Patients seen for standard rheumatoid arthritis care have significantly better articular, radiographic, laboratory, and functional status in 2000 than in 1985

Abstract

Objective

A comprehensive quantitative analysis of measures of disease activity and joint damage has not been available to compare patients in different eras in the same clinical setting. This study was undertaken to determine whether the clinical status of patients with rheumatoid arthritis (RA) has improved on average in recent years.

Methods

A quantitative cross-sectional evaluation, which included joint count, radiographic, laboratory, patient questionnaire, and physical function measures, and therapies, was performed in 125 consecutive RA patients seen from 1984 through 1986 (“1985 cohort”). A virtually identical assessment was performed in 150 patients seen from 1999 through 2001 (“2000 cohort”), in the same weekly academic clinic. Measures were compared using descriptive statistics and a median regression model, adjusted for age, duration of disease, level of formal education, and rheumatoid factor.

Results

Patients in 1985 had significantly poorer status compared with those in 2000: median 12 versus 5 swollen joints, Larsen radiographic score 20 versus 3, erythrocyte sedimentation rate 33 mm/hour versus 20, and modified Health Assessment Questionnaire 1.0 versus 0.4 (P < 0.019). Severe Disease Activity Scores >5.1 were seen in 69% of 1985 patients, compared with 30% in 2000. Methotrexate was taken by 10% of patients in 1985, versus 76% in 2000. The proportion of patients not taking any disease-modifying antirheumatic drugs was 66% in 1985 versus 13% in 2000.

Conclusion

Patients receiving standard care for RA in this setting had significantly better status, including radiographic scores, in 2000 than in 1985, associated with aggressive treatment strategies, prior to the introduction of biologic agents.

During the 1980s and 1990s, the traditional “pyramid” approach to management of rheumatoid arthritis (RA) was replaced by aggressive treatment strategies directed toward long-term remission (1–7), with increasing use of methotrexate (MTX) (8) as monotherapy or in combinations with other disease-modifying antirheumatic drugs (DMARDs) (9), early in the disease course. Several reports have suggested that patients with RA at this time have better status than in previous decades according to the Ritchie Articular Index (10, 11), functional capacity (11, 12), and radiographic scores (13–15), including lower mortality rates in patients whose disease has responded to MTX (16, 17). Furthermore, most patients seen in standard rheumatology care appear not to meet inclusion criteria (which had been developed 2–3 decades ago) for most recent RA clinical trials, including trials of new anti–tumor necrosis factor biologic agents (18, 19).

These observations suggested that patients in the 1990s may have considerably better clinical status than in previous decades, even before availability of biologic agents such as etanercept (20), infliximab (21), anakinra (22), and adalimumab (23). Recognition of better status in individual patients with RA requires multiple measures, as in the American College of Rheumatology (ACR) core data set (24) and Disease Activity Score (DAS) (25), because a single measure (unlike blood pressure or serum cholesterol) is not sufficient to describe status in all individuals (26). Furthermore, measures of inflammatory activity, such as joint swelling and erythrocyte sedimentation rate (ESR), may be improved or unchanged over 5–10 years, while measures of damage, such as joint deformity or radiographic scores, indicate disease progression (27–37). Therefore, both measures of disease activity and measures of joint damage are needed to document better status of patients with RA.

Ideally, data concerning differences in the status of patients with RA at different times might be available from many sites to enable a multicenter study. However, quantitative data, other than laboratory tests, generally are not collected in routine care of patients with RA at most clinical sites (38). Data to compare patients at different times from a single clinical setting appear preferable to data from different settings. Ironically, possible better status of patients seen in routine care cannot be documented from randomized controlled clinical trials in RA patients conducted to date because, as noted above, inclusion criteria have been similar over the last 2 decades (18, 19).

In 1984–86, a cross-sectional single-center assessment had been conducted in consecutive patients with RA who were receiving standard clinical care in Nashville, Tennessee. Fries (39) and Moses (40) had suggested that a database of consecutive patients may be highly informative, comparable with clinical trials. Inclusion of all consecutive patients led to a database of patients that appears representative in retrospect, as it was used to develop a 28-joint count (41), describe the absence of correlation between joint tenderness and radiographic findings (42), rediscover radiographic abnormalities in most patients within the first 2 years of RA (43), recognize laboratory associations of HLA–DR4 with radiographic changes but not with measures of function (44), demonstrate reliability of physical measures of functional status (45), and document significant correlations of patient questionnaire scores for functional status with traditional measures (46).

In 1999–2001, a virtually identical assessment was conducted in all consecutive patients by the same observer (TP) in the same weekly academic rheumatology clinic. This report presents a cross-sectional comparison of the cohorts seen in 1984–86 (“1985 cohort”) versus 1999–2001 (“2000 cohort”). The results demonstrate considerably better clinical status in 2000 versus 1985, according to both measures of disease activity and measures of damage. These differences were associated with aggressive therapy, and appear largely to antedate the introduction of biologic agents.

PATIENTS AND METHODS

Patients.

Two cross-sectional cohorts of consecutive patients who were receiving standard care for RA were analyzed. All patients were diagnosed with RA according to the criteria of the ACR (formerly, the American Rheumatism Association) (47). The 1985 cohort included 210 patients with RA identified from 3 sites: a weekly rheumatology clinic at Vanderbilt University conducted by one of the authors (TP), as well as the Nashville Veterans Administration Medical Center, and the private practice of Dr. Joseph Huston in Nashville (41, 43–46). In order to compare patients seen in the same rheumatology clinical setting 15 years apart, only the 125 consecutive patients seen at Vanderbilt University are included in this report; the 85 patients seen at the 2 other sites are not included, particularly because data from the Veterans Administration Medical Center may bias the data toward poorer status in 1985.

The 2000 cohort included all 150 patients with RA seen at the weekly Vanderbilt University Clinic (TP) between January 1999 and December 2001. Seven patients were seen during both periods; they are included in the 1985 cohort in this report, so that each cohort includes entirely different patients. Results were virtually identical if these 7 patients were included in either the 2000 cohort or in both cohorts. An earlier report of patients seen between January 1998 and June 2001 (17) included 129 of these patients, 7 of whom were excluded here as noted above, and 122 who are included in this report. Data were available concerning 28 additional patients seen between July and December 2001, most of whom were seen for their first visit. If a patient was seen more than once over either study period, as was frequently the case, the first visit during the study period was chosen for these analyses.

This research program was approved by the Vanderbilt University Institutional Review Board for the protection of human subjects. Patients gave informed consent to complete questionnaires, be monitored in a longitudinal database, and have blood drawn for supplementary research studies.

Patient assessment.

In 1982, it was recognized that morbidity and mortality of RA could be predicted and monitored effectively by patient questionnaire and physical measures of function (48, 49). Since then, every patient (regardless of diagnosis) seen at the weekly rheumatology clinic of TP has completed at every visit a patient questionnaire version of the modified Health Assessment Questionnaire (M-HAQ) (50, 51) or its successor multidimensional HAQ (MD-HAQ) (52). This questionnaire assesses physical function in activities of daily living according to a 4-point scale, as well as pain and global status on 10-cm visual analog scales (VAS). A metrologist has assessed physical measures of function, grip strength, walk time, and button test (45, 53) for every patient at every visit. Patients with RA have had ESR and, since 1995, C-reactive protein (CRP) measured at each visit. Joint counts are assessed at initiation of a new DMARD, and at visits for specialized research studies, including the first visit during the 2 study periods for each patient analyzed in each cohort in this study. A patient questionnaire has been found to provide ∼80% of the data of a joint count assessment with 10% of the effort (38, 54).

The more extensive assessment performed in 1985 in all consecutive patients with RA was developed into a “standard protocol to evaluate rheumatoid arthritis” (SPERA) (55). This protocol now includes all measures in the ACR core data set (23) and the DAS (24), which are the 2 most widely used measures in contemporary RA clinical trials, as well as other quantitative measures, recorded on 5 short (1-page) forms. The SPERA now records (a) clinical review of RA, including ACR criteria for RA (47), extraarticular disease, comorbidities, laboratory results (ESR, hemoglobin, and rheumatoid factor [RF]); (b) complete medication review; (c) a 28-joint count (41) to assess joints for swelling, tenderness, or pain on motion, limited motion or deformity, and previous surgery, as well as physical measures of functional status (grip strength, walk time, and button test) (45); (d) a patient self-report MD-HAQ (50–52), including an M-HAQ to assess functional disability in 8 activities of daily living (ADL), advanced ADL, and patient self-assessment of pain and global status on 10-cm VAS; (e) a score for radiographs of the hands and feet.

The 1985 review had not included a patient or physician assessment of global status. Therefore, the standard DAS28, which includes a tender joint count, a swollen joint count, ESR or CRP levels, and patient assessment of global status, could not be calculated for the 1985 cohort. A version of the DAS28 that does not include a patient assessment of global status (56) was calculated for the 1985 and 2000 cohorts. The standard DAS28 (56) was also calculated for the 2000 cohort.

Radiographs of the hands were available in 104 patients (83%) in the 1985 cohort, and in 129 patients (81%) in the 2000 cohort. All were read by TS without knowledge of the identity of the patient or cohort at the time of the reading, according to a modified Larsen score, with 20 joints scored 0–5, for a total score of 0–100 (57).

Statistical analysis.

Data were entered into an Access database designed for the SPERA assessment, and analyzed using Stata 8.0 (Stata, College Station, TX). Mean and median values with interquartile ranges were computed for the clinical status variables. Differences between normally distributed continuous variables, including age and education, were compared according to mean values and tested for statistical significance with Student's t-test. Differences between non-normally distributed continuous variables were compared according to median values, and tested for statistical significance using the Mann-Whitney test. Differences between categorical values were compared for statistical significance using the chi-square test. Clinical status measures in the 2 cohorts were compared using a median regression model, adjusted for age, education, duration of disease, and RF status, which differed in the 2 cohorts.

The cohorts were compared graphically for the major clinical status measures according to duration of disease, using 2 types of analyses. 1) Standard scatterplots were computed according to duration of disease, with depiction of locally weighted scatterplot smoothing regression curves; different symbols are used to depict values in patients who were positive or negative for serum RF, and locally weighted smoothing regression curves for 2 groups according to RF status for radiographic findings, the one clinical measure that is correlated at high levels with RF (42, 44). 2) Bar graphs were computed separately for patients with RA for durations of <5 years, 5–15 years, and ≥15 years, to depict median values with 95% confidence intervals, adjusted for age, education, and RF status.

RESULTS

Patients in the 2 cohorts.

The 2 patient cohorts were similar in age, sex, and race (Table 1). The 1985 cohort was less educated, had shorter duration of disease, shorter followup period in the clinic, and was more likely to have positive RF. All subsequent statistical analyses were performed to include adjustment for age, education, duration of disease, and RF status.

Table 1. Demographic characteristics and disease clinical status measures in the 2 cohorts of patients with rheumatoid arthritis seen in 1984–86 (“1985 cohort”) and 1999–2001 (“2000 cohort”)

Variable*	Cohort		P†
	1985 (n = 125)	2000 (n = 150)
* TP = Dr. Theodore Pincus; M-HAQ = modified Health Assessment Questionnaire; DAS = Disease Activity Score. † By Student's t-test, Mann-Whitney test, or chi-square test, as appropriate. ‡ Values are the unadjusted median (interquartile range). P values were derived from a median regression model with adjustment for age, education, duration of disease, and rheumatoid factor status.
Demographic measures
Age, mean ± SD years	55 ± 15	58 ± 14	0.12
Female, no. (%)	81 (65)	107 (71)	0.25
White race, no. (%)	108 (86)	138 (92)	0.13
Education, mean ± SD years	11 ± 4	13 ± 3	<0.001
Positive rheumatoid factor, no. (%)	105 (86)	84 (61)	<0.001
Disease duration, years, median (interquartile range)	7 (2–15)	9 (4–18)	0.021
Years of followup by TP, median (interquartile range)	0 (0–1)	3 (0–7)	<0.001
Clinical status measures‡
Swollen joint count, 0–28	12 (6–16)	5 (2–10)	<0.001
Larsen radiographic score, 0–100	20 (2–36)	3 (0–13)	<0.001
Erythrocyte sedimentation rate, mm/hour	33 (16–50)	20 (9–33)	0.016
Hemoglobin, gm/liter	129 (116–138)	136 (128–143)	0.002
Functional disability score on M-HAQ, 0–3	1.0 (0.6–1.4)	0.4 (0.1–1.0)	<0.001
Pain visual analog scale, 0–10	52 (32–80)	49 (15–73)	0.38
Grip strength, mm Hg, 30–300	82 (65–115)	120 (91–166)	<0.001
Walk time, seconds, 0–20	8.6 (7.0–11.8)	7.4 (6.6–8.7)	0.018
Button test, seconds, 0–300	50 (39–71)	40 (32–52)	<0.001
DAS, 0–10	5.7 (4.9–6.5)	4.4 (3.2–5.3)	<0.001

DMARD use in the 2 cohorts.

Substantial differences were seen in the drugs taken by the 125 patients in the 1985 cohort and the 150 patients in the 2000 cohort (Table 2). Sixty-six percent of patients in the 1985 cohort were taking no DMARDs, compared with 13% in the 2000 cohort. Only 10% of patients in the 1985 cohort were taking MTX, compared with 77% of those in the 2000 cohort (Table 2). The mean MTX dose was 9.2 mg/week in 1985, compared with 12.7 mg/week in 2000, and the median dosage of MTX was 7.5 mg/week compared with 12.5 mg/week, respectively (Table 3).

Table 2. Disease-modifying antirheumatic drugs (DMARDs) used in the “1985 cohort” seen in 1984–1986 and the “2000 cohort” seen in 1999–2001*

	1985 cohort (n = 125)	2000 cohort (n = 150)
* Values are the number (%).
No DMARDs, may be taking prednisone	83 (66.4)	20 (13.3)
No DMARDs, no prednisone	46 (36.8)	5 (3.3)
Methotrexate + any other DMARD or prednisone	13 (10.4)	115 (76.7)
Prednisone + any other DMARD	64 (51.2)	129 (86.0)
Prednisone only	37 (29.6)	15 (10.0)
Methotrexate only or + prednisone or + hydroxychloroquine	13 (10.4)	92 (61.3)
Methotrexate + other traditional DMARDs	0 (0)	17 (11.3)
IM gold only or + prednisone and/or + hydroxychloroquine	12 (9.6)	1 (0.7)
Hydroxychloroquine only or + prednisone	6 (4.8)	6 (4.0)
Azathioprine only or + prednisone and/or + hydroxychloroquine	4 (3.2)	2 (1.3)
Auranofin only or + prednisone and/or + hydroxychloroquine	5 (4.0)	1 (0.7)
D-penicillamine + prednisone	1 (0.8)	0 (0)
Cyclophosphamide + prednisone	1 (0.8)	0 (0)
Leflunomide + any other DMARD or prednisone	0 (0)	5 (3.3)
Infliximab + any other DMARD or prednisone	0 (0)	3 (2.0)
Etanercept + any other DMARD or prednisone	0 (0)	3 (2.0)

Table 3. Dosages of methotrexate and prednisone in patients with rheumatoid arthritis in the 1985 cohort and the 2000 cohort

	1985 (13/64)*	2000 (115/129)*
* Values in parentheses are n values for patients taking methotrexate/patients taking prednisone.
Methotrexate, mg/week
Mean	9.2	12.7
Median	7.5	12.5
Minimum	7.5	5.0
Maximum	15.0	30.0
Prednisone, mg/day
Mean	7.8	4.6
Median	5	4
Minimum	4	1
Maximum	30	15

Hydroxychloroquine was taken by 5% of patients in 1985, compared with 27% in 2000 (in combination with another agent in all but 3 in 2000). In contrast, gold injections were taken by 9.6% of patients in 1985 compared with only 2% in 2000; 2 of the 3 patients taking gold in 2000 were also taking MTX. The 2000 cohort included 5 patients who took leflunomide, 3 who took etanercept, and 3 who took infliximab (Table 2).

In 1985, 51% of 86 patients were taking prednisone, compared with 86% in 2000. However, 30% of the patients were taking prednisone monotherapy in 1985 (in addition to nonsteroidal antiinflammatory drugs), compared with 10% in 2000 (Table 2). The mean prednisone dosage was 7.8 mg/day in 1985, compared with 4.6 mg/day in 2000; median doses were 5 mg/day in 1985, and 4 mg/day in 2000 (Table 3).

Measures of clinical status in the 2 cohorts.

The median unadjusted number of swollen joints (0–28) was 12 in 1985 and 5 in 2000 (Table 1 and Figure 1). The median Larsen radiographic score (0–100) was 20 in 1985 compared with 3 in 2000 (Table 1, Figure 1). Rheumatoid factor–negative patients were virtually without significant radiographic damage in 2000, and even RF-positive patients with ≥15 years of disease scored <10% of maximum, compared with 30% of maximum in 1985 (Figures 1 and 2). Median ESR was 33 mm/hour in 1985 versus 20 mm/hour in 2000 (P = 0.016), and the median hemoglobin level was 129 gm/liter in 1985 versus 136 gm/liter in 2000 (P = 0.002) (Table 1).

Figure 1. Measures of clinical status in 2 cohorts of consecutive patients with rheumatoid arthritis (1985 cohort and 2000 cohort). Patients who were positive for rheumatoid factor (RF) are depicted by asterisks, and patients who were negative for RF by squares. Locally weighted scatterplot smoothing regression curves from 0 to 15 years are shown. Two curves for Larsen scores include a solid line for RF-positive patients and a dashed line for RF-negative patients, as Larsen scores are the only measure associated significantly with RF status (42, 44). ESR = erythrocyte sedimentation rate; M-HAQ = modified Health Assessment Questionnaire.

Figure 2. Bar graphs indicating medians and 95% confidence intervals for measures of clinical status in patients with <5, 5–15, and >15 years of disease in 2 cohorts of consecutive patients with rheumatoid arthritis. Hatched bars indicate data for patients in the 1985 cohort; open bars indicate data for patients in the 2000 cohort. All values were adjusted for age, education, and rheumatoid factor status. ESR = erythrocyte sedimentation rate; M-HAQ = modified Health Assessment Questionnaire; VAS = visual analog scale.

Median M-HAQ scores for functional status (0–3) were 1.0 in 1985 and 0.4 in year 2000 (Table 1, Figure 1). Physical measures also indicated better status in 2000: median grip strength was 82 mm/Hg in 1985 compared with 120 mm/Hg in 2000 (P < 0.001); median walk time was 8.6 seconds in 1985 versus 7.4 in 2000 (P = 0.018); and median button test was 50 seconds in 1985 versus 40 seconds in 2000 (P < 0.001) (Table 1). Pain score was the only measure that did not differ significantly in the 2 cohorts (Table 1 and Figure 2), with median scores of 52 (0–100) in 1985 compared with 49 in 2000, although the lower bound of the interquartile range was 32 in 1985 and 15 in 2000.

Analyses of median and 95% confidence intervals in 3 subcohorts according to duration of disease were performed (Figure 2). No overlap between the 1985 and 2000 cohorts was seen for swollen joint count in patients with disease duration of 0–4 years or 5–14 years, while overlap existed in the ≥15-year group; differences for all groups were statistically significant (P < 0.001). Radiograph scores overlapped in patients with 0–4 years of disease, but did not overlap in the 5–14-year groups, and differed substantially in patients with ≥15 years of disease, in whom median Larsen scores were 37 in 1985 versus 9 in 2000 (P < 0.001). ESR values overlapped in both cohorts, but were consistently lower (P = 0.016) in the 2000 cohort compared with the 1985 cohort; hemoglobin levels (P = 0.002), grip strength (P < 0.001), and walk time (P = 0.018) also indicated better status in 2000 than 1985.

The M-HAQ scores did not overlap between the cohorts in the 0–4 and 5–14-year subgroups, with substantially better scores in 2000 compared with 1985, although overlap was seen in the ≥15-year groups (P < 0.001). Pain scores, as noted, were the one measure that did not differ significantly in the 2 groups, although scores were somewhat lower in the patients in 2000 with 0–4 and 5–14 years of disease, but higher in patients with ≥15 years of disease. This was the only subgroup in which 1985 patients had better status than 2000 patients (Figure 2).

The median DAS28 score, calculated without patient assessment of global status, was 5.7 in 1985 compared with 4.4 in 2000. The proportion with scores of ≥5.1, indicating severe disease activity, was 68.9% in 1985 compared with 30.1% in 2000 (Table 4). The DAS28 results for 2000 were virtually identical when patient assessment of global status was included, although the proportion was higher than in our previous report (57), probably because most additional patients were seen on their first visit.

Table 4. Twenty-eight–joint Disease Activity Scores (DAS28) in the 1985 cohort and the 2000 cohort*

DAS28	Disease activity level	1985 cohort, DAS28 without patient global status	2000 cohort
			DAS28 without patient global status	DAS28 with global status
* Values are the number (%).
<2.6	Remission	3 (2.5)	14 (13.6)	14 (13.6)
2.6–3.19	Mild	5 (4.2)	12 (11.7)	15 (14.6)
3.2–5.09	Moderate	29 (24.4)	46 (44.7)	43 (41.7)
>5.1	Severe	82 (68.9)	31 (30.1)	31 (30.1)
Total		119 (100.0)	103 (100.0)	103 (100.0)

DISCUSSION

The results of this study clearly indicate better status in the 2000 cohort versus the 1985 cohort, with differences in the range of 50%, comparable with findings in current clinical trials of biologic agents. These data are consistent with reports indicating better status in recent years of patients with RA, as determined by joint examination (10), functional capacity (11), and radiographic scores (13–15) in addition to substantially better mortality outcomes associated with responses to MTX (16, 17). More favorable outcomes are seen both with measures of inflammatory activity, which may be improved over time, and measures of joint damage, which may show disease progression in the same patients (27–37). Only 7% of all the patients in the study had been treated with more powerful new DMARDs and biologic agents such as leflunomide, etanercept, and infliximab (anakinra and adalimumab were not available at this time). Therefore, the data suggest that radiographic progression can be slowed substantially in many patients by MTX, with significantly less cumulative damage in 2000 compared with 1985. The Finland Rheumatoid Arthritis Combination Trial (FinRACo) (59, 60), Early Rheumatoid Arthritis clinical trial (61), and the Tight Control for Rheumatoid Arthritis (TICORA) study (62) also indicate substantial slowing of radiographic progression with MTX.

This study has several limitations. It might appear desirable that documentation of improved clinical status be available from a randomized controlled clinical trial, but such a trial cannot be performed over 15 years in patients with RA for ethical and logistic reasons (63). Furthermore, although selection bias regarding the therapy given to an individual patient is overcome in clinical trials, inclusion and exclusion criteria and other limitations of the methodology (18, 64–80) may introduce other types of selection bias which may compromise generalizability and representativeness.

A second limitation is that these findings were obtained in a single clinical setting, and comparisons with patients in other settings are clearly desirable. However, few institutions rigorously collected quantitative data in all patients in 1985, and even at this time (38, 54) quantitative data in standard care, outside of research studies or laboratory tests, are unusual. Longitudinal data in several reports suggest better clinical status of patients with RA in recent years (10–15), as well as lower mortality rates in patients whose RA has responded to MTX (16, 17), consistent with data in this report. Furthermore, previous reports have suggested that data concerning patients with RA from this setting (and most individual settings) appear generalizable to most RA patients, including a report in 1984 indicating severe functional declines, frequent work disability, and premature mortality in patients with RA (27), and reports from analyses of the 1985 cohort of a 28-joint count (41), radiographic damage within the first 2 years of RA (43), and significant correlation of patient questionnaire scores (46) compared with traditional measures. In addition, many important concepts concerning RA were derived initially from single-center observational studies, including evidence of substantial work disability (81), functional declines (27), radiographic damage (82), and premature mortality in RA (83), as well as use of weekly low-dose MTX in clinical care (84), which were then confirmed at other sites. Data similar to those reported here were found at the only other site known to the authors at which patient questionnaires were also completed by all patients at all visits in 1985 and 2000 (Wolfe F: personal communication). Even if the data are not generalizable, they are nonetheless rigorously collected data from clinical care in one setting, with highly significant differences in the range of 50% between 2 patient groups, as seen in contemporary clinical trials.

A third limitation is that there may have been changes in the observer (TP) over a 15-year period leading to measurement differences and, furthermore, grip strength, walk time, and button test were assessed by different assessors in 1985 and 2000. The swollen joint count was performed by an assessor (Ms Raye Brooks) in 1985 and by TP in 2000, ESR and hemoglobin are laboratory measures, M-HAQ and pain are self-reported measures, and all radiographs were read by one observer (TS) in 2002, who did not know any other data about the patient or the year in which the patient was seen.

A fourth limitation is that patients in the 2000 cohort may have been self-selected for attendance at a mature clinical practice and response to therapy, and statistically significant differences were seen in education level, RF, disease duration, and duration of treatment. However, all analyses were adjusted for variables that differed significantly in the 2 cohorts. Stratified analyses in patients with varying duration of disease (Figure 2) also indicated significant differences after 5 years, although radiographic and pain scores during years 0–4 were similar in both cohorts. Rheumatoid factor status may convert from positive to negative with effective treatment. Differences of the magnitude seen cannot be explained by known variables that differed in the 2 cohorts, although unknown variables may have been contributory.

The multicenter randomized controlled clinical trial remains the best method to directly compare efficacy of therapies. However, clinical trials have many limitations (18, 19, 64–80), and analysis of consecutive patients according to quantitative measures in different periods may provide a useful approach to assess possible changes in status of patients with RA and other chronic diseases (39, 40). This design also mimics a “scientific experiment,” as does a clinical trial (67, 71), by isolating a single variable and keeping all other variables as constant as possible. In a clinical trial, the isolated variable is a drug versus placebo or another drug. In analyses of cross-sectional data from consecutive patients at different periods, the variable is the time seen, e.g., 1985 versus 2000. Both randomized controlled clinical trials and cross-sectional data from consecutive patients should be supplemented with long-term longitudinal data to provide improved accuracy concerning results of therapeutic interventions and outcomes (27, 33, 39, 40, 49, 85).

The data presented here indicate better status for patients with RA in 2000 compared with 1985, over which an aggressive treatment strategy was implemented (2, 4, 5, 86). These findings are consistent with other reports of improving status of patients with RA (10–17). The data may be explained in part by a suggestion that RA may be becoming a milder disease (87–90), supported by the observation that patients with <4 years of disease had better status in the 2000 cohort than in the 1985 cohort according to many measures. However, radiographic and pain scores during years 0–4 were similar in both cohorts, and in some patients in the 2000 cohort, measures were comparable with those in the 1985 cohort (Figures 1 and 2).

The data are consistent with the concept that an aggressive treatment strategy to control inflammation in RA is more effective than a traditional conservative approach. In particular, early use of low-dose MTX is associated with better clinical status in our study, consistent with findings in the FinRACo (59) and TICORA (62) studies. Regardless of possible explanations, it appears that the outlook for patients with RA at this time is considerably more favorable than in prior decades.

Acknowledgements

The authors thank Melissa Gibson for maintenance of clinical records, Christopher Swearingen for expert computer support, and Frederick Wolfe for helpful discussions and sharing of unpublished data.