aggressive tumours can grow in the low-androgen environment) or whether a low-androgen environment directly promotes the development of a more aggressive phenotype is unclear. As such, though a low androgen environment may certainly be one explanation for our observations, much more research is needed regarding the role of obesity, diabetes, androgens and PCa progression.
As diabetes and obesity both disturb endocrine homeostasis, their combined effect on growth factor signalling may explain our observations of higher metastasis risk in men with both disorders [30-32]. Notably, several previous studies have found elevated fasting insulin and C-peptide levels, a marker of insulin secretion, to predict higher PCa-specific mortality [4, 30]. Studies have also linked attributes of metabolic syndrome, which is characterized by insulin resistance and secondary hyperinsulinaemia, to aggressive PCa [4, 31, 32]. Together, these findings implicate insulin as a key promoter of PCa progression, which is particularly suggestive given hyperinsulinaemia is a hallmark of both obesity and early type 2 diabetes.
Interestingly, while early type 2 diabetes leads to insulin excess, advanced diabetes is characterized by diminished insulin secretion due to progressive pancreatic β-islet cell loss, so men with type 2 diabetes eventually become hypo- rather than hyperinsulinaemic. Indeed, the majority of reports examining the risk of PCa diagnosis (as opposed to progression, which was the focus of the present study) have noted a lower PCa risk with longer diabetes duration, supporting the hypothesis that a transition from a hyper- to hypoinsulinaemic state is protective [33-36]. Interestingly, we found herein that a longer duration of diabetes predicted a higher metastasis risk. Thus, while the hypoinsulinaemia of long-standing diabetes may reduce the risk of PCa diagnosis, our latest findings suggest that for men already diagnosed with PCa, long-standing diabetes may actually increase the likelihood of metastasis, perhaps by starving the tumour of needed growth stimuli and thereby selecting for the biologically fittest cells within. That a lesion is able to survive and expand in such a harsh environment to the point of being detected clinically would suggest it is likely to be highly aggressive. Since one consequence of inadequate insulin is poor glycaemic control, it is noteworthy that several other studies have indeed linked elevated HbA1c levels with higher grade tumours at RP [37, 38]. Our results showing that longer duration of diabetes is associated with a higher metastasis risk will, of course, require confirmation through other studies. If verified, however, the clinical implication is that men with long-standing diabetes who develop PCa may harbour a more aggressive variant, and this possibility should be kept in mind when determining treatment.
Unfortunately, a true comparison of metastasis risk between type 1 and type 2 diabetes could not be performed as the vast majority of our cases (97%) had type 2 disease. We also lacked information on diabetes severity and treatment. Since these factors might alter plasma insulin levels, such information would be important to future studies investigating the biological underpinnings of our epidemiological observations; however, these data were not available to us. Additionally, the present analysis accounted only for men with diabetes at the time of RP, which may have biased our results toward the null if many in the group without diabetes demonstrated glucose intolerance at surgery but were only formally diagnosed with diabetes later. Nevertheless, on sensitivity analysis including all men ever diagnosed with diabetes, results remained similar. Our groups also differed significantly in baseline clinical characteristics, with men with diabetes more likely to be non-white and to have a higher BMI and lower PSA levels; however, these differences accurately reflect the known demographics of type 2 diabetes, and we further controlled for these covariates in our regression models [1, 13, 14].
Another significant limitation was the rarity of metastatic disease (n = 49). Despite this, the present results are consistent with previous findings from largely the same population of men within the SEARCH database examining the association of diabetes with BCR risk, an intermediate endpoint predictive of progression to metastasis and mortality, for which substantially more events (n = 401) were available for analysis [10, 39].
Finally, the present analysis was retrospective in nature and prone to selection bias (specifically, it included only men neither too obese nor too ill to preclude surgery), changes in clinical practice over time, and unknown confounders.
The strengths of our study include the fact that it comprised a large, racially diverse cohort drawn from a single equal-access healthcare system, determination of BMI from direct measurements, and identification of diabetes through detailed chart review. The SEARCH database also did not include men receiving preoperative radiation or ADT, thus eliminating confounding by multiple treatment types. We further performed sensitivity analysis by eliminating men undergoing adjuvant radiation or ADT and observed that our results did not change. It should be noted, however, that our cohort included only men undergoing RP. Thus, our results may not be applicable to all men with PCa, and additional studies will be required to validate these findings in other patient populations such as men treated with external beam radiation or brachytherapy, or those enrolled in active surveillance. Nevertheless, given the paucity of studies on diabetes and PCa progression, our results shed valuable light on this important yet little-examined topic and suggest that diabetes in combination with obesity results in particularly aggressive PCa.
To conclude, in retrospective analysis, diabetes was not associated with metastasis risk after RP; however, when stratified by obesity, diabetes resulted in a nearly fourfold higher metastasis risk in obese men despite predicting more aggressive secondary treatment. By contrast, among non-obese men, diabetes resulted in a nonsignificant trend toward lower metastasis risk. Obesity, but not race, modified the effect of diabetes on metastasis risk. Finally, among men with diabetes, a longer duration of diabetes was associated with a higher risk of developing metastases. Our findings add to the limited research on diabetes, obesity, and PCa progression, suggesting men with diabetes and obesity together might harbour particularly aggressive PCa.
This work was supported by the American Urological Association Foundation/Astellas Rising Star in Urology Award (S.J.F.) and a Howard Hughes Medical Institute Medical Research Fellows scholarship (C.W.).