|Home | About | Journals | Submit | Contact Us | Français|
HMG Co-A reductase inhibitors (statins) have shown inverse associations with cancer risks, but the results have been inconsistent. Since there is no previous published data in brain tumors, we conducted a case-control study to investigate statin therapy and risk of glioma. We also further evaluated the use of nonsteriodal anti-inflammatory drugs (NSAIDs) in the risk of these tumors. We recruited newly diagnosed glioma cases and frequency matched controls at Columbia University and the University of California San Francisco. Standardized questions on statins and NSAIDs were used at both institutions. Intakes of these drugs were defined as > 6 months of at least twice weekly use versus less than this amount or never use. From July 2007 to January 2010, we recruited a total of 517 cases and 400 controls. Simvastatin and lovastatin showed significant inverse associations with glioma (OR = 0.49, 95% CI 0.30, 0.81 and OR = 0.47, 95% CI 0.24, 0.93, respectively). In NSAIDs, aspirin use was also inversely related to glioma risk (OR = 0.68, 95% CI 0.49, 0.96). Both statins and NSAIDs showed significant inverse trends between duration of drug use and glioma risks (trend tests p = 0.03 and p = 0.02, respectively), and drug intake > 120 months demonstrated the most significant associations for both types of medication. The inverse association between statin therapy and risk of glioma supports the roles of Ras/Rho GTPases or inflammatory cytokines in gliomagenesis, and similar relationship between NSAIDs and glioma highlights the importance of cyclo-oxygenase-2 in glioma pathogenesis.
The diagnosis of glioma carries a poor prognosis despite recent therapeutic advances. With standard of care treatment, glioblastoma multiforme (GBM, WHO Grade IV) has a median survival of approximately 14.6 months.1 With the exception of ionizing radiation, many past observational studies failed to establish consistent environmental determinants for this disease.2
The effect of commonly prescribed medications on the risk of cancers is actively being investigated in systemic cancers. An example is HMG Co-A Reductase Inhibitors (statins).3 To date, some studies had shown an inverse correlation between long-term statin use and risk of systemic cancers, but results had been conflicting. For example, one population-based cohort study evaluated long-term statin use and incidence of 10 common cancers among 133,255 participants.4 Statin use for five or more years was significantly associated with lower risk of melanoma, endometrial cancer and non-hodgkin lymphoma, but long-term use of these drugs did not change the incidence of all cancers. However, another analysis of 62,842 subjects using the New England Veterans Integrated Service Network-1 (VISN-1) pharmaco-epidemiology database found that statin users had a statistically significant lower risk for all cancers than nonusers after adjustment for age and multiple confounders.5 Similarly, conflicting results were reported in many individual cancer sites. For example, The Molecular Epidemiology of Colorectal Cancer Study showed a significant inverse association between the use of statin and colorectal cancer risk (CRC),6 but subsequent studies failed to confirm an inverse relationship with CRC, or an inverse association was only observed in stage IV colon cancer.7,8
Other than statins, NSAIDs have been extensively evaluated in the risk of cancers.9 Although observational study results are also conflicting with regard to cancer risks, many studies consistently demonstrated NSAID use is associated with a reduced risk of colon cancers.10 Two prior case-control studies suggested an inverse relationship between NSAIDs use and glioma risk.11,12 None has directly examined a duration-response relationship. Moreover, the two previously published pilot studies used different definitions of NSAID intake.
To date, there has been no epidemiologic report on statin intake and glioma risk. Therefore, we conducted a case-control study at Columbia University Medical Center (CUMC) and the University of California San Francisco (UCSF) to evaluate the association between statin use and risk of glioma and further investigated a duration-response relationship between NSAID intake and these tumors. We hypothesized that medication intake and duration of use for both types of drugs were inversely related to the development of glioma.
This study was approved by the Institutional Review Board of each institution. At CUMC, case eligibility criteria were: 1) age at or over 20; 2) histologically confirmed glioma (International Classification of Diseases for Oncology 3rd edition (ICD O-3) codes 9380-9460); 3) initial contact within six months of diagnoses. Spouses were permitted to serve as proxies if cognitive impairment precluded patients from participating. Newly diagnosed cases were first approached either in the hospital or during their initial visits to the neuro-oncology clinic.
At UCSF, population-based cases in the six San Francisco Bay Area counties (Alameda, Contra Costa, Marin, San Mateo, San Francisco and Santa Clara counties) were ascertained using the Northern California Cancer Center’s rapid case ascertainment system. Case eligibility criteria were the same as at CUMC. Physicians of eligible patients were contacted to assure no contraindications for patient contact. Then, if the patient did not refuse by postcard or telephone, an interviewer telephoned the subject to arrange an in-person interview. Proxy responses were also allowed for subjects with cognitive impairment or who were deceased. The number of proxy responses was detailed in Table 1.
At CUMC, the control group consisted of hospital visitors who could be family members or friends of hospitalized patients in CUMC. Visitors to intensive care units and post-operative areas were excluded due to medical and emotional intensity of those areas. Visitors with a pre-existing diagnosis of brain tumors were excluded. The research assistant systematically surveyed all regular hospital wards for recruitment. The control group was frequency matched (1:1) to cases according to age, gender and ethnicity.
At UCSF, population-based controls were identified through random digit dialing, using methods described by Waksberg and refined by Harlow and Davis.13 They were frequency matched (1:1) on the same factors as Columbia’s. Once eligible controls were found, the research assistant sent a letter to them and then telephoned them to arrange an in-person interview. Eligible controls were residents in the San Francisco Bay Area.
The participation rates for cases at CUMC and UCSF were 82% and 64%, respectively, and for controls, 86% and 77%, respectively.
We used an interviewer-administered questionnaire at both centers. Subjects were interviewed to obtain demographic information, family history of cancer, medical history, occupational, smoking, alcohol intake and physical activities. Medication intake and important demographic questions were standardized between CUMC and UCSF.
Ever use of statin and NSAID intakes were defined as taking drugs at least twice a week for 6 months or longer. This definition was chosen in accordance with other published pharmaco-epidemiological studies in the literature.14 Ever use was dichotomized as ≤ 6 months and > 6 months. The interview also included questions relating to duration of drug use (months), which was categorized as ≤ 6 months, 7 – 24 months, 25 – 60 months, 61 – 120 months and > 120 months. Statins that we studied were: simvastatin, lovastatin, atorvastatin, pravastatin and rosuvastatin. NSAIDs that we included were aspirin, ibuprofen and naproxen. Although acetaminophen inhibits prostaglandin synthesis, it possesses little anti-inflammatory property. Nevertheless, we have included it in regression models of NSAIDs for a comparison. The associations between statins/NSAIDs and glioma were adjusted for age, gender, ethnicity, educational level, NSAID use (or statin use when evaluating NSAIDs) and study center. It was necessary to adjust for NSAID when evaluating statins (and vice versa) because one third of subjects who took statins also had used NSAIDs. The evaluation of an individual statin or NSAID was adjusted by other statins or NSAIDs, because some subjects had taken more than one of these drugs.
We performed adjusted unconditional logistic regression models to evaluate the associations between intake of statins, NSAIDs and glioma. For the association between duration of drug use and glioma risk, we only analyzed data for all gliomas, because some duration intervals had relatively small number of cases and controls. We conducted four sensitivity analyses: 1) excluding subjects who required proxy respondents; 2) excluding subjects who took statins and NSAIDs for 24 months or less; 3) subjects with GBM only; 4) analyses based on the subgroups of men and women. All analyses were performed using STATA version 9.2 (College Point, Texas).
From July 2007 to Jan 2010, we recruited a total of 517 glioma cases and 400 matched controls. Table 1 showed demographic and other characteristics of cases and controls for the combined dataset and the individual study center. Control participants were more likely to be female, but overall, cases and controls were well matched with regard to age, ethnicity and educational level. Over 60% of cases had GBM and less than one out of five cases required proxy respondents.
For those who took statins > 6 months (ever use), there were significant inverse associations between simvastatin and lovastatin and the risk of glioma in the combined dataset (Table 2). Inverse associations were observed at both centers. The results for atorvastatin, pravastatin and rosuvastatin were not significant. In summary, there was a borderline significant inverse relationships for all statins combined.
For those who took NSAIDs > 6 months, we observed a significant inverse relationship between aspirin and the risk of all gliomas in the combined dataset (Table 2). Similar to statins, inverse associations were observed at both centers. Overall, all NSAIDs showed a borderline inverse association with glioma risk. Acetominophen, which does not possess anti-inflammatory effect, was not associated with these tumors (OR 0.89, 95% CI 0.56 – 1.38).
Results for duration of statin use were summarized in Table 3. All statins use showed a significant trend test, indicating there was an inverse association between duration of therapy and glioma risk. The duration category which showed the most significant inverse relationship with glioma was the group that took > 120 months of statins. A significant trend was also observed for simvastatin. Similarly, all NSAIDs combined showed a significant inverse trend between duration of drug use and glioma risk (Table 3); again, the most significant inverse association was attributed to > 10 years of use.
A sensitivity analysis that evaluated men and women separately showed that simvastatin, lovastatin and all statins combined had significant inverse relationships in men only (Supplementary Table 1). In NSAIDs, aspirin and all NSAIDs combined showed significant inverse associations in women only. There were no significant associations between statins and glioma in women, and between NSAIDs and glioma in men. However, interaction effect between gender and statin, and between gender and NSAIDs were not significant (p = 0.37 and p = 0.63, respectively).
After the exclusion of proxy respondents, the results for statins did not change (data not shown). All NSAIDs combined showed a greater association (OR 0.62, 95% CI 0.49 – 0.91) when proxy cases were eliminated from analyses, but the results for the individual NSAID did not change significantly (data not shown). When we excluded subjects who took statins or NSAIDs for 2 years or less, the results for both types of drugs were not altered (data not shown). Similarly, the associations between these drugs and the GBM subgroup were not significantly different from those of all gliomas (data not shown).
To our knowledge, this is the first study that evaluated the association between statin use and risk of glioma. Perhaps the reason that simvastatin and lovastatin were associated with a significant protective effect is that they are the most lipophilic statins and thus most readily cross the blood brain barrier (BBB).15 For example, a study showed lovastatin was detected in cerebrospinal fluid (CSF) at concentrations that might have a pharmacologic effect, whereas pravastatin, which is hydrophilic, was not detected in CSF.16 Moreover, in vivo brain perfusion technique in rats demonstrated that the in vivo BBB permeability coefficients of lovastatin and simvastatin were high, but in contrast, the coefficient for pravastatin was low.17
Through inhibition of HMG-CoA reductase, statin may inhibit downstream prenylation of G-protein subunits Ras and Rho family of GTPases.3 Ras activity is elevated in malignant glioma, and activated Ras stimulates other pathways essential for proliferation, progression through the cell cycle and inhibition of apoptosis in malignant glioma.18 Activated Rho is necessary for Ras-induced oncogenesis.19
Increasing evidence also suggests that statins may shift Th1 to Th2 cytokine profile.20 In glioma, a meta-analysis showed that atopic diseases, which are characterized by Th2 cytokine activations, were inversely related to glioma.21 Therefore, statins may impact the risk of glioma through their promotion of Th2 cytokine activities.
Two previous studies have evaluated the association between the intake of NSAIDs and glioma. Using a different and earlier dataset from UCSF, Sivak-Sears et al. reported an inverse association between NSAIDs and GBM.12 However, when the authors stratified subjects by proxy status, they found an inverse association with the self-reported cases but not the proxy-reported cases. Scheurer et al also observed an inverse association between anti-inflammatory medication use and adult glioma.11 However, they did not have information on specific anti-inflammatory medications, and neither study directly evaluated a duration-response relationship. In the present study, we confirmed an overall inverse association between NSAIDs and risk of glioma, but only the result of aspirin was significant. After the removal of proxy respondents, the result for all NSAIDs changed from borderline significant to significant, which indicated proxy reporting might have introduced some bias in the analysis.
Non-steriodal anti-inflammatory drugs (NSAIDs) exert their anti-neoplastic effect via suppression of cyclo-oxygenase 2 (COX2). High expression of COX2 is an independent poor prognostic factor in malignant glioma, after accounting for age, MIB-1, p53, retinoblastoma protein, p16 and Bcl-2 immunostaining.22 Thus NSAIDs may prevent tumorigenesis via inhibition of COX2.
In both statins and NSAIDs, the overall trend tests showed significant inverse relationships between duration of use and glioma risk. The most significant result was associated with > 10 years of drug use. In other cancers, statin use for at least five years was associated with a significantly reduced risk of colorectal cancer and advanced prostate cancer.6,23 The protracted time interval observed suggested that long-term user may derive more benefit from a protective effect of these drugs.
This study found statin use was inversely correlated with glioma in men, whereas NSAID intake was inversely associated with glioma in women. It is possible these drugs may have differential effects on tumors depending on hormonal context. However, these results are preliminary and need to be replicated before undertaking further investigative work to explain this phenomenon.
It was possible that patients sought medical attention due to symptoms of their growing brain tumors, which led to the discovery of hypercholesterolemia on blood tests and initiation of statin therapy. If statins were prescribed for this reason, we would have discovered positive associations between these drugs and risk of glioma. Also, our definition of taking statins at least twice a week for at least six months prior to diagnoses was designed to minimize this potential bias. In our sensitivity analyses, we further excluded those who used these drugs for two years or less prior to diagnoses and found no changes to our results. Furthermore, confounding by indication is also unlikely as statins are mostly used for the prevention of vascular diseases and treatment of hypercholesterolemia. To our knowledge, there has been no report on positive associations between vascular diseases and glioma, and the relationship between serum cholesterol level and glioma had been inconsistent.24
NSAIDs, on the other hand, might have been initiated by patients for headache associated with growing brain tumors, but this would have also resulted in a positive association between NSAID use and these tumors. Moreover, eliminating those cases that initiated NSAIDs within two years of diagnoses also did not change the results. NSAIDs are primarily taken for the prevention of coronary artery disease, arthritic and degenerative joint conditions, which are diseases that have not shown any association with glioma; therefore, confounding by indication is also not likely.
Despite potential methodological limitations, our study suggested that statins and NSAIDs are inversely associated with glioma. There is a need to further study our research questions in other populations. Future study may also benefit from the evaluation of pharmacogenetic modifications of these drugs, which may help to identify subjects who are most likely to benefit from therapy and least likely to suffer from toxicity.
This research was supported by grants from the National Cancer Institute (grants T32-CA09529, K05CA89155, R01CA52689, P50CA097257, K07CA127468, K07127468-02S and R01CA139020-27064).
The authors thank Drs. Peter Canoll and Tarik Tihan for central neuropathology review at CUMC and UCSF, respectively. The San Francisco Adult Glioma Study thanks the Northern California Cancer Center for identifying glioma cases.