Cholesterol and prostate cancer: deconstructing a complex relationship

After more than a hundred years of research into the topic of cholesterol and abnormal cell growth, much debate and substantial doubt remain concerning the effect of hypercholesterolemia on prostate disorders. These unresolved issues include whether drugs used to treat hypercholesterolemia alter the risk of prostate cancer. Much of the controversy stems from limitations in the existing literature and interpretations thereof that go beyond the data. This review highlights some of these points of debate and draws contrasts between established findings, inconsistencies between randomized trials and observational studies, and what we believe to be misinterpretations of published data sets. What emerges from a careful analysis of a large body of literature, going back many decades, is an integration of older findings and data in newer reports that supports a conclusion that hypercholesterolemia is a risk factor for aggressive prostate cancer.

The prostate synthesizes cholesterol at rates equivalent to the liver and an age-dependent shift in cholesterol homeostasis allows cholesterol to accumulate in the prostate at high levels in older individuals. Cholesterol makes up about 30% of the lipid content of plasma membranes and is an absolute requirement for new membrane synthesis. This neutral lipid also contributes to the physiologic properties of cell membranes by regulating membrane fluidity, promoting negative membrane curvature, and by interdigitating with acyl chains of phospholipids to create liquid ordered membrane microdomains. Cholesterol-rich microdomains are believed to play an important role in signal transduction and in other physiologic mechanisms such as solute transport. Because of the multiple roles of cholesterol in the cell, perturbations in cholesterol metabolism might conceivably alter epithelial, stromal, and inflammatory cell infiltrates in the prostate. In a scenario involving rapid cell proliferation, the requirement for cholesterol assembly into new membranes may be a rate-limiting step in the process of tissue growth. Tumor cell proliferation, in turn, may affect circulating cholesterol levels. This article reviews various aspects of this complex relationship.

Cancer and its effects on circulating cholesterol: the U-shaped curve

The concept of the U-shaped curve refers to the higher levels of mortality found on either end of the cholesterol level spectrum that many investigators discovered when studies were performed to assess the potential relationship between circulating cholesterol and mortality. Although it is well established that the increase in mortality at the high end of the cholesterol spectrum is caused by cardiovascular disease, the higher level of mortality at the low end of the spectrum is more mysterious and seems to emanate from a variety of sources. These sources include an unexplained higher frequency of accidental deaths in individuals with low cholesterol and an association between endemic hepatitis and low cholesterol. However, another important reason for higher mortality at the low end of the U arises from the cholesterol-lowering effect of cancer itself; therefore, the low end of the curve likely includes an excess of individuals with preexisting cancer.

Although earlier articles speculated on this point, evidence implicating a direct effect of cholesterol lowering on increased cancer risk was first reported in 1971 by Pearce and Dayton. These investigators examined cancer incidence and cancer-specific deaths in 2 patient cohorts, one that received a control diet and the other an experimental cholesterol-lowering diet. The experimental diet was essentially the same as control, but with less cholesterol and with additional polyunsaturated fats. These investigators found that within 10 years of initiation of the study there was more total cancer (81 vs 66), more cancer deaths (31 vs 17), and, for our purposes, more prostate cancers (12 vs 10) in the experimental diet cohort, although the prostate cancer differences are small. A review of 5 diet trials published in 1971 did not support a cholesterol-lowering diet and cancer association (odds ratio [OR] for cancer incidence was 1.15 95% confidence interval [CI] [0.81–1.63]; for cancer death it was 1.08 95% CI [0.71–1.69]) and other, subsequent, cholesterol-lowering diet trials did not confirm the observations of Pearce and Dayton. A review of prestatin cholesterol-lowering trials in 1988, in which 22 randomized trials including ≈40,000 individuals were discussed (only 2 of which are cited), found no evidence suggesting an association between low cholesterol and cancer risk. This finding contrasts with a 1990 meta-analysis of 6 cholesterol-lowering trials, which used stringent criteria for selection of included studies. This review reported a significant increase in overall cancer mortality (OR 1.43 95% CI [1.08–1.90]) that remained even when studies not including drugs (2 studies) were used in the analysis (OR 1.62 95% [CI 1.03–2.57]).

Randomized, placebo-controlled or diet-controlled cholesterol-lowering trials were largely equivocal in their conclusions concerning a cholesterol-cancer association, and population studies were equally ambiguous. Rose and colleagues pooled data from 6 prospective population studies (more on these studies later) to show a low cholesterol-colon cancer link, with patients with cancer of the colon having on average a 10-mg/dL reduction in total cholesterol versus the population mean. The conclusions of this early study were buttressed by several additional population studies that showed excess cancers in the cohorts with the lowest cholesterol.

We have reviewed 52 population studies that reported on cholesterol and total cancer incidence and/or mortality that were published from 1972 to 2009. In total, these reports cover 79.5 million men and women ages 15 to 99 years (average 38–63 years) from Finland, Yugoslavia, the United States, New Zealand, Italy, France, Japan, China, Scotland, England, Norway, Israel, Australia, and Sweden. Thirty-two studies report an inverse association between cancer risk and cholesterol level, 16 show no association, 2 provided no statistics, 1 was a follow-up report, and 1 study was largely equivocal. Studies that report excess risk usually find this association more prominently in men and, most frequently, the associated cancers are those of the liver, colon, and lung (probably because of the high rates of mortality seen with these cancers). Many of these studies were long in duration (as long as 40 years); consequently, some of the investigators transformed the data by removing cancer cases that appeared in the early years of the studies (defined by investigators as those occurring anywhere from 2 to 20 years after study initiation). In 12 of a total of 30 of these reports, removal of cancers that appeared early in the study either diminished or eliminated the significance of the low cholesterol-increased cancer risk association, suggesting that lower cholesterol was not the cause but the result of cancer.

One intriguing aspect of these population studies that seems to strongly support the hypothesis that the inverse correlation between cancer and cholesterol level is caused by an effect of cancer on cholesterol level is that there is no absolute low level of cholesterol associated with cancer. For example, an association is found when the low cholesterol level for a population is ≤230 or ≤134 (ie, the 20% of any cohort with the lowest cholesterol in any population seems to have a greater prevalence of cancer than groups within that population with higher cholesterol). Logically, one would predict that if low cholesterol triggered cancer it would do so at a relatively uniform cholesterol level, and that if such an association existed, populations with low cholesterol would show an excess of certain cancers, an outcome that population studies do not support. Instead, if cancer reduces cholesterol level we would expect the 20% of the population with the lowest cholesterol to have excess cancer regardless of the endemic cholesterol level of the population, a prediction that is supported by the population studies when considered in aggregate.

A second point that seems to show that low cholesterol is the result of the effect of cancer on the host, and not the cause of cancer, arises from studies that measured cholesterol proximal to time of death. Keys and colleagues show that men who died of cancer within 2 years of study onset had cholesterol values 9.48% lower than the average of all men at entry. Sherwin and colleagues reported that men who developed cancer exhibited a 22.7-mg/dL drop in cholesterol level versus matched survivors. The International Collaborative Group showed that individuals dying from cancer 1 year after cholesterol measurement had cholesterol levels that were 24 to 35 mg/dL lower than controls (ie, those not dying); those dying 2 to 5 years after cholesterol measure reported values that were 4 to 5 mg/dL lower than controls; and those dying of cancer 6 to 10 years after cholesterol measure had cholesterol values 2 mg/dL lower than controls. These findings likely reflect the tendency of cancer to depress circulating cholesterol levels.

Particularly revealing are studies of cholesterol level variation over time in relation to cancer incidence or death. There are only a few of these reports because they require multiple cholesterol measures over time and few studies were designed to capture these data. The first such report we analyzed is by Sorlei and Feinleib, which used Framingham Heart Study data and showed that in some individuals who develop cancer, cholesterol levels are lower than the mean value up to 18 years before cancer diagnosis, whereas in other individuals cholesterol levels decline near the time of diagnosis. The robustness of the analysis is hampered because the investigators split the group by decades of age and sex, thus making each cohort small. The investigators write “Although the Framingham data are not conclusive, they do suggest that in some cancer cases where the serum cholesterol level was lower than that expected at as much as 16–18 years before cancer diagnosis, the depressed level was likely to be a precursor to the tumor growth. However, consistent with the metabolic consequences of tumor growth, the data show that in some cancer cases, serum cholesterol had decreased at measurements made close to the time of cancer diagnosis.” Pekkanen and colleagues, studying Finnish men (aged 55–74 years) analyzed the change in cholesterol levels from 1959 to 1974 (every 5 years) in individuals who did not have cardiovascular disease in 1974. The investigators found that older men (aged 65–74 years in 1974) with the steepest decline in cholesterol had a higher risk of dying from cancer. In 1992 Pocock and Seed reported cholesterol time trend data from British men and women with hypertension (aged 65–74 years) reporting that cholesterol levels fell an average of 11.2 mg/dL in men within a year before a cancer death. Sharp and Pocock, again using Framingham Heart Study data, report several highly relevant findings: (1) 61% of individuals have a 12.6 mg/dL (95% CI [8.46–16.70]) decline in cholesterol within 2 years before a cancer death; (2) the mean level of cholesterol 2 to 4 years before death from cancer is 10 mg/dL lower than the population norm; (3) the odds of dying of cancer within 4 years were increased (2.11 95% CI [1.41–3.14]) in individuals whose decline in cholesterol was greater than 38 mg/dL.

Whether low cholesterol causes cancer or low cholesterol is the result of cancer has important implications for public health. Not surprisingly, the National Institutes of Health (NIH) was concerned, especially in light of the large body of evidence that high cholesterol is associated with death from cardiovascular disease (nearly all the population studies of cholesterol and mortality verified this) and the international reaction to reduce cholesterol levels broadly in the human population. At least 3 NIH conferences were held to explore the relationship between cholesterol and mortality, and cancer mortality specifically: the first was held in February of 1980, and the second soon after in May of the same year ; a third was held in October of 1990. Because the report of Jacobs and colleagues is the most thorough, with its inclusion of 19 population studies in its analysis, we briefly explore this report with regards to cancer. After elimination of cancer deaths occurring within 5 years of the onset of each individual study, the cancer rate ratio for men with cholesterol less than 160 mg/dL was between 1.18 ( P <.05) and 1.23 ( P <.001), depending on how the studies were analyzed. For women the cancer rate ratio was 1.05 (not significant). Significant associations between low cholesterol and cancer were found for lung cancer in both men and women, and other cancers (not specified) in women. No association was found between low cholesterol and colon cancer. The summarized conclusion from the 3 NIH-sponsored conferences was that there were insufficient data regarding the association between cholesterol and cancer. They cite multiple concerns, including (1) the modest increase in cancer among the low-cholesterol population, (2) the lack of a clear association in women, (3) the presence of the cancer-cholesterol association in some populations but not in others, (4) the presence of a presumptive effect in some but not all studies, and (5) the absence of a plausible biologic mechanism that would explain why lower, but still physiologic, levels of cholesterol might trigger cancer. These considerations, as well as an absence of information about the effect of disease on cholesterol levels over time, prevented the conferees at any of the 3 meetings mentioned earlier from adopting specific recommendations on cholesterol reduction therapy. The recommendation of all 3 NIH conferences on low cholesterol and mortality was not to alter the prescription of lowering cholesterol levels to improve public health, but to continue to study the issue.

The concern that low cholesterol might increase cancer risk persisted until the early 1990s; however, it has almost entirely disappeared in the poststatin era. Multiple meta-analyses report that statins, which potently reduce cholesterol levels, do not cause an increase in cancer. This finding leaves the conclusion that cancer itself reduces circulating cholesterol as the only plausible explanation for the presumptive associations reported in the population studies described earlier.

What do the large population studies of overall disease incidence, mortality, and cholesterol level, in which prostate cancer was not an exclusive focus, indicate about prostate cancer, specifically? A small minority of the studies (10 of 52 publications) we examined for this review include prostate cancer in the analysis, and the total number of prostate cancer cases combined in the studies is only 1652. The following discussion summarizes all of the major population studies that include prostate cancer as an end point. Knect and colleagues reported a positive association between low cholesterol and increased prostate cancer risk (n = 45), but did not follow their cohort for greater than 5 years. Kark and colleagues found that individuals with prostate cancer (n = 12) had significantly reduced cholesterol levels. Hiatt and Fireman reported on 601 prostate cancers and found that removing cases occurring within the first 2 years after study inception eliminated the positive correlation between low cholesterol and cancer incidence. Williams and colleagues reported on 44 prostate cancer cases, but do not comment specifically on a cholesterol-prostate cancer association. Wingard and colleagues reported on 49 prostate cancer cases, but found no cholesterol-cancer association. Morris and colleagues found a significant low-cholesterol–cancer association (n = 26). These investigators show that the 5-year prostate cancer incidence rates decreased stepwise with lower to higher cholesterol levels (6.9 deaths/1000 cases when cholesterol was ≤204 mg/dL; 6.5/1000 in the 205–230 mg/dL quartile; 2.2/1000 in the 231–260 mg/dL quartile and 1.7 deaths/1000 cases in the ≥261 mg/dL quartile). Schatzkin and colleagues reported on 95 prostate cancer deaths and found no cholesterol-cancer association. Tulinius and colleagues describe no statistically significant association between cholesterol level and 524 incident prostate cancer cases. Davey-Smith and colleagues reported on 92 prostate cancer deaths and in quintile analysis report a nonsignificant trend, with lower levels of cholesterol associated with reduced cancer risk; the hazard ratio (HR) corresponding to a 1 standard deviation decline in cholesterol was 0.85 (95% CI 0.69–1.04). Iso and colleagues found no low cholesterol-prostate cancer association in their 164 cases; instead they report a significant trend of higher levels of cholesterol associated with increased cancer risk (p for trend 0.0023), an association that disappeared when advanced cancers were removed from the analysis (p for trend 0.12). Collectively, these studies point to a modest association between low cholesterol and increased prostate cancer risk.

In contrast to the reports cited earlier that include limited numbers of prostate cancer cases, for the most part were of short duration, and almost never report or comment on late-stage disease (except as it pertains to death), studies that specifically address the potential association between serum cholesterol level and prostate cancer have large enough cohort sizes to allow a more thorough analysis. Thompson and colleagues found no cholesterol-prostate cancer association (n = 100). The Asia Pacific Cohort Studies Collaboration reported on 308 prostate cancer deaths and found a greater number of deaths in the population with the highest cholesterol based on tertile analysis, but the difference was not significant (possibly because of small numbers). Platz and colleagues in a case-control analysis of men in the Health Professionals Follow-up Study reported that men (n = 698) with low cholesterol (Q1 in quartile analysis; cholesterol level is undefined because of assay complications) had a lower risk of high-grade prostate cancer (OR, 0.61 95% CI [0.39–0.98]). Platz and colleagues examined 1251 incident prostate cancers and found that men with low cholesterol (<200 mg/dL) had a lower risk of high-grade disease (Gleason 8–10; OR, 0.41 95% CI [0.22–0.77]). Mondul and colleagues examined 438 incident prostate cancers and determined that men with cholesterol lower than 240 mg/dL were less likely to develop high-grade prostate cancer then men with cholesterol greater than 240 mg/dL, results that were unchanged after eliminating cholesterol-lowering drug users. Batty and colleagues, in a study that included 578 prostate cancer deaths, reported a greater number of cancer-specific deaths in the highest cholesterol tertile (<175 mg/dL HR 1.0 [reference population]; 175–214 mg/dL HR 1.07 95% CI [0.87–1.32]; >214 mg/dL HR 1.35 95% CI [1.11–1.65] p-value for trend = 0.003). These reports do not support a low cholesterol-increased prostate cancer risk association, but instead suggest that men with high cholesterol are either at increased risk for prostate cancer or castrate-resistant disease.

Cancer and its effects on circulating cholesterol: the U-shaped curve

The concept of the U-shaped curve refers to the higher levels of mortality found on either end of the cholesterol level spectrum that many investigators discovered when studies were performed to assess the potential relationship between circulating cholesterol and mortality. Although it is well established that the increase in mortality at the high end of the cholesterol spectrum is caused by cardiovascular disease, the higher level of mortality at the low end of the spectrum is more mysterious and seems to emanate from a variety of sources. These sources include an unexplained higher frequency of accidental deaths in individuals with low cholesterol and an association between endemic hepatitis and low cholesterol. However, another important reason for higher mortality at the low end of the U arises from the cholesterol-lowering effect of cancer itself; therefore, the low end of the curve likely includes an excess of individuals with preexisting cancer.

Although earlier articles speculated on this point, evidence implicating a direct effect of cholesterol lowering on increased cancer risk was first reported in 1971 by Pearce and Dayton. These investigators examined cancer incidence and cancer-specific deaths in 2 patient cohorts, one that received a control diet and the other an experimental cholesterol-lowering diet. The experimental diet was essentially the same as control, but with less cholesterol and with additional polyunsaturated fats. These investigators found that within 10 years of initiation of the study there was more total cancer (81 vs 66), more cancer deaths (31 vs 17), and, for our purposes, more prostate cancers (12 vs 10) in the experimental diet cohort, although the prostate cancer differences are small. A review of 5 diet trials published in 1971 did not support a cholesterol-lowering diet and cancer association (odds ratio [OR] for cancer incidence was 1.15 95% confidence interval [CI] [0.81–1.63]; for cancer death it was 1.08 95% CI [0.71–1.69]) and other, subsequent, cholesterol-lowering diet trials did not confirm the observations of Pearce and Dayton. A review of prestatin cholesterol-lowering trials in 1988, in which 22 randomized trials including ≈40,000 individuals were discussed (only 2 of which are cited), found no evidence suggesting an association between low cholesterol and cancer risk. This finding contrasts with a 1990 meta-analysis of 6 cholesterol-lowering trials, which used stringent criteria for selection of included studies. This review reported a significant increase in overall cancer mortality (OR 1.43 95% CI [1.08–1.90]) that remained even when studies not including drugs (2 studies) were used in the analysis (OR 1.62 95% [CI 1.03–2.57]).

Randomized, placebo-controlled or diet-controlled cholesterol-lowering trials were largely equivocal in their conclusions concerning a cholesterol-cancer association, and population studies were equally ambiguous. Rose and colleagues pooled data from 6 prospective population studies (more on these studies later) to show a low cholesterol-colon cancer link, with patients with cancer of the colon having on average a 10-mg/dL reduction in total cholesterol versus the population mean. The conclusions of this early study were buttressed by several additional population studies that showed excess cancers in the cohorts with the lowest cholesterol.

We have reviewed 52 population studies that reported on cholesterol and total cancer incidence and/or mortality that were published from 1972 to 2009. In total, these reports cover 79.5 million men and women ages 15 to 99 years (average 38–63 years) from Finland, Yugoslavia, the United States, New Zealand, Italy, France, Japan, China, Scotland, England, Norway, Israel, Australia, and Sweden. Thirty-two studies report an inverse association between cancer risk and cholesterol level, 16 show no association, 2 provided no statistics, 1 was a follow-up report, and 1 study was largely equivocal. Studies that report excess risk usually find this association more prominently in men and, most frequently, the associated cancers are those of the liver, colon, and lung (probably because of the high rates of mortality seen with these cancers). Many of these studies were long in duration (as long as 40 years); consequently, some of the investigators transformed the data by removing cancer cases that appeared in the early years of the studies (defined by investigators as those occurring anywhere from 2 to 20 years after study initiation). In 12 of a total of 30 of these reports, removal of cancers that appeared early in the study either diminished or eliminated the significance of the low cholesterol-increased cancer risk association, suggesting that lower cholesterol was not the cause but the result of cancer.

One intriguing aspect of these population studies that seems to strongly support the hypothesis that the inverse correlation between cancer and cholesterol level is caused by an effect of cancer on cholesterol level is that there is no absolute low level of cholesterol associated with cancer. For example, an association is found when the low cholesterol level for a population is ≤230 or ≤134 (ie, the 20% of any cohort with the lowest cholesterol in any population seems to have a greater prevalence of cancer than groups within that population with higher cholesterol). Logically, one would predict that if low cholesterol triggered cancer it would do so at a relatively uniform cholesterol level, and that if such an association existed, populations with low cholesterol would show an excess of certain cancers, an outcome that population studies do not support. Instead, if cancer reduces cholesterol level we would expect the 20% of the population with the lowest cholesterol to have excess cancer regardless of the endemic cholesterol level of the population, a prediction that is supported by the population studies when considered in aggregate.

A second point that seems to show that low cholesterol is the result of the effect of cancer on the host, and not the cause of cancer, arises from studies that measured cholesterol proximal to time of death. Keys and colleagues show that men who died of cancer within 2 years of study onset had cholesterol values 9.48% lower than the average of all men at entry. Sherwin and colleagues reported that men who developed cancer exhibited a 22.7-mg/dL drop in cholesterol level versus matched survivors. The International Collaborative Group showed that individuals dying from cancer 1 year after cholesterol measurement had cholesterol levels that were 24 to 35 mg/dL lower than controls (ie, those not dying); those dying 2 to 5 years after cholesterol measure reported values that were 4 to 5 mg/dL lower than controls; and those dying of cancer 6 to 10 years after cholesterol measure had cholesterol values 2 mg/dL lower than controls. These findings likely reflect the tendency of cancer to depress circulating cholesterol levels.

Particularly revealing are studies of cholesterol level variation over time in relation to cancer incidence or death. There are only a few of these reports because they require multiple cholesterol measures over time and few studies were designed to capture these data. The first such report we analyzed is by Sorlei and Feinleib, which used Framingham Heart Study data and showed that in some individuals who develop cancer, cholesterol levels are lower than the mean value up to 18 years before cancer diagnosis, whereas in other individuals cholesterol levels decline near the time of diagnosis. The robustness of the analysis is hampered because the investigators split the group by decades of age and sex, thus making each cohort small. The investigators write “Although the Framingham data are not conclusive, they do suggest that in some cancer cases where the serum cholesterol level was lower than that expected at as much as 16–18 years before cancer diagnosis, the depressed level was likely to be a precursor to the tumor growth. However, consistent with the metabolic consequences of tumor growth, the data show that in some cancer cases, serum cholesterol had decreased at measurements made close to the time of cancer diagnosis.” Pekkanen and colleagues, studying Finnish men (aged 55–74 years) analyzed the change in cholesterol levels from 1959 to 1974 (every 5 years) in individuals who did not have cardiovascular disease in 1974. The investigators found that older men (aged 65–74 years in 1974) with the steepest decline in cholesterol had a higher risk of dying from cancer. In 1992 Pocock and Seed reported cholesterol time trend data from British men and women with hypertension (aged 65–74 years) reporting that cholesterol levels fell an average of 11.2 mg/dL in men within a year before a cancer death. Sharp and Pocock, again using Framingham Heart Study data, report several highly relevant findings: (1) 61% of individuals have a 12.6 mg/dL (95% CI [8.46–16.70]) decline in cholesterol within 2 years before a cancer death; (2) the mean level of cholesterol 2 to 4 years before death from cancer is 10 mg/dL lower than the population norm; (3) the odds of dying of cancer within 4 years were increased (2.11 95% CI [1.41–3.14]) in individuals whose decline in cholesterol was greater than 38 mg/dL.

Whether low cholesterol causes cancer or low cholesterol is the result of cancer has important implications for public health. Not surprisingly, the National Institutes of Health (NIH) was concerned, especially in light of the large body of evidence that high cholesterol is associated with death from cardiovascular disease (nearly all the population studies of cholesterol and mortality verified this) and the international reaction to reduce cholesterol levels broadly in the human population. At least 3 NIH conferences were held to explore the relationship between cholesterol and mortality, and cancer mortality specifically: the first was held in February of 1980, and the second soon after in May of the same year ; a third was held in October of 1990. Because the report of Jacobs and colleagues is the most thorough, with its inclusion of 19 population studies in its analysis, we briefly explore this report with regards to cancer. After elimination of cancer deaths occurring within 5 years of the onset of each individual study, the cancer rate ratio for men with cholesterol less than 160 mg/dL was between 1.18 ( P <.05) and 1.23 ( P <.001), depending on how the studies were analyzed. For women the cancer rate ratio was 1.05 (not significant). Significant associations between low cholesterol and cancer were found for lung cancer in both men and women, and other cancers (not specified) in women. No association was found between low cholesterol and colon cancer. The summarized conclusion from the 3 NIH-sponsored conferences was that there were insufficient data regarding the association between cholesterol and cancer. They cite multiple concerns, including (1) the modest increase in cancer among the low-cholesterol population, (2) the lack of a clear association in women, (3) the presence of the cancer-cholesterol association in some populations but not in others, (4) the presence of a presumptive effect in some but not all studies, and (5) the absence of a plausible biologic mechanism that would explain why lower, but still physiologic, levels of cholesterol might trigger cancer. These considerations, as well as an absence of information about the effect of disease on cholesterol levels over time, prevented the conferees at any of the 3 meetings mentioned earlier from adopting specific recommendations on cholesterol reduction therapy. The recommendation of all 3 NIH conferences on low cholesterol and mortality was not to alter the prescription of lowering cholesterol levels to improve public health, but to continue to study the issue.

The concern that low cholesterol might increase cancer risk persisted until the early 1990s; however, it has almost entirely disappeared in the poststatin era. Multiple meta-analyses report that statins, which potently reduce cholesterol levels, do not cause an increase in cancer. This finding leaves the conclusion that cancer itself reduces circulating cholesterol as the only plausible explanation for the presumptive associations reported in the population studies described earlier.

What do the large population studies of overall disease incidence, mortality, and cholesterol level, in which prostate cancer was not an exclusive focus, indicate about prostate cancer, specifically? A small minority of the studies (10 of 52 publications) we examined for this review include prostate cancer in the analysis, and the total number of prostate cancer cases combined in the studies is only 1652. The following discussion summarizes all of the major population studies that include prostate cancer as an end point. Knect and colleagues reported a positive association between low cholesterol and increased prostate cancer risk (n = 45), but did not follow their cohort for greater than 5 years. Kark and colleagues found that individuals with prostate cancer (n = 12) had significantly reduced cholesterol levels. Hiatt and Fireman reported on 601 prostate cancers and found that removing cases occurring within the first 2 years after study inception eliminated the positive correlation between low cholesterol and cancer incidence. Williams and colleagues reported on 44 prostate cancer cases, but do not comment specifically on a cholesterol-prostate cancer association. Wingard and colleagues reported on 49 prostate cancer cases, but found no cholesterol-cancer association. Morris and colleagues found a significant low-cholesterol–cancer association (n = 26). These investigators show that the 5-year prostate cancer incidence rates decreased stepwise with lower to higher cholesterol levels (6.9 deaths/1000 cases when cholesterol was ≤204 mg/dL; 6.5/1000 in the 205–230 mg/dL quartile; 2.2/1000 in the 231–260 mg/dL quartile and 1.7 deaths/1000 cases in the ≥261 mg/dL quartile). Schatzkin and colleagues reported on 95 prostate cancer deaths and found no cholesterol-cancer association. Tulinius and colleagues describe no statistically significant association between cholesterol level and 524 incident prostate cancer cases. Davey-Smith and colleagues reported on 92 prostate cancer deaths and in quintile analysis report a nonsignificant trend, with lower levels of cholesterol associated with reduced cancer risk; the hazard ratio (HR) corresponding to a 1 standard deviation decline in cholesterol was 0.85 (95% CI 0.69–1.04). Iso and colleagues found no low cholesterol-prostate cancer association in their 164 cases; instead they report a significant trend of higher levels of cholesterol associated with increased cancer risk (p for trend 0.0023), an association that disappeared when advanced cancers were removed from the analysis (p for trend 0.12). Collectively, these studies point to a modest association between low cholesterol and increased prostate cancer risk.

In contrast to the reports cited earlier that include limited numbers of prostate cancer cases, for the most part were of short duration, and almost never report or comment on late-stage disease (except as it pertains to death), studies that specifically address the potential association between serum cholesterol level and prostate cancer have large enough cohort sizes to allow a more thorough analysis. Thompson and colleagues found no cholesterol-prostate cancer association (n = 100). The Asia Pacific Cohort Studies Collaboration reported on 308 prostate cancer deaths and found a greater number of deaths in the population with the highest cholesterol based on tertile analysis, but the difference was not significant (possibly because of small numbers). Platz and colleagues in a case-control analysis of men in the Health Professionals Follow-up Study reported that men (n = 698) with low cholesterol (Q1 in quartile analysis; cholesterol level is undefined because of assay complications) had a lower risk of high-grade prostate cancer (OR, 0.61 95% CI [0.39–0.98]). Platz and colleagues examined 1251 incident prostate cancers and found that men with low cholesterol (<200 mg/dL) had a lower risk of high-grade disease (Gleason 8–10; OR, 0.41 95% CI [0.22–0.77]). Mondul and colleagues examined 438 incident prostate cancers and determined that men with cholesterol lower than 240 mg/dL were less likely to develop high-grade prostate cancer then men with cholesterol greater than 240 mg/dL, results that were unchanged after eliminating cholesterol-lowering drug users. Batty and colleagues, in a study that included 578 prostate cancer deaths, reported a greater number of cancer-specific deaths in the highest cholesterol tertile (<175 mg/dL HR 1.0 [reference population]; 175–214 mg/dL HR 1.07 95% CI [0.87–1.32]; >214 mg/dL HR 1.35 95% CI [1.11–1.65] p-value for trend = 0.003). These reports do not support a low cholesterol-increased prostate cancer risk association, but instead suggest that men with high cholesterol are either at increased risk for prostate cancer or castrate-resistant disease.

A question of prostate-specific antigen

The answer to the question of why some studies support an association between low cholesterol and prostate cancer, whereas others support an association with high cholesterol, most likely reflects when the study was conducted. The US Food and Drug Administration approved serum prostate-specific antigen (PSA) as a prostate cancer biomarker in 1994, forever changing the diagnostic landscape in the field. With PSA testing, men now generally present clinically with early stage disease, years before any clinical symptoms would otherwise appear. Thus, cancer populations considered in studies published before 1994 include many more advanced cancers than studies published in the last 10 years. This important milestone also suggests that cholesterol readings in the pre-PSA era have a greater chance of being a product of tumor metabolism, leading to a low cholesterol-cancer association, whereas cholesterol measures in post-PSA studies are more likely to reflect the cholesterol environment before the development of cancer. This finding would lead to a positive correlation between high cholesterol and prostate cancer risk.

We propose a unifying model that reconciles the data from the pre-PSA and post-PSA studies ( Fig. 1 ). The most recent evidence indicates that high circulating cholesterol is a risk factor for prostate cancer. In the pre-PSA era a low cholesterol reading is more likely to be associated with a higher risk of a prostate cancer death; in the post-PSA era this association is reversed. This contradiction can be explained by the different patient cohorts (with regard to extent of disease progression) analyzed in the older versus the new studies, by preclinical findings that high circulating cholesterol promotes prostate growth cancer, and by epidemiologic data showing higher cholesterol levels increase prostate cancer risk. Additional evidence includes the apparent protective effects of long-term statin drug therapy on prostate cancer risk (considered later).

Theoretic representation of the relationship between low cholesterol and the risk of prostate cancer death in the pre-PSA and post-PSA eras.

However, there are important caveats that alter this simple equation, and these are best understood by considering the period of time between a low cholesterol measure and a prostate cancer diagnosis versus the relative risk of prostate cancer death (see Fig. 1 ). In both the pre-PSA and post-PSA eras a low cholesterol measure within 1 year of a prostate cancer diagnosis raises the risk of a prostate cancer death, whereas in both the pre-PSA and post-PSA eras a low cholesterol measure more than 6 years before a prostate cancer diagnosis reduces the risk of a prostate cancer death (see Fig. 1 , left end of the curves vs the right end of the curves). Between 1 and approximately 6 years before a prostate cancer diagnosis the tendency for low cholesterol to correlate with increased risk of prostate cancer death is different in the pre-PSA and post-PSA eras. In the pre-PSA era a low cholesterol measure about 1 to 6 years before prostate cancer diagnosis raises the risk of a prostate cancer death, whereas in the post-PSA era a low cholesterol measure ∼1 to 6 years before a prostate cancer diagnosis decreases the risk of prostate cancer death.

The statin controversy

Recent epidemiologic studies from several groups have shown that cholesterol-lowering drugs (primarily 3-hydroxy-3-methylglutaryl coenzyme A [HMG CoA] reductase inhibitors, known as statins) may lower prostate cancer risk, and in particular, the risk of advanced disease. Platz and colleagues assessed potential statin drug effects specifically on prostate cancer in an analysis powered to measure differences in cancer incidence and progression in 2579 cancer cases, with 316 cases of advanced disease. The adjusted relative risk of castration-resistant cancer among statin users in this study was 0.51 95% CI (0.30–0.86) and of metastatic or fatal disease was 0.39 95% CI (0.19–0.77) for statin users versus nonusers. These investigators also showed that risk of advanced disease was lower with longer statin use (the relative risk for statin users of <5 years of use was 0.60 95% CI [0.35–1.03] and for ≥5 years of use was 0.26 95% CI [0.08–0.83]). In contrast to advanced disease, this study found no association between statin use and prostate cancer risk, suggesting that incidence alone is inadequate for evaluating potential chemopreventives in this disease.

Several more recent prospective studies from independent groups have largely confirmed the conclusions of Platz and colleagues that statins reduce the risk of aggressive prostate cancer. Flick and colleagues performed a case-control analysis using a population of 69,047 participants in the California Men’s Health Study that included 888 total cases of prostate cancer, with 131 advanced cases. They concluded that use of statins for ≥5 years was associated with a 28% lower disease risk (adjusted rate ratio of 0.72 95% CI [0.53–0.99]). Jacobs and colleagues reported on a case-control study using 55,454 men from the Cancer Prevention Study II Nutrition Cohort, which included 3413 cases of prostate cancer, with 317 cases of advanced disease. This group did not show a change in overall prostate cancer risk in the statin group; however, they found a marginally significant effect on advanced disease among the statin users (adjusted rate ratio of 0.60 95% CI [0.36–1.00]). Murtola and colleagues presented a case-control study of a large study population using data from the Finnish Cancer Registry, the Population Register Center, and the Social Insurance Institution of Finland (24,723 cases with an equal number of matched controls). This group found a significant reduction of risk of advanced prostate cancer in users of atorvastatin, lovastatin, and simvastatin (approximately 77% of the statin drug usage in the cohort), with an adjusted OR of 0.61 95% CI (0.37–0.98); 0.61 95% CI (0.430.85); and 0.78 95% CI (0.61–1.01), respectively (the overall OR for all statins was 0.75 95% CI [0.62–0.91]). Murtola and colleagues also noted that nonstatin cholesterol-reducing drugs (fibrates specifically) modestly reduced the risk of advanced prostate cancer, but these data did not reach significance, possibly a result of a small sample size. We have proposed that the chemopreventive effects of statin drugs arise predominantly or exclusively from cholesterol lowering, not from other statin effects, such as inhibition of isoprenoid synthesis ; consequently, if this hypothesis is true, we would expect that other methods of cholesterol reduction would also modify risk. Recent observational studies of statin effects on prostate cancer risk, which contain large numbers of subjects, are largely supportive of the hypothesis that statins reduce the risk of advanced prostate cancer.

Although the recent literature indicates that long-term statin therapy is chemopreventive against aggressive prostate cancer, large randomized trials of statin drugs that report on cancer (including prostate cancer) do not support this claim. There are several reasons that these 2 types of studies tend to present different pictures regarding prostate cancer risk and statin use.

Large randomized, placebo-controlled trials of statins include small numbers of prostate cancers. We reviewed 49 trial reports that included 134,516 individuals and identified only 5 prostate cancer deaths and 1142 incident prostate cancer cases, and these trials were of short duration (4.2 years on average). Observational studies of prostate cancer risk and statin use include many more prostate cancer cases (77,325 in a total study population of 4,168,049), and they include follow-up to 14 years. This situation explains some of the discrepancy between randomized, placebo-controlled studies and observational studies, but there are some additional important differences that deserve attention.