Metabolic Syndrome and Prostate Cancer Risk in a Population-Based Case–Control Study in Montreal, Canada

Audrey Blanc-Lapierre; Andrea Spence; Pierre I. Karakiewicz; Armen Aprikian; Fred Saad; Marie-Élise Parent


BMC Public Health. 2015;15(913) 

In This Article


In this population-based case–control study including some 4,000 subjects, we observed a significant inverse relationship between MetS and PCa, regardless of the criteria used to define MetS. This negative association did not vary according to PCa aggressiveness, and was particularly pronounced when MetS was developed at a young age (≤40 years). The analysis of the risk according to the number of MetS components suggests a dose–response relationship with MetS severity.

Selection Bias

Response rates could have affected results if socioeconomic characteristics associated with MetS influenced subjects' participation. However, according to Canadian census tract data for 2006, the rates for recent immigration, unemployment, low educational level and low household income were similar in living areas of participants and non-participants, both among cases and controls, indicating that selection bias is not of major concern in the study.

Detection Bias

Misclassification of PCa status due to under-detection is possible in both prospective and case–control studies. However, the present study is set in a population with very high PCa screening rates, thereby minimizing chances of a detection bias. Indeed, as a result of a universal access to health care in Montreal, study participants were relatively uniformly and regularly screened for PCa, whatever their socioeconomic position. However controls with MetS were more likely to have been screened recently, which may reflect a closer medical follow-up related to their condition. Nevertheless, this would have increased the probability of PCa diagnosis among MetS subjects, leading to underestimate a true negative association. Furthermore, the similar results observed in analysis restricted to subjects recently screened do not support an important impact of screening frequency on our findings.

Subjects with MetS are known to have decreased PSA levels,[30] which may result in a PCa diagnosis at a higher stage and/or grade. This could lead to differential misclassification and overestimation of a negative association between MetS and PCa, especially regarding localized and/or low-grade cases. However, our results did not change substantially when MetS risk was stratified according to PCa grade. Moreover, the association was of the same magnitude among subjects recently screened with DRE, which was shown to improve the predictive value of PCa screening among obese men.[32] Finally, controls with MetS were not less likely to have been referred for prostate biopsy compared to controls without MetS, and PSA levels among cases did not vary significantly according to the presence of MetS at diagnosis. Taken together, these observations argue against an important detection bias due to a decreased sensibility of PSA screening among subjects with MetS.

MetS Exposure Classification

A recall bias could be suspected in view of the study design, which would have led to underestimate a negative association. We were able to check the reliability of self-reported MetS-related conditions by referring to patient files in the hospitals where PCas were diagnosed. For respectively 84 %, 94 % and 93 % of cases reporting dyslipidemia, diabetes or hypertension at diagnosis, these disorders were also mentioned in patients' files, indicating that ascertainment of individual MetS components was reasonably valid. With regard to controls, there is little reason to suspect them of having over-reported MetS-related conditions, thereby driving results towards a protective association. Furthermore subjects and interviewers were blinded to the hypothesis under study and metabolic disorders were not the primary focus of the questionnaire.

Waist circumferences were measured by trained interviewers. While advanced stage PCa can lead to weight loss, only negligible weight losses were recorded in the two years preceding index dates. Moreover in contemporary newly diagnosed PCa, cancer-related weight loss is virtually unseen within two years of diagnosis. When WC was missing, we used the standard threshold for obesity, i.e., a BMI of 30, rather than a BMI value that better corresponds to a WC of 102 cm in our data, to facilitate comparison with previous studies.

Using NCEP-ATPIII criteria, we observed a MetS prevalence of 22 % and 34 % among controls aged 40–59 and ≥60 years, respectively. This prevalence was similar in the younger group but about 20 % lower in the older one in comparison with that reported in two Canadian surveys based on clinical values for blood pressure, TGs, HDL-C and blood glucose.[33,34] This may reflect a better participation of healthy subjects with limited cardiovascular risk factors in this age group. Indeed, MetS is an indicator for cardiovascular disease risk,[35] and competing risk of death from cardiovascular causes was found to bias the association towards negative values[36,37] or the null value[38] in cohort studies, and may bias the association towards the null value in case–control studies. According to this, our conclusion for a negative association was likely in the correct direction, albeit conservative in terms of magnitude. Besides, the negative association remained significant, and more pronounced with a young age at MetS onset, among subjects aged 65 years and more.

As documented in other investigations,[20] T2D was less frequent among PCa patients. However, inclusion of this condition in the MetS definition does not solely explain the inverse relationship observed. The prevalence of diabetes was low compared to other components (35 % of the subjects with MetS) and the negative association with MetS was still significant after adjustment for diabetes or exclusion of diabetic subjects.

About half of subjects with a history of dyslipidemia ever took statins. Statins use has been shown to be associated with a reduced risk of PCa, especially advanced ones.[39] However, statin intake did not contribute to the negative association observed with dyslipidemia in our sensitivity analysis.

Confounding/Effect Modification

We were able to consider a wide range of potential confounders. Lifestyle components, including physical activity and diet, did not emerge as important confounding factors in our study population. Subjects of sub-Saharan ancestry were found to have a distinct MetS profile, as observed previously in an American population.[40] However the risk of PCa associated with MetS was similar across ancestries.

Comparison With Previous Studies

The literature on the role of MetS in PCa development is divided. No association was found between MetS and PCa risk in two recent meta-analyses, one based on 14 studies (RR = 1.12 [0.93–1.35])[22] and another based on nine cohort studies (RR = 0.96 [0.85–1.09]).[23] However, there was a trend for a positive relation between MetS and PCa aggressiveness among PCa patients in another meta-analysis including seven studies (high grade PCa: OR = 1.36 [0.90–2.06], advanced PCa: OR = 1.37 [1.12–1.68])[23]

The conflicting findings across studies may relate to differences in PCa detection practices between populations. Positive associations were more often observed in Europe,[41–44] whereas negative associations, such as ours, were found in North America,[45,46] except among African Americans.[40,47] This trend might be explained by a less frequent systematic PCa screening in Europe,[48] where studies have observed a globally more aggressive cancer profile. Supporting evidence comes from a recent Swedish prospective cohort, using a composite score combining z scores of MetS components.[49] No association was observed with overall risk of PCa, while a positive association emerged for PCa mortality. However, in analyses restricted to cases diagnosed since 1997, the MetS score was significantly associated with a decreased PCa risk. This period was characterized by an increase in low-grade PCa incidence due to more frequent PSA testing. In a more recent analysis taking into account competing events, the decrease in PCa risk observed among men with metabolic disorders was more pronounced in the PSA era.[38]

Even when stratified on PCa aggressiveness, results observed in highly screened populations cannot be interpreted in the same way as in other populations. The PCa grade captured at the time of diagnosis depends on screening practices and does not take into account the whole history of PCa progression. For instance, the observed association between MetS and aggressive PCa[42,50,51] may ensue from delayed detection among subjects with MetS. Besides, non-screened controls may include non-detected cases, directly impacting case ascertainment.

Other reasons may explain discrepant findings across studies. Two studies and a multicenter clinical trial have been conducted on Spanish,[51] Canadian[52] and worldwide[53] patients referred for prostate biopsy. They observed an increased risk of high-grade PCa with MetS, but either no association or a positive one was found with overall PCa risk. In the three studies, PSA levels were elevated among participants,[51–53] and the prevalence of MetS was low in the clinical trial.[53] It is thus unclear how findings based on biopsy series compare to those from population-based studies, since exclusion of asymptomatic and untested subjects from control series may have compromised representativeness to the population base, including with respect to MetS prevalence.

Previous investigations have applied different MetS definitions. Interestingly, the four studies that used the NCEP-ATP III MetS criteria with a WC of 102 cm for abdominal obesity and that considered subjects with less than three MetS components as the reference group[46,50,51,54] observed, like us, an inverse association between MetS and PCa risk. While using different MetS definitions did not substantially alter our own findings, it may not be so when applying different criteria to populations presenting diverse anthropometric or clinical patterns.

Finally, methodological issues may be at play. Most previous studies were based on limited numbers of exposed cases. Our study was especially well-powered with some 2000 cases, almost 500 of whom having MetS. Three other recent powerful studies, with more than 300 exposed cases, were recently published on this topic, but these did not consider potential confounding by socioeconomic status, medical history (comorbidities, medication) and/or lifestyle.[49,51,52] In another study using a questionnaire-based MetS assessment similar to ours, and reporting a positive association especially in low-grade cases,[43] selection (or classification) biases can be suspected in view of the very low MetS prevalence and of socioeconomic differences between cases and hospitalized controls.

One salient advantage of this study is its ability to investigate the role of age at MetS onset in PCa risk which, to our knowledge, has never been investigated. Using a retrospective approach, we could trace the MetS history all along the potential period of PCa development. About half of our subjects with MetS had developed it after age 60. Prospective cohorts have typically relied on a single MetS assessment at baseline including participants of different ages,[36,42,45,49,52,54,55] or possibly too young to capture an eventual occurrence of MetS during adulthood.[41,42] We observed a stronger risk decrease with a young age at MetS onset. In a cohort of men aged 40–49 years, MetS defined using NCEP values was not predictive of PCa.[41] However diabetics, which account for 18 % of our subjects with prevalent MetS at age 40, were excluded from this former cohort.

Potential Mechanisms

Biological pathways involving low insulin, IGF-1 and testosterone levels have been suggested to explain a negative association between MetS and PCa.[13] The pronounced negative association observed with a young age at MetS onset may relate to the timing of diabetes occurrence. Indeed, a reduced risk of PCa is usually observed only several years after diabetes diagnosis,[56] probably because long-standing diabetics may experience low insulin levels in later years. Besides, the apparent contribution of dyslipidemia to the negative association observed raises the issue of the role of cholesterol in prostate malignancy, although some evidence suggests that it might be positively related with PCa growth.[57–60] Conversely, a cholesterol-lowering effect of cancer has been suggested, as a result of tumor metabolism. Low cholesterol-cancer associations have been mostly observed in studies conducted before introduction of PSA-testing, including more advanced PCas, or in PSA-screened populations where cholesterol was measured within one year of cancer diagnosis.[61] The high PSA-screening rate in our population and our application of a two-year-lag in the analysis between MetS assessment and PCa diagnosis do not support such a reverse relation. Future experimental research exploring potential biological mechanisms of MetS should consider the synergistic interaction of MetS components.