The statin class of drugs has been unequivocally proved to reduce cardiovascular morbidity and mortality in primary and secondary prevention studies.1^–3 The cholesterol-lowering properties of these drugs arise from inhibition of the conversion of 3-hydroxy-3-methylglutaryl coenzyme A to mevalonate. Statins may offer other benefits that contribute to their protection against vascular disease. Of these, most attention has focused on anti-inflammatory effects as there is increasing evidence that inflammation is a feature of the atherogenic process4 and markers of inflammation independently predict coronary heart disease. Observations that statins reduce circulating levels of C-reactive protein5 6 and related pro-inflammatory molecules6 7 are thus noteworthy and may help to explain the early plaque-stabilising effects of these drugs.

As data on the spectrum (and potential mechanisms) of the anti-inflammatory properties of statins have accrued, they suggest that these drugs may offer therapeutic possibilities in other conditions with an inflammatory component in their pathogenesis. Data from several different areas of research are supportive of the hypothesis. For example, clinical studies involving transplant recipients point towards the immunosuppressive actions of statins, with reduced graft rejection episodes recorded after statin treatment.8^–11 In addition, a 71% lower risk of dementia was reported in patients previously prescribed statins using information derived from the UK-based General Practice Research Database (GPRD).12 Significantly, inflammation plays a key role in the pathogenesis of Alzheimer’s disease.13 14 Statins have also been shown to reduce significantly the extent of footpad oedema in mice after carrageenan injection,15 which is representative of acute inflammation. We extended this work by showing a therapeutic anti-inflammatory effect of statins in vivo, in collagen-induced arthritis, a representative animal model of inflammatory arthritis.16 Finally, there is also a variety of other animal studies, as well as ex vivo and in vitro studies in humans that support a potential anti-inflammatory role for statins.17^–21 Perhaps the most compelling evidence for a genuine anti-inflammatory effect of statins came from our 6-month clinical trial reporting a reduction in the disease activity score and swollen joint score in rheumatoid arthritis (RA) subjects treated with atorvastatin compared with placebo recipients.22

On the basis of the above data, we hypothesised that statin treatment may lower the risk of developing RA. We addressed this hypothesis by conducting a nested case–control study using data from the GPRD.

SUBJECTS AND METHODS

The GPRD has previously been described in detail.23 24 Briefly, since 1987, more than 5 million residents in the UK have been enrolled with selected general practitioners (GPs) who have agreed to provide data for research purposes to the database. The information recorded includes patient demographics and characteristics (eg, height, weight, smoking status), symptoms, medical diagnoses, referrals to consultants, hospital admissions and drug prescriptions (including the specific preparation, route of administration, dose and number of tablets for each prescription). On request, hospital discharge and referral letters are available for review to validate the diagnoses recorded in the computer record. All of the information received by investigators is anonymised. The database has been the source for numerous epidemiological studies in recent years, and the accuracy and completeness of these data have been well documented and validated. The patients in the GPRD are representative of the UK population, making the database an excellent resource for medical research.

Base population

The study period included the period 1 January 1992 to 31 December 2001. Within the database, we identified a study population consisting of three separate groups of subjects all aged 40–89 years. Group 1 included all patients with at least one prescription for a statin at any time (ie, atorvastatin, cerivastatin, fluvastatin, pravastatin, or simvastatin) or a lipid-lowering agent (LLA) other than a statin (ie, bezafibrate, ciprofibrate, clofibrate, fenofibrate, gemfibrozil, colestipol, cholestyramine, acipimox, or niacin/nicotinic acid). Group II included all patients with a computer-recorded International Classification of Diseases (ICD) coded diagnosis of hyperlipidaemia who did not receive any lipid-lowering drug treatment. Group III was a random sample of 25 000 individuals between the ages of 40 and 89 years who had neither a computer-recorded diagnosis of hyperlipidaemia nor a prescription for a lipid-lowering drug at any time.

Case definition

From the base population we identified all subjects with a first-time diagnosis of RA (ICD 8 code 712.9 and Read codes N040.xx). The computerised medical record of each identified subject was reviewed by hand to determine whether they were a probable, possible or unlikely case. Case status was determined using the following criteria. Probable cases: referral to outpatient specialist or inpatient hospitalisation and stated diagnosis of RA, plus the use of disease-modifying antirheumatic drugs (DMARD). Possible cases: the use of DMARD and a diagnosis stated in GP notes but no inpatient or outpatient referral. Unlikely cases: those with an unsupported diagnosis of RA (only a diagnosis recorded in the record without any referrals or DMARD prescriptions).

We also sent for the original medical records of a subset of these patients for review to ensure diagnostic accuracy (American College of Rheumatology, 1987 classification).

Those classified as unlikely cases of RA were excluded from further study as were individuals with psoriatic arthritis, reactive arthritis, or any history of cancer (excluding non-melanoma skin cancer).

The index date for the cases was defined as the date exactly one year before the date of the first RA diagnosis. We chose this date because the diagnosis of RA does not occur until the disease has been present and symptomatic for some period of time.

Control selection

From the study base population, we randomly selected up to four controls, ie, people without a diagnosis of RA, matched to each individual with RA by age (within 1 year), sex, calendar time (by using the same index date as for cases), practice and number of years of previous recorded history in the database. Controls were required to be registered in the practice at the index date. The same exclusion criteria that were applied to cases were also applied to controls.

Statistical analysis

A matched analysis using conditional logistic regression was used to evaluate the association between RA and hyperlipidaemia and the association between RA and exposure to statins. To assess the independent effect of hyperlipidaemia on the risk of RA we compared those with a diagnosis of hyperlipidaemia (defined as any code for hyperlipidaemia or hypercholesterolaemia) with the reference group of all those with no diagnosis or treatment for hyperlipidaemia. To assess the independent effect of statins on the risk of RA we compared those who were currently exposed to a statin and who had a diagnosis of hyperlipidaemia with those with a diagnosis of hyperlipidaemia in the absence of a statin. Current use of a statin was defined as the receipt of at least two prescriptions within 1 year preceding the index date (the receipt of only one prescription for a statin within 1 year preceding the index date was defined as unexposed). All other receipt of statins was considered past use. We also evaluated the effects of the duration of use and the effects of the individual statins. In addition to controlling for age, sex, calendar time, practice and years of recorded history in the database before the index date by matching, we controlled for smoking in the regression model. We further evaluated the effects of a previous history of diabetes, inflammatory bowel disease, hypertension and alcohol abuse, but we did not include them in the regression models because in no instance did they confound the association between statins and RA. We also looked at systemic lupus erythematosus, multiple sclerosis and psoriasis, but there were few patients with these diseases and we did not evaluate them further. Calculations were performed using SAS release 9.1.

RESULTS

We identified 313 cases of probable or possible RA and 1252 matched controls. Cases were predominantly female (61%) and were aged 50–69 years (70%). Cases were more likely than controls to be current smokers and to have hyperlipidaemia. Table 1 provides further descriptive characteristics of cases and controls.

We requested the original clinical records of 40 randomly selected patients diagnosed with RA and classified as possible or probable cases, in order to validate the diagnosis in the computer record. Of these, three had died and one had transferred out of the general practice and their records were therefore not available. In six instances the GP was not able to send records (for example the GP was too busy, or could not locate the record). We received and reviewed the records of 30 patients. In 28 of these patients (93%), 21 were considered probable and seven possible cases of RA. The remaining two were considered unlikely to be true cases of RA. In both of these cases the patient had been classified as a “possible” case based on the computer information. All cases classified as probable were confirmed as probable RA cases. We determined that using the computer record and the criteria described above was satisfactory for identifying established cases of RA.

Cases with RA were more likely than controls to have hyperlipidaemia. When patients with no diagnosis or treatment for hyperlipidaemia constituted the reference group we found a small increase in the risk of RA among those with untreated hyperlipidaemia (adjusted odds ratio (OR) 1.33, 95% CI 0.95 to 1.86; table 2).

Similarly, there was an increase in risk among current and past users of non-statin lipid-lowering drugs (adjusted OR 1.71, 95% CI 0.97 to 3.00 and 3.32, 95% CI 1.31 to 8.41). In contrast, no such increased risk was observed among statin users and the risk direction was slightly inverse among current statin users, although insignificant (adjusted OR 0.88, 95% CI 0.55 to 1.40). Correspondingly, compared with those with hyperlipidaemia and no statin treatment, people with no history of hyperlipidaemia were at lower risk of RA (adjusted OR 0.68, 95% CI 0.50 to 0.91; table 3).

In order to evaluate the independent effect of statins on the risk of RA, we subsequently compared those with hyperlipidaemia who received statins to those with hyperlipidaemia who did not use statins (ie, untreated hyperlipidaemia patients or those treated with non-statin lipid-lowering drugs). In this analysis the adjusted OR for the development of RA in subjects taking statins compared with the reference group was 0.59 (95% CI 0.37 to 0.96). There was no effect of past statin use but the numbers were very small in this exposure group (table 3).

We further evaluated the effects of the individual statins. There were too few users of cerivastatin or fluvastatin to evaluate these drugs independently, but there was a marginally sufficient number of simvastatin, atorvastatin and pravastatin users to assess their effects. The adjusted OR for RA in users of simvastatin compared with non-statin users with hyperlipidaemia was 0.74 (95% CI 0.45 to 1.23). The adjusted OR for pravastatin and atorvastatin users compared with non-statin users with hyperlipidaemia were 0.41 (95% CI 0.15 to 1.12) and 0.42 (95% CI 0.12 to 1.49), respectively. The results were thus broadly similar for the three major statins used (see table 4).

We evaluated both the duration of any statin use and the number of statin prescriptions received to see if there was evidence of a duration effect. There was no effect of increased use in either analysis (data not shown).

In order to assess the accuracy of the RA diagnosis we stratified the results by the RA likelihood rating. We found that the effects were similar in those with probable RA compared with those with possible RA (data not shown). We also evaluated the effect of smoking and found that smoking, although independently associated with RA, did not confound the statin/RA relation.

DISCUSSION

In these data we found that individuals with hyperlipidaemia who were receiving prescriptions for statins were at a lower risk of developing RA compared with people with a diagnosis of hyperlipidaemia who did not receive a statin. We also found that hyperlipidaemia was an independent risk factor for RA. Individuals with no history of hyperlipidaemia were at lower risk of RA compared with those with a diagnosis of hyperlipidaemia, without statin treatment.

An adverse lipid profile has been observed in active RA in a few previous studies, and thus may be partly responsible for the increased early cardiovascular risk in patients with RA.25^–29 Data from a recently published paper suggest that patients who later developed RA had a more atherogenic lipid profile than matched controls at least 10 years before the onset of symptoms. This lipid profile was characterised by higher total cholesterol, triglyceride and apolipoprotein B levels and lower high-density lipoprotein cholesterol levels.30 When we limited the reference group to those with no hyperlipidaemia and no treatment (“healthy people”) the adjusted OR for RA in statin users with hyperlipidaemia was 0.88. This result further supports the hypothesis that the use of statins may prevent the underlying increase in the risk of RA conferred by hyperlipidaemia.

The fact that non-statin LLA did not demonstrate a protective effect on RA does not necessarily mean that the protective mechanism of statins was not through lowering cholesterol. Statins have a much greater effect on lowering low-density lipoprotein cholesterol than any other class of LLA, and it remains possible that raised cholesterol could contribute to inflammation either directly via more substrate for oxidation or via its association with lipid rafts.31 Other human studies have suggested that the anti-inflammatory properties of statins are substantial in multiple sclerosis patients,32 diabetes,33 in the kidney transplant graft rejection rate,34 35 in colon cancer36 and dementia (including Alzheimer’s disease).12 The data on kidney transplantation, diabetes, dementia and cancer demonstrate the protective effects of statins, especially against the risk of new cases of these conditions, which corresponds to our study outcome (the risk of incident RA). Moreover, particularly in RA patients, the recent double-blind placebo controlled trial (Trial of Atorvastatin in Rheumatoid Arthritis; TARA) showed that atorvastatin significantly reduced swollen joint counts (mean reduction of 2.2, p<0.001) and C-reactive protein (50% decline, p<0.001), compared with placebo.22 Our current study findings expand directly on these trial data from established RA cases, and suggest that statins may be protective against the development of RA in those with hyperlipidaemia. A series of molecular studies has demonstrated important anti-inflammatory properties of statins, such as the inhibition of leucocyte–endothelial adhesion, anti-inflammatory effects on reactive oxygen and nitrogen intermediate production, the suppression of inflammatory cytokine release, inhibition of the inflammatory signal pathways, activation of anti-inflammatory transcription factors and inhibition of T-cell activation and co-stimulatory molecules.37 38

In order to ensure that cases and controls were comparable with respect to their underlying risk of RA, we compared statin users with hyperlipidaemia with people with other LLA-treated hyperlipidaemia or untreated hyperlipidaemia. Statin users in this study were older, more likely to be male, and had more hypertension than other LLA users, which is consistent with the results from an earlier study conducted using the GPRD.39 The earlier study also found that statin users were more likely to have concomitant severe cardiovascular co-morbidities, and had more GP visits than those with other LLA-treated hyperlipidaemia or untreated hyperlipidaemia. The presence of these associations, however, would not explain the protective effect of statins on the development of RA. In addition, female gender and smoking increase the risk of RA, but our analysis was matched by sex and adjusted for smoking, so confounding by these factors is unlikely. Nonetheless, residual confounding by indication or by some other unknown factors cannot be completely eliminated in an observational study.

To ensure the accuracy of the case definition, all RA cases were required to have both multiple prescriptions for a DMARD and a computer-recorded diagnosis of RA. When we analysed these data based on the certainty of the diagnosis we found similar results in those with probable and possible RA diagnoses, providing reassurance that our case definition was satisfactory.

The strengths and limitations of our study deserve comment. Exposure recall bias was not of concern in this analysis because exposure information was recorded on the computer at the time of prescription receipt, which preceded disease onset. Nor is biased recording of the outcome or potential confounders a concern, because all were also recorded on computer before the diagnosis of RA. The finding that current smokers have an increased risk of RA has consistently been found in many studies.40^–44 This finding and the finding that hyperlipidaemia is associated with RA are therefore both consistent with previous research. There was a limited sample size for individual statin use in this database, and therefore we were not able to evaluate the effects of each individual statin. Although we did have sufficient data to analyse the effects of simvastatin, pravastatin and atorvastatin, the numbers were still small and the effect measures were imprecise. In addition, data on statin use in non-hyperlipidaemia subjects in this study were small and could not be interpreted. Additional studies would be useful to evaluate statins further in the presence and absence of hyperlipidaemia in relation to the development of RA. We have limited information on users of long duration so we cannot address the issue of a duration effect. Nor can we assess whether statin use protects against the development of RA or whether it delays the onset of disease, or a combination of both possibilities. Further studies with more information on the long-term use of statins will be useful to evaluate these questions.

Signs and symptoms of RA begin insidiously and accumulate over months (>6 months),45^–47 so it is likely that some cases had the illness for some time before their diagnosis. Chan et al48 reported that the median lag time was 36 weeks between symptom onset and first diagnosis (interquartile range 15–87 weeks) among 81 newly diagnosed RA patients in central Massachusetts. (Among 14 patients for whom we sent for original clinical records and for whom we had information on dates of first symptoms, we found eight had a lag time of 1 year or less.) In order to account for this diagnosis delay, the assigned index date was exactly one year before the diagnosis date, thus limiting the misclassification of exposure due to assessing exposure after disease onset. As we could not exclude the possibility that some cases occurred before the imputed index date, we re-analysed the data using a new index date of exactly 2 years before the diagnosis date. In this analysis we found a stronger protective effect for current statin use on the risk of RA. When we re-analysed the data using the date of diagnosis as the index date, the results were not materially different from those presented.

This study evaluated the effect of statins on the risk of the development of RA. There are other inflammatory diseases such as irritable bowel disease, which, if the statin mechanism works through the inflammatory process, may also benefit from statin treatment. The aim of this study did not include other outcomes but it would be important to look at the effect of statins on the risk of other chronic inflammatory diseases to corroborate and extend our findings.

In summary, these data provide further evidence to suggest that hyperlipidaemia is a risk factor for the development of RA. More intriguingly, the data also support the hypothesis that statins, as a class, are potentially protective against the development of RA in those with hyperlipidaemia.