Baseline characteristics of the study population eligible for the RA analysis cohort and for the OA analysis are given in Tables 1 and ​and2,2, respectively. Patients with high PDC were more likely to be older, belong to a lower socioeconomic level, have higher prevalence of chronic conditions such as cancer, diabetes, and cardiovascular diseases, and have more frequent hospitalizations and visits to primary physicians. During 4.97 and 4.79 y of mean follow-up in the RA and OA analyses, there were 2,578 (3.07 per 1,000 person-years) RA cases and 17,878 OA cases (24.34 per 1,000 person-years), respectively. The age- and sex-specific incidence rates of RA and OA in the study population during the 10 y of study period are depicted in Figure 1. In most age groups, the incidence rate of RA and OA was 2- to 3-fold higher in women compared to men, and increased with age to a maximum in women aged 65–74 y (4.78 for RA and 50.45 per 1,000 for OA). RF tests were available for 76.6% of the diagnosed RA cases, of whom 1,478 (76.0%) were positive for RF at ≥10 IU/ml and 714 (36.7%) were positive for RF at ≥40 IU/ml.

The crude incidence density rate of RA among patients with a PDC level of less than 20% (3.89 per 1,000 [95% CI 3.62–4.17]) was higher by 51% compared to patients with a PDC level of 80% (2.57 per 1,000 [2.37–2.78]). No similar pattern was observed in the OA analysis where the incidence density rate of OA in the lowest PDC level category (23.61 per 1,000: 95% CI 22.92–24.31) was comparable to the incidence in the highest PDC level (24.12 per 1,000) (Table 3).

Baseline characteristics associated with increased risk for RA and OA included age, female gender, and frequent visits to primary physicians (Tables 4 and ​and5).5). Morbid obesity was related to a substantially higher risk of OA (HR=1.72; 95% CI 1.65–1.78), but not to RA. Similar results were obtained when analyses were limited to patients with at least 5 y of follow-up.

After adjustment for the variables in Table 4, patients with a PDC with statins of 80% or above, had a HR of 0.58 (95% CI 0.52–0.65) for RA compared with patients with PDC levels of less than 20% (Table 4). Similar results were obtained when analyses were restricted to patients with ≥5 y of follow-up. In the OA cohort analysis (Table 5), high persistence with statins (PDC level ≥80%) was associated only with a modest decrement in risk ratio (HR=0.85; 0.81–0.88) compared to nonadherent patients (PDC level <20%). In contrast with the RA analysis, the negative association between OA risk and PDC with statins was not observed when analyses were limited to patients with at least 5 y of follow-up (Figure 2).

When PDC with statins was analyzed as a continuous variable, an increase of 10% in PDC level was associated with an adjusted HR of 0.95 (95% CI 0.94–0.96) or a 5.3% lower risk of RA. Similar results were obtained when analyses were limited to patients with ≥5 y of follow-up. In the OA analysis, a 10% increase in PDC was related with an adjusted HR of 0.99 (95% CI 0.99–1.00) for OA. No association (p=0.23) was calculated when analyses were limited to patients with ≥5 y of follow-up. In stratified analyses, younger age at index date was associated with larger differences in RA risk. In patients aged 35–44 y, an increase of 10% in days covered with statins was associated with an adjusted HR of 0.90 (95% CI 0.87–0.97), whereas in patients aged 75 y or above, there was no significant association between persistence with statins and reduced RA risk. However, a heterogeneity test between all age strata did not reach statistical significance. Similarly, substantially lower risk of RA was calculated for patients persistently treated with high efficacy statins with a statistically significant heterogeneity (Figure 3).

Several of the cholesterol-independent effects of statins are exerted by the modulation of the synthesis of isoprenoids. Post-translational modification by isoprenylation is inhibited by statins; statins decrease isoprenylation of the GTP-binding proteins Ras and Rho, which consequently leads to the modulation of signaling pathways involving endothelial nitric oxide synthase [27], tissue plasminogen activator [28], endothelin 1 [29], plasminogen activator inhibitor 1 [30], and CRP [31]. Statins were also shown to effectively suppress the induction of class II major histocompatibility complex (MHC) protein and gene expression by interferon-γ (IFNγ), resulting in suppressed class II MHC-mediated T cell activation [32]. Interestingly, statins also selectively block β2 integrin and lymphocyte function-associated antigen 1 (LFA-1). In addition LFA-1 binding to intercellular adhesion molecule 1 (ICAM-1) was also abrogated by statins [33].

In a murine model of autoimmune encephalomyelitis, atorvastatin altered cytokine secretion favoring Th2 cytokine (interleukin [IL]-4, IL-5, IL-10, and transforming growth factor β) secretion while inhibiting the expression of Th1 cytokines (IL-2, IL-12, IFNγ, and tumor necrosis factor α [TNFα]). Interestingly, CD40 signaling, which is implicated in the pathogenesis of many autoimmune conditions, was reduced by statins in atheroma-associated cells in vitro and in atherosclerotic lesions; this change hinders proinflammatory cytokines, chemokines, matrix metalloproteinases, and tissue factor secretion and activity [34].

Additional features of statins are bone-specific anabolic and antiresorptive effects that may prevent osteoporosis, which often occurs in patients with active RA. These mechanisms may elucidate the improved, yet modest overall, RA disease activity, swollen joint scores, and reduced CRP in the TARA (Trial of Atorvastatin in Rheumatoid Arthritis) trial, a 6-mo, placebo-controlled, randomized clinical trial with Atorvastatin [6].

The JUPITER (Justification for the Use of Statins in Prevention: an Intervention Trial Evaluating Rosuvastatin) trial was designed to investigate the preventive effects of Rosuvastatin against vascular events among individuals with LDL cholesterol <130 mg/dl and enhanced innate immune response, as determined by a high-sensitivity CRP (hsCRP) level. The results of the JUPITER trial indicated that statin therapy may also reduce all cause mortality in patients with low LDL cholesterol, probably by decreased inflammation [35]. Although RA patients were not included in the JUPITER trial, the questionable validity of its conclusions [36], and the uncertainty of whether or not CRP itself is a marker of risk or the target for therapy, our findings may support the potential relationship between statin use and RA risk [37].

The strengths of the current study include prospective data collection, the use of administrative databases to avoid the problem of differential recall bias, the systematic and comprehensive collection of personal sociodemographic data, medical history, and utilization of health services prior to the index date, which reduces the possibility for bias due to study outcomes. The present retrospective cohort is one of the largest undertaken to date on the relationship between statin therapy and RA, with respect to the length of follow-up and the size of the study population. Furthermore, survivor-treatment selection bias and competing medical issues, two potential problems with observational studies of treatment efficacy [38], are unlikely to affect our results, since the association of statins and RA onset was not attenuated when analyses were limited only to patients with more than 5 y of follow-up data.

In addition, the present study used internal comparisons among patients who had at least one dispensed prescription of statins, reducing the threat of confounding by indication that might have occurred in previous investigations that compared statin users and nonusers [26]. For example, a recently published cohort study [39] on more than 2 million patients from 368 general practices in England and Wales found no significant association between statin use and RA. In their analysis, Hippisley-Cox and Coupland compared the risk of RA in statin users and nonusers. The study groups differed considerably in many important characteristics such as mean age (57 versus 44 y), body mass index (BMI) (28.3 versus 26), and potentially in other important variables that were not included in the multivariable analysis (e.g., cholesterol level, LDL levels, cardiovascular diseases, and other comorbid conditions, etc.). In addition, the authors reduced the statistical power of their tests by stratifying the analyses by sex and five types of statins, and did not take into account the effect of time and amount of statin purchased. All of the above mentioned aspects could have masked a significant association between persistent use of statins and RA.

This study has the following limitations. Our analysis was retrospective in nature and allocation of statin therapy was not randomized. Despite adjustment for baseline differences and an abundance of poor prognostic factors, a higher proportion of days covered with statins could still be a surrogate for other unmeasured variables that reflect a higher quality of care and more aggressive treatment strategies. In our analysis we did not address different temporal variations in patients' use of statins over the study period. However, a majority of RA patients (79%) purchased statins at least 3 mo prior to diagnosis. Also, the similarity in study results when limiting the analysis to patients with at least 5 y of follow-up reduces the possible effect of this potential limitation.

The most important methodological limitation in estimating effectiveness in observational studies is the potential for “healthy adherer” bias. Previous studies have indicated that in patients with a chronic illness, good adherence to medication, or even to placebo, is more likely to lead to better health and improved survival [40]. A recent study [41] aimed to examine whether adherence with statins is associated with a decreased risk of accidental events that were thought to be unrelated to statins (e.g., motor vehicle and workplace accidents, burns, and falls). As expected, they found a modest (10%–15%) overall reduction in accident rate among adherent patients compared to nonadherent ones. In order to evaluate this potentially important bias in our study, we conducted a similar analysis with OA as an outcome. Our results indicated that persistent use of statins was associated with a 15% lower OA risk, a relatively small difference compared to RA risk, but one that needs to be noted when considering the results of the study overall. In addition, the reduced risk for OA in patients with high PDC with statins was limited to patients with short follow-up periods and was not found in patients with a follow-up of 5 y or more. This finding supports the notion that most of the RA risk reduction is due to a real biological effect. The threat of “healthy adherer bias” in this study was further diminished by findings from our previous study [42] on the present cohort indicating that older age and the presence of chronic diseases and other risk factors for cardiovascular events are associated with higher persistence with statins.

The incidence of RA and OA in our study, based on physician diagnoses, is higher compared with previous studies [2],[43],[44]. This higher incidence can be explained by the relatively older age and the frequent visits to physicians in our study population, leading to earlier detection. Also, RA cases in most epidemiological studies were defined by the 1987 ACR criteria and not on diagnosis alone as in our study. The association between LDL levels and risk of RA is not fully understood. Some studies [45]–[47], but not all [48], suggested that patients with active untreated RA have reduced LDL levels. A recent large retrospective cohort from the Rochester Epidemiology Project showed a decrement in LDL levels during the 5-y period before RA incidence. This decrement could not be explained by usage of lipid-lowering drugs alone. This conclusion is in agreement with our finding that high LDL levels (>190 mg/dl) at index date were significantly associated with a lower short-term (<5 y) risk of RA. Since patients with higher LDL levels at index date are more likely to be persistent with statin therapy [49], a residual confounding cannot be excluded.

Mild muscle pains are one of the frequent side effects of statins documented in 5% to 10% of outpatients on statins [50],[51] and commonly result in discontinuation of treatment. Muscle symptoms frequently begin within several months after initiation of therapy [50]. Since muscle pains can be misclassified as OA symptoms and result in result in a mistaken diagnosis of OA shortly after treatment initiation. This potential differential information bias may explain the disappearance of the small negative association between statin use and OA when analysis was limited to patients with at least 5 y of follow-up. Also, one might claim that persistent use of statins may have been associated with more frequent physician visits increasing the chance of being diagnosed with RA or OA. However, the direction of this potential detection bias may only support our conclusions. In the present research, not only did the risk of RA not rise with increasing persistence, it decreased, indicating that the true association between persistent statin use and RA could be stronger than observed. Interestingly, starting statin treatment at a younger age (35–44 y) was associated with a more pronounced decline of onset of RA; this finding probably underlines the importance of inflammatory processes during this age and the potential role statins may have to block these mechanisms, in other words, the different effects that statins have at different ages once again implies that RA is a heterogeneous disease.