The data presented herein were obtained from the EMECAR cohort (Estudio de la Morbilidad y Expresión Clínica de la Artritis Reumatoide), a prospectively designed cohort of RA patients promoted by the Spanish Society of Rheumatology in 1999. The main aim of this study was to determine the clinical situation, the management and the progression rate of patients with RA in Spain, as well as to estimate the frequency of both complications and associated comorbidity of this disease.17

The cohort was established by random sampling of patients diagnosed of RA registered in clinical databases of 34 participating centers in Spain. From a total eligible population of 13260 patients, 1427 were randomly chosen and examined to confirm the RA diagnosis according to the ACR 1987 criteria.18 Twenty-five patients per center (a total of 850 patients) were initially selected from each confirmed pool. The rest remained in reserve. Patients not included (135 did not meet RA criteria, 96 had died and 327 were not found) were replaced by the subsequent patients on the selection list (Fig. 1). Eighty patients who did not want to take part in the study were not replaced. The protocol was revised and approved by ethical committees from the centers taking part in the study. Those patients who joined the cohort signed a consent agreement after being informed in detail of the nature and the purpose of the study.

Fig. 1.

Flow chart of the study. *Includes an initially participating centre that closed (10 patients) patients who changed their address or could not be found (48 patients) and the refusal, of the research collaborator to remain in the study (127 patients).

(0,21MB).

Finally, the baseline visit included 789 patients, 73% of them women and 75% with a positive rheumatoid factor. The mean age at diagnosis was 48±15 years and the mean disease duration at entry was 10±7 years, with 11% of the patients showing a disease evolution of less than two years (Table 1).

Baseline visits of the enrolled patients were carried out between November 1999 and November 2000. Data about the evolution of the disease, complications, comorbidities, and treatments were obtained retrospectively from the clinical records and contrasted with the patients at the start-up visits. The clinical and radiological status of each patient was also recorded. The prospective follow-up of the cohort lasted for 4 years and data acquisition concluded in November 2004. In the results section, we will refer to the visits by years; thus, the basal visits took place in 2000 and the last ones in 2004. Fig. 1 shows the flow of patients throughout the study, as well as the reasons for attrition, which ranged from 7.2% to 13.5% by year.

The rheumatologists involved in the study were instructed in collecting the data and performing the joint counts and other measurements. In all visits, patients were physically examined and laboratory tests were carried out. The functional state and quality of life were assessed through the Spanish versions of the HAQ19 and of the SF-12,20 respectively. Disease activity was determined through the disease activity score obtained from 28 joint counts (DAS28) using the formula with three parameters.21 Hand X-rays were taken at the baseline visit (year 2000), in the third year (2002), and in the last one (2004), and the radiological progression was assessed centralized by an experienced reader using the Larsen method with the Scott modification (range 0–150).22 In addition, treatments with NSAID, corticoids, and DMARDs, as well as their combinations, reasons for withdrawals or substitutions, necessity of intra- and periarticular infiltrations, the use of painkillers, the use of antiulcer drugs (specifically PPI) and other concomitant treatments were registered in every visit. Side effects attributed to NSAID, were also recorded and classified, according to their seriousness, as mild or severe. Mild side effects included headache, diarrhea, dyspepsia, rash, blood hypertension, insomnia, sickness, nausea, tenesmus, and dizziness; severe side effects included perforation, ulcer, and gastrointestinal bleeding, as well as cardiovascular events (congestive heart failure, ischemic heart disease and stroke).

For the description of the cohort, means, medians and absolute and relative frequencies were used. Student's t-test was used for comparisons between groups with continuous variables and normal distribution. Discrete variables not normally distributed were analyzed with the Mann–Whitney's U test and categorical variables with the chi-square test. A two-tailed alpha level (p) of 0.001 was agreed for statistical significance, instead of the more common p<0.05 level, to account for multiple comparisons. An estimation of the rate of gastrointestinal hemorrhages and cardiovascular events with 95% confidence intervals was also carried out. The association of factors with NSAID intake during follow-up was measured and tested by means of Cox-proportional hazard models. In a first step, bivariate models were performed, and then multivariate analyses were carried out including only those variables initially found to have a p<0.01 in the bivariate models. Two sets of analyses were carried out: models explaining the use of NSAIDs by variables related to the indication—namely stiffness, CRP, disease activity by DAS28 and VAS of global pain— and models explaining the use of NSAIDs by safety-related variables—age, steroids use, previous ulcer disease, hypertension, hypercholesterolemia, smoking habit, and diabetes—. Data analyses were re-run for COXIB only. We also analyzed the longitudinal risk of gastric ulcer, cardiovascular events, and cancer, associated with NSAIDs, controlling for steroids use, age, previous ulcer disease, and gastric protection, also with Cox-proportional hazard models. All analyses were performed with Stata 9.2 for Windows (StataCorp LP, College Station, TX, USA).

In the first visit, 613 (78%) patients confirmed, with no gender-related differences, that they took or had taken NSAID at any time because of RA. A complete description of the demographic characteristics, drugs used, complications, disease activity, and comorbidity of the patients included in the EMECAR cohort is shown in Table 1. Patients who took NSAID for RA were significantly younger, their disease had an earlier onset, and they had less cardiovascular and gastric comorbidity than patients who had never taken NSAID. RA activity and health-related quality of life were worse in patients on NSAID than in patients who had not taken these compounds. The use of DMARDs was higher in patients on NSAID. With regard to antiulcer medications, only 11% of the RA patients, independently of the use of NSAID, had taken these compounds by year 2000. Surprisingly, the group of patients treated with NSAID had accumulated less osteoporotic fractures than patients without NSAID (see Table 1).

In 2000, indomethacin was the most frequently prescribed NSAID in RA patients, followed by diclofenac and naproxen (Table 2). In 1999, COXIB (rofecoxib) started to be available in the market. Thus, very few patients were using these compounds at the end of 2000. In the baseline visit (ending by November 2000), 68 (11%) of patients in the NSAID group reported having had at least one side effect attributable to NSAID, most of which (72%) were minor gastrointestinal effects (Table 3). The frequency of accumulated side effects reported among the most commonly used NSAID was: 15% for naproxen, 12% for diclofenac, 11% for aceclofenac, and 10% for indomethacin.

There was a slight, but gradual decline in the percentage of patients with RA using NSAID during the follow-up period (from 78% in 2000 to 66% in 2004) (Fig. 2). Most patients in the NSAID group used a single NSAID throughout the study and only a small percentage of them tried two or more of these compounds consecutively (Fig. 2). The majority of patients who took NSAID used a classic one (Fig. 3). In the year 2000, the percentage of RA patients using COXIB was very low (1.5%). One year later, in 2001, 15.8% of the patients were already using these compounds. Since then, the percentage of COXIB-consumers remained stable along the observational period and at the end of 2004, almost 17% of patients were taking rofecoxib or celecoxib (Figs. 3 and 4). Throughout the study, only 8 patients replaced a COXIB by a classic NSAID. As shown in Fig. 4, indomethacin was the NSAID which experienced the largest decline in consumption, from 32% in 2000 to 23% in 2004.

Fig. 2.

Proportion of patients on NSAID throughout the follow-up.

(0,12MB).

Fig. 3.

Proportion of patients on NSAID, according to the group of NSAID used.

(0,13MB).

Fig. 4.

Proportion of patients taking specific NSAID throughout the follow-up.

(0,15MB).

The use of NSAID (all NSAID) was related to the presence of pain (hazard ratio 1.16, 95%CI: 1.06–1.26; p<0.001), but not to the duration of morning stiffness or to disease activity (DAS28). Conversely, COXIB were preferentially used in patients with high levels of DAS28 (hazard ratio 1.27, 95%CI: 1.14–1.41; p<0.0001). With respect to safety, the presence of risk factors for NSAID-induced gastropathy (all NSAID) was not apparently taken into account by physicians, except in the group of patients older than 65 years (hazard ratio: 0.83, 95%CI: 0.75–0.92; p<0.001). When cardiovascular risk factors were analyzed, only the presence of hypercholesterolemia significantly reduced the use of NSAID (hazard ratio: 0.88: 95%CI: 0.79–0.98; p=0.02). The simultaneous use of antiulcer drugs and NSAID increased during follow-up, from 11% at the initial visit to 75% by the end of the study (Fig. 5). Surprisingly, patients treated with COXIB used antiulcer medications at a similar percentage as patients who were taking classic NSAID (Fig. 5).

Fig. 5.

Co-prescription of NSAID and COXIB with antiulcer medication. The last three years of the study are presented as there are more reliable data on antiulcer medication and higher numbers of patients treated with COXIB.

(0,11MB).

The frequency and type of side effects reported to be associated with NSAID consumption during the study are shown in Table 3. Frequencies varied from 11% in the initial cross-sectional assessment—that actually reflects the accumulated rate of side-effects up to first visit—to an annual average of 7% in the longitudinal phase. During the follow-up period, 21 gastroduodenal ulcers, with or without bleeding, were detected. The hazard ratio of using NSAID for presenting an ulcer was 0.88 (95%CI: 0.36–2.19; p>0.05), without adjusting for other variables. When antiulcer drugs were included in the model, the risk of gastroduodenal ulcer did not change substantially (hazard ratio: 1.49, 95%CI: 0.58–4.35; p>0.05). The only independent risk factor for gastroduodenal ulcer was the presence of a history of previous ulcer disease (hazard ratio: 31.36, 95%CI: 9.04–108.74; p<0.0001). The use of NSAID, with or without antiulcer medications did not modify the association with this important risk factor.

Throughout the study, there were 65 cardiovascular events (ischemic heart disease, stroke and congestive heart failure) with an incidence rate of 2.63 (95%CI: 2.06–3.35) per 100 patient∗years. In the group of patients without NSAID, with classic NSAID and with COXIB, the incidence rate of this complication was 3.2 (95%CI: 2.2–4.8), 2 (95%CI: 1.4–2.9) and 4 (95% CI: 2.3–7.1) per 100 patients∗years, respectively. There was no association between NSAID intake (classical and COXIB) and cardiovascular events (hazard ratio: 0.78, 95%CI: 0.47–1.30). The only variables associated with cardiovascular event were age (hazard ratio: 1.04, 95%CI: 1.02–1.07; p<0.0001) and hypertension (hazard ratio: 2.49, 95%CI: 1.68–3.71; p<0.0001). With regard to cancer, a rate of 12.3 per 1000 patient*years (95%CI: 6.4–23.6) was observed in patients without NSAID, 11.9 (95CI%: 7.4–19.2) in those who used classic NSAID and 3.4 (95%CI: 0.4–24.1) in COXIB-treated patients. The only independent risk factor for cancer was age (hazard ratio: 1.04, 95%CI: 1–1.07; p<0.01).