We found that plasma cholesterol was positively associated with increased risk of aggressive prostate cancer but not with overall risk of developing the disease in this population-based prospective cohort study. Our findings are consistent with others reported on United States populations [15
]. Similarly, data from the Swedish Apolipoprotein Mortality Risk (AMORIS) study reported no evidence of a relationship between hypercholesterolemia and overall prostate cancer risk [28
]. However, a large study of male Finnish smokers, reported a positive association between increasing total cholesterol level and overall risk of prostate cancer particularly advanced stage prostate cancer [29
]. The association further strengthened when they restricted the analysis to the cases diagnosed after 10 years from the baseline. However, this study did not find any association between aggressive disease, which may be because Gleason score was only available for 25% of the prostate cancer patients [29
Several underlying mechanisms by which cholesterol and prostate carcinogenesis may be linked have been proposed. Prostate cancer cells tend to over accumulate cholesterol in their cell membrane, forming large lipid rafts which in the cancer cells may facilitate pro-carcinogenic cell signalling [29
]. Moreover, several other pathways which are considered vital in carcinogenesis, such as sonic hedgehog and Akt pathways, are also cholesterol sensitive [32
]. Thus, having a lower cholesterol level may inhibit these pro-carcinogenic activities in the prostate cells.
Generally, an association has been reported between low cholesterol and increased risk of many cancer types and their associated mortality [9
] which has been ascribed to reverse causality; that is, early undiagnosed cancers lead to behavioural and physiological changes that reduce plasma cholesterol. The longer period between baseline cholesterol assay and diagnosis in our study (about 21 years for grade-specific analyses) compared to others suggests that reverse causality is unlikely to have been responsible for the observed association. Moreover, any such effect would have been expected to attenuate rather than exaggerate the association.
The potential clinical implications of our findings are that increasing obesity and associated dyslipidemia may have been responsible for the increasing incidence of prostate cancer and that modifying cholesterol may reduce incidence of more aggressive disease. The evidence that statins may reduce prostate cancer incidence remains equivocal. Two meta-analyses and a subsequent cohort study did not find any relationship between statin use and prostate cancer [34
]. However, Platz and Jacobs found associations between statin use and lower risk of advanced prostate cancer only and suggested that plausible biological mechanisms may existed, for example 3-hydroxy-3-methylglutaryl (HMG) coenzyme A reductase inhibition may reduce prostate cancer cell survival by interfering with membrane-associated signalling. In the absence of more consistent evidence on the effects of statins on prostate cancer, the most effective means of reducing incidence of the disease may therefore be through effective weight management.
Our research is based on one of the largest population based prospective studies in the United Kingdom and used cancer incidence data for the grade-specific analyses, rather than death records for cancer outcomes. Mortality data are a product of both incidence and case fatality, and do not allow risk factors to be individually differentiated. Furthermore, high cholesterol may increase the risk of death from other causes in prostate cancer patients and not necessarily be a causal factor for prostate cancer itself. Our study has larger numbers of incident cancers (n = 650), longer follow up and lower losses to follow up (0.1%) compared with earlier studies [15
]. However, our study has some weaknesses. The number of cases with aggressive PC was smaller; this may have influenced our results in some analyses. The Midspan questionnaire lacked information on family history of PC, prostate specific antigen screening and use of statins. We used plasma total cholesterol level because other measures of cholesterol, such as lipoprotein fractions (high and low density lipoproteins), were not available. Prostate specific antigen and disease stage data were not available which could be used to stratify the analyses based on localised and metastatic prostate cancer. The mean age at prostate cancer diagnosis is high and a large proportion of men are likely to die before diagnosis. The risk estimates we present might therefore have been affected by differential competing mortality risks. However, Cancer Registry data include Death Certificate Only diagnoses--and may have included, prostate cancers detected at post-mortem--that will attenuate such survival biases. The proportion of men who develop PC is higher among those who do not smoke, have a desirable BMI and are taller. The higher proportion of PC among these men results from those factors which confer a survival advantage. They live longer and therefore experience a longer risk time. However, the observation that height is associated with PC does raise the question whether some of those characteristics which promote longevity, are also associated with an increased risk of PC or whether such associations spuriously result from the influence of competing risk. Height is linked with development of many adult cancers [37
], a meta-analysis of 58 studies also suggested that height is positively associated with incident prostate cancer with stronger effect for advance stage and aggressive disease [40
]. There is some suggestion that height might be confounded by the socioeconomic status of the individual, however in our study height was associated with overall risk of prostate cancer independent of socioeconomic status.
Cholesterol, height and obesity are related to mortality from cardiovascular disease [27
] and early death from cardiovascular disease may be an important consideration. One possible scenario is that early cardiovascular disease mortality exhausts the pool of those men who would otherwise be susceptible to PC in later life, consequently systematic selection of more resilient individual may take place (men with low risk of PC but with high levels of traditional cardiovascular risk factors). This potentially could explain the positive association between height and overall incident PC, but would fail to explain the association between cholesterol and aggressive PC, which we report. A further consideration is detection bias and whether some groups are more likely to report their symptoms or have frequent medical examinations.
Further research is needed on cancer registry data to determine whether high-grade prostate cancer has differentially increased following increases in metabolic factors associated with hypercholesterolaemia. Longer-term follow-up of clinical trials of statins is also required, with exclusions of early tumours to minimise the potential effects of "reverse causality" on any association.