In 2014, pancreatic cancer (PC) represented the fourth-leading cause of cancer-related mortality among U.S. adults, with a projected 46,420 new cases and 39,590 deaths¹. Approximately 75% of patients die within 1 year following diagnosis and 5-year survival is 6%¹. Only 15–20% of patients have potentially operable tumors at diagnosis; the majority present with incurable locally advanced or metastatic disease². Population-based risk-reduction strategies require improved understanding of factors that modulate risk, particularly modifiable factors. Known and suggested risk factors include male sex, increasing age, African-American ethnicity, smoking, obesity, type 2 diabetes (T2D), pancreatitis, and family history of PC³.

Statin medications, currently indicated for coronary heart disease and its risk equivalents⁴, lower serum cholesterol levels via competitive inhibition of HMG-CoA reductase (the rate limiting enzyme in cholesterol synthesis). Recently updated preventive health guidelines vastly expand the cohort of U.S. adults deemed likely to benefit from statins⁵. Due to their pleiotropic effects, they have been of considerable interest for cancer prevention and treatment. Data safety analyses of early randomized control trials (RCTs) of statins revealed an inverse association between statin use and cancer incidence⁶. Pre-clinical studies have demonstrated growth suppressive effects on various tumors, and epidemiologic studies have shown inverse associations with overall cancer risk^{7, 8} and risk of other gastrointestinal cancers including esophageal⁹, colorectal¹⁰, and liver¹¹.

Previous studies of statin use and PC risk are inconsistent. Some RCTs and cohorts were underpowered^12–17 and of non-trial studies^7,18–26, few examined associations by sex with the exception of a large study of predominantly male veterans²⁶ and a UK study¹⁹ where inverse relationships were observed only in men and male smokers, respectively.

To further examine the association between use of statin medications and PC risk among women and men combined and separately, we analyzed data collected in our large clinic-based case-control study of PC in the San Francisco, California Bay Area. As anti-neoplastic effects may vary with drug characteristics^{27, 28}, we also explored differential effects of statins individually and grouped by pharmacologic properties.

Of potentially eligible cases aged 21–85 years, ~12% were ineligible due to language problems and ~3% were cognitively impaired, not located or dead, leaving 698 eligible PC cases. Of these, 16% refused, 6% were too ill and 2% expressed privacy concerns or participation in another study, for a final participation rate of 76%. Cases were recruited at a median of 61 days after PC diagnosis (interquartile range 25–148 days). Among controls, ~34% of patients approached did not meet eligibility requirements e.g. language problems, incompatible age-group. Of eligible clinic controls, 35% had no time, 3% refused, 6% were too ill and 3% had privacy concerns, for a final participation rate of 53%. These analyses include the eligible 536 cases and 869 frequency-matched controls who completed interviews.

Compared with controls, cases were slightly older and a greater proportion were men, consumed alcohol, were overweight or obese, or had T2D or pancreatitis (Table 1). Nearly half of both cases and controls reported a history of hypercholesterolemia.

Hypolipidemics were ever used by 34.0% of cases and 36.9% of controls (Table 2). Statins were the most commonly used (32.6% cases, 35.7% controls) and 23.9% of statin users took two or more different statins. Both cases and controls with T2D were more likely to have hypercholesterolemia and use cholesterol-lowering drugs including statins. Of note, one-third of the 39 statin users without hypercholesterolemia had T2D and 180 patients with hypercholesterolemia never used statins.

Ever use of statins was associated with a reduced PC risk (adjusted OR=0.66, 95%CI 0.47–0.92, Table 3). Sex-stratified analyses showed this was mainly due to the association in men (men: adjusted OR=0.50, 95%CI 0.32–0.79; women: adjusted OR=0.86, 95%CI 0.52–1.43). In contrast, use of other hypolipidemics was not associated with PC risk regardless of exclusivity of use, although estimates were imprecise.

Duration of statin use was inversely associated with risk, particularly among long-term users (used >10 years: fully adjusted OR=0.51, 95%CI 0.31–0.85, p_trend=0.01, Table 3). Age at first use was not associated with PC risk in fully adjusted models and did not confound the association between PC risk and duration of use (Table 3). Median age at first use in men, 57 years (interquartile range 50–64), was similar to that in women, 58 years (interquartile range 52–65) (data not tabled). In sex-stratified analyses, a trend of reduced risk with increased duration of use was statistically significant in men only (men, p_trend=0.006; women p_trend=0.44; Table 3).

Atorvastatin was the most common exclusively-used statin (Table 4). Only exclusive pravastatin use was associated with a statistically significant decreased PC risk (fully adjusted OR=0.22, 95%CI 0.06–0.82) and the magnitude of the OR was similar for men and women, although imprecise. Use of multiple statins was associated with a 56% reduced PC risk, statistically significant in men only (OR=0.30, 95%CI 0.10–0.86). Among those who took multiple statins, 86.2% used atorvastatin and 72.4% used simvastatin.

Exploratory analyses of the pharmacologic characteristics of statins disclosed few associations with PC risk (Table 5) with the exception of drug bioavailability. Compared with exclusive use of statins with low bioavailability (≤5%), those who exclusively used high bioavailability (≥12%) statins had reduced risk of PC (p=0.01). However, this was observed in men only (men: p=0.01, women: p=0.26). Interestingly, participants in the “mixed” use group were at lowest risk and were the only group with statistically significantly decreased risk compared with non-users. Other pharmacologic characteristics were not associated with risk when compared with non-users or in comparisons of high and low intensity exposure. However, those in the “mixed” use groups for derivation, renal excretion, and elimination had statistically significantly reduced PC risk compared with non-users.

To our knowledge, this is the largest published case-control study to demonstrate an inverse association between statin use and PC risk, and to conduct detailed analyses by sex and pharmacologic properties. Associations were observed in men only. PC risk was inversely associated with duration of statin use and lowest in men long-term users regardless of age at first use. The decreased risk of PC with exclusive pravastatin use was novel to our study but requires confirmation and should be interpreted cautiously.

Statins are hypothesized to decrease cancer risk partly via the downstream effects of HMG-CoA reductase inhibition in the mevalonate pathway. Specifically, inhibition disrupts synthesis of cholesterol and farnesyl or geranylgeranyl disphophates (FPP, GGPP) which function in prenylation of the G-proteins Rho and Ras, as well as other proteins involved in cell signaling^6,30,31. Relevant to pancreatic cancer, aberrant Ras signaling is integral to pancreatic tumorigenesis^32,33 whereas Rho mediates epidermal growth factor signaling and is implicated in other cancer-related mechanisms including angiogenesis^34,35, tumor invasion and metastasis^36,37, and activation of the NF-κB pathway³⁸. Statins also might impact cancer development via other direct anti-inflammatory and immunomodulatory effects^6,39. Interestingly, results from recent meta-analyses of RCT data show that statins may reduce risk of acute pancreatitis and increase diabetes risk⁴⁰, conditions associated with increased PC risk. The association between statin use and cancer is complex, organ dependent and confounded by intermediary health conditions requiring carefully designed studies to better understand the mechanisms driving the observed pleiotropy.

Our findings of sex-specific associations expand on results from a large nested case-control study of predominantly male U.S. veterans that showed statistically significantly reduced PC risk in statin users (OR=0.33), particularly long-term users (>4 years; OR=0.20)²⁶, as well as a recent UK case-control study that reported a reduced risk of PC with statin use among male smokers (OR=0.11)¹⁹. Other studies that reported null associations with PC risk^{7,18,20,21,25} typically had matched on sex and had too few cases to conduct sex-stratified analyses. However, earlier results for all-cause mortality and stroke where decreased rates were observed in men only^41,42 provide additional evidence of statins’ sex-specific effects. Integration of the accumulating data suggests sex-related effects of statins, including potential anti-neoplastic associations. Underlying mechanisms to explain these differences are uncertain and poorly understood. With few published studies among women, further research is warranted.

Our detailed analyses of pharmacologic properties of statins are unique to our study. Results were null from earlier analyses that assessed medication derivation or solubility^43,44. Our observed inverse associations with bioavailability are intriguing but could be spurious given that the lowest PC risk was observed among those in the “mixed” group of users. Overall, our results do not suggest that specific pharmacologic characteristics might confer exceptionally low PC risk.

Strengths of this study include the large sample size, the short duration between case diagnosis and interview, and the experienced, well-trained interviewers who administered a structured questionnaire in-person to collect data about potential and known confounders, effect modifiers and risk factors of PC in a standard manner. Data about most exposures excluded the year before diagnosis/interview to diminish effects of reverse causation. Cancer registry and medical record data were used to confirm PC diagnoses. Study limitations include the potential for recall bias inherent in case-control studies although direct interviews and use of cue cards to facilitate recall helped to diminish exposure misclassification. A potential for selection bias among controls is possible as a high frequency of controls “had no time” or could only partially complete an interview during the clinic visit. Our use of clinic-based controls, which compared with population-based controls may include a greater prevalence of unhealthy persons, i.e. smokers, also may have influenced our findings although our controls were largely being seen for acute conditions or healthy annual exams. Also, UCSF is a tertiary care center and cancer patients are often referred for surgery or clinical trial consideration, options that generally necessitate good functional status. Compared with 2006–2011 San Francisco-Oakland SMSA SEER data⁴⁵ for pancreatic cancer patients, a greater proportion of study patients were non-Hispanic white (85% vs 61%), were younger at diagnosis (median age 63 vs 70 years), and had earlier stage disease at diagnosis (regional stage, 42% vs 34%; advanced stage 41% vs. 56%). Thus, our results may pertain to a healthier population of PC patients than in the broader community. Finally, statins have been previously associated with increased adherence to preventive health measures⁴⁶, suggesting that statin use may be a surrogate for a healthier lifestyle or better functional status. This potential bias should be considered in the interpretation of our findings, although potential confounding due to lifestyle factors associated with increased risk of PC such as obesity, smoking and alcohol consumption was adjusted for in our analyses.

In conclusion, this represents the largest case-control study to demonstrate an inverse relationship between statin use and PC risk, particularly in men and in long-term users. Prospective clinical evaluation of statins as preventive therapy, e.g. in individuals at particularly high risk of PC, represents an intriguing possibility.