Higher mortality rates in rheumatoid arthritis were described initially in 1953 [19] and in all subsequent published clinical series [6, 16, 20]. However, estimation of mortality rates according to quantitative data has not been incorporated into the standard assessment of individual patients with rheumatoid arthritis, as has advanced clinical assessment in other chronic diseases, notably cardiovascular [21] and neoplastic diseases [22]. One reason for the relative absence of quantitative prediction of mortality in rheumatoid arthritis is that traditional clinical measures with documented prognostic value for mortality [13, 18, 23] appear too complex for routine clinical use, whereas global, radiographic, and laboratory data have not yet provided comparable prognostic utility for mortality.

Studies during the last decade indicate that simple quantitative clinimetric [24] measures, including questionnaires to assess activities of daily living that include only 20 [25] or 8 [26] activities and joint counts that include 36 [27] or 28 [28] joints, provide valid and reliable data to assess clinical status in patients with rheumatoid arthritis. These measures can be ascertained in usual clinical practice in 10 to 15 minutes and provide information comparable to more elaborate traditional measures, which may involve 70 or more activities of daily living [13] or joints [28]. Most studies of these simplified measures have been cross-sectional, and relatively little information is available about their prognostic value over long periods.

We have reported that mortality rates over 9 years in a cohort of patients with rheumatoid arthritis are increased compared with expected rates [13, 18] and are similar to most previous series from rheumatology settings [16]. Increased mortality rates, which were not seen over the first 2 to 3 years after baseline, were predicted over 5 years or longer by more severe clinical involvement at baseline and were comparable to other chronic diseases in certain patients. Clinimetric measures that predicted increased mortality rates included joint counts, functional status questionnaires, and other clinical measures, as well as age and formal education level [13, 18, 29].

In this report, we extend these studies to analyze whether simplified questionnaires and joint counts, which can be obtained in 10 to 15 minutes in a physician's office, provide effective data to identify increased probability of mortality over the subsequent 15 years in individual patients.

Methods

Top Methods Results Discussion Author & Article Info References

Patients

Seventy-five patients with rheumatoid arthritis were evaluated during the fall of 1973. The patients included 53 women and 22 men (73 white and 2 black persons). The mean age at baseline in 1973 was 54.7 years, the mean duration of disease was 11.2 years, and the mean formal education level was 10.9 years. The mean joint count level was 16.2 of 50 joints, the mean morning stiffness was 82.1 minutes, and 12.3% of the patients had cardiovascular disease (in addition to hypertension). Mean values for baseline measures of functional capacity were as follows: Questionnaire responses about activities of daily living were 88.3% of activities done "with ease"; grip strength was 99.2 mm Hg; modified walking time was 18.5 seconds; the button test was 76.6 seconds; and the disease adjustment score was 1.8 [13, 18]. The standardized mortality ratio was 1.62 at 15 years. That is typical of series of patients with rheumatoid arthritis, as were the findings that rheumatoid arthritis was listed on only 14 of 34 death certificates and the overall similarity of attributed causes of death to those in the general population [6, 16]. The demographic composition and baseline values for patients in this study were similar to those seen in 155 other consecutive patients with rheumatoid arthritis studied at Vanderbilt University from 1982 to 1984 [13]. Evidence has been presented that patients seen in private rheumatology practices are similar to patients seen at medical centers [30], although it is not known whether patients seen by rheumatologists may differ from patients seen by nonspecialist physicians. All 75 patients who were studied in 1973 were accounted for 15 years later in 1988.

Protocol for Baseline Evaluation

The baseline evaluation in 1973 included assessment of demographic, clinical, articular, questionnaire, and functional capacity variables. Demographic variables at baseline included age, sex, and level of formal education. Clinical variables included 1) duration of disease [the years from disease onset to 1973 as reported by the patient and recorded by a physician]; 2) morning stiffness [the number of minutes from awakening to optimal status for that day]; 3) treatments [parenteral gold salts and oral corticosteroids]; and 4) comorbid cardiovascular disease (including diseases other than hypertension only). Articular variables were based on a joint count in which swelling, limited motion, and deformity were recorded for each of 70 joints in each patient, and any abnormality was scored as an "involved" joint. The standard joint count studied as potentially prognostic for mortality included 50 joints: 8 distal interphalangeal joints in the hand, 10 proximal interphalangeal joints in the hand, 10 metacarpophalangeal joints in the hand, 2 shoulder, 2 knee, 2 hip, 2 ankle, 2 wrist, 2 elbow, and 10 metatarsophalangeal joints in the feet. Certain analyses were directed at the prognostic value of reduced joint counts for mortality for 15 years. A 36 joint count [27] did not include shoulder or hip joints, distal interphalangeal joints in the hand, or metacarpophalangeal joints in the thumb. A 28 joint count [28] did not include hip or ankle joints, distal interphalangeal joints in the hand, or metatarsophalangeal joints in the foot. A 12 joint count included only 8 metacarpophalangeal joints (not including the thumb), the shoulders, and the knees; a count of 6 joints included only shoulders, knees, and hips. Survival rates were also analyzed according to baseline involvement of 0, 1, or 2 knees.

Two questionnaire measures of functional status were obtained.

1. Activities of daily living: An interviewer reviewed a list of 87 activities with each patient, and the patient was asked to rate his or her performance into one of three categories, "able to do with ease" = 1; "able to do with some help" = 2; and "unable to do" = 3. Certain activities were irrelevant to certain patients (for example, use of cane or wheelchair), resulting in patients being asked a variable number of questions between 74 and 87 (mean, 82.9). Overall functional capacity was calculated by determining the percentage of activities that the patient was "able to do with ease" and calculating the mean score.

In order to analyze the prognostic value of a decreased number of items, survival was also analyzed according to only 20 or 8 of these activities, to simulate the 20-item Stanford Health Assessment Questionnaire (HAQ) [25] and the 8-item modified version of this questionnaire (MHAQ) [26]. Responses in the initial review had been obtained by an observer rather than by self-report as in the contemporary questionnaires.

2. Adjustment scale: Three questions were asked in a self-report format at baseline in 1973: 1) "Do you feel all is being done that can be done for your care?"; 2) "Do you feel you have been well educated on how to live with your condition?"; 3) "Do you neglect medicine or treatment because of the expense?" A score of 1 was assigned for a "yes" response to the first two questions and a "no" response to the third question. The total adjustment score, therefore, ranged from 0 to 3.

Three physical measures of functional status [31] were obtained.

1. Grip strength: A blood pressure cuff was inflated to 30 mm Hg, and the patient was asked to squeeze it as hard as he or she could. The test was repeated three times for each hand, and the score recorded was the mean of these six measures.

2. Modified walking time: The standard procedure, in which the patient is asked to walk at a normal pace for 7.6 metres (25 feet) [31], was modified in the 1973 evaluation by having the patient get up from a chair before walking and sit in a chair at the conclusion of the 25-foot walk. The score was recorded in seconds.

3. Button test: A standard button board (J. A. Preston, Clifton, New Jersey) was used. The patient was asked to unbutton five buttons and then button them as quickly as he or she could, using the dominant hand only, unlike the present method [31]. The score was recorded in seconds [32]. Trivial differences in values of tests of functional capacity that were previously reported [13] resulted from small amounts of missing data found after that report was published.

The baseline evaluation 15 years earlier had included more quantitative data than are usually collected in the care of patients with rheumatoid arthritis but unfortunately had not included the American Rheumatism Association Criteria for rheumatoid arthritis [2, 3] or the formal American Rheumatism Association Functional Class [33]. Most patients had a complete blood count, but only a few patients had assessments of erythrocyte sedimentation rate or tests for rheumatoid factor; therefore, meaningful analyses according to baseline laboratory data could not be done.

Statistical Analysis

The data were analyzed using the BMDP [34] and SAS [35] statistical software packages. Standard mortality ratios were computed by comparison to age- and sex-specific U.S. mortality [36]. The proportion of survivors at various time intervals was analyzed for each individual baseline measure using actuarial life tables. Patients were classified into four categories for each baseline variable using clinically meaningful cut points, and statistical significance was computed using the Wilcoxon method. In analyses of simplified questionnaire and joint count indices, patients were classified into four groups using quartiles, to minimize possible bias in the selection of any cut points to provide effective prognostic data. Only three groups were included for analyses of the eight-item questionnaire and of knee involvement.

Cox proportional-hazards models [37] were computed to analyze whether the increased probability of death identified using a clinical or functional status variable might be explained by age, gender, duration of disease, or formal education level as possible confounding variables [37]. Different study variables were analyzed as continuous variables, with models computed for 5-, 10-, and 15-year survival, using the stepwise method for maximum partial likelihood ratios [34].

The relative risk for mortality was calculated according to each baseline variable, as described by Kleinbaum and colleagues [38]. The variables were dichotomized according to different cut point levels, including the mean, median, lowest quartile, and various clinically sensible levels. Results were similar using different cut point levels, and data are depicted only for results using the median. Crude estimates, as well as adjusted estimates controlling for age, duration of disease, gender, and formal education level, were determined. Confidence intervals (95%) were calculated for each unadjusted and adjusted point estimate of baseline variables.

Results

Top Methods Results Discussion Author & Article Info References

Mortality over 5, 10, and 15 Years

Standard mortality ratios in these patients were 1.86 at 5 years (95% CI, 0.98 to 3.00; 13 deaths compared with 7 expected in the U.S. population matched for age and gender); 1.92 at 10 years (CI, 1.21 to 2.78; 23 deaths compared with 12 expected); and 1.62 at 15 years (CI, 1.12 to 2.21; 34 deaths compared with 21 expected). Rheumatoid arthritis was listed on the death certificates of only 14 of 34 patients (41%) who died. The attributed causes of death in these 34 patients during the 15-year period Table 1 included cardiovascular disease (44%), cancer (21%), gastrointestinal disease (6%), pulmonary disease (12%), infectious disease (15%), and rheumatoid arthritis (3%). The findings for the standard mortality ratios, the absence of rheumatoid arthritis from death certificates, and the attributed causes of death are similar to published series of patients with rheumatoid arthritis [6, 39-41].

Actuarial Survival according to Various Individual Baseline Measures

Patient survival over 15 years was analyzed according to various baseline demographic, clinical, articular, gastrointestinal, and other functional status variables, which were analyzed for each variable individually Figure 1, Figure 2, Figure 3, Figure 4, and Figure 5. Among demographic variables, survival was decreased in patients who were older [P = 0.01] Figure 1, panel A) and in patients with fewer than 12 years of formal education [P = 0.04] Figure 1, panel C), after adjusting for age, functional status, and other variables [see below]. Survival was lower in men than in women, but this difference was not statistically significant Figure 1, panel B) (P = 0.08).

View larger version (17K): [in this window] [in a new window]   Figure 1. Survival in 75 patients with rheumatoid arthritis during 15 years according to demographic variables. The variables included age, gender, and formal education level.

View larger version (24K): [in this window] [in a new window]   Figure 2. Survival in 75 patients with rheumatoid arthritis over 15 years according to disease variables. The variables included duration of disease, status of joints (joint count), morning stiffness, comorbid cardiovascular disease, parenteral gold use, and oral corticosteroid use.

View larger version (21K): [in this window] [in a new window]   Figure 3. Survival in 75 patients with rheumatoid arthritis over 15 years according to measures of functional status. The variables included modified walking time, button test, grip strength (men and women), questions about activities of daily living, and adjustment index.

View larger version (23K): [in this window] [in a new window]   Figure 4. Survival in 75 patients with rheumatoid arthritis over 15 years according to responses to questions about activities of daily living with 87, 20, or 8 included activities. ADL = activities of daily living; HAQ = Health Assessment Questionnaire; and MHAQ = modified HAQ.

View larger version (27K): [in this window] [in a new window]   Figure 5. Survival in 75 patients with rheumatoid arthritis over 15 years according to joint counts. Joint counts included 50, 36, 28, 12, and 6 joints or 1 or 2 knees.

Among clinical measures, no statistically significant differences were noted in survival according to duration of disease Figure 2, panel A) (P > 0.2); any small differences were explained by differences according to age (see below). When 50 joints were included, patients with more than 30 involved joints did not survive as long as patients with fewer than 10 involved joints [P = 0.01] Figure 2, panel B). Thus, patients with more than 30 involved joints had survival that was less than 50% at 5 years and less than 30% at 15 years, whereas patients with fewer than 10 involved joints had survival that was more than 90% at 5 years and more than 70% at 15 years. Survival was also decreased according to morning stiffness [P = 0.11] Figure 2, panel C). Survival of patients with comorbid cardiovascular disease Figure 2, panel D) was 50% at 5 years compared with 90% in patients with no cardiovascular disease (P = 0.02). Survival did not vary in patients who did or did not receive gold compounds [P > 0.2] Figure 2, panel E). Survival was decreased in patients who received corticosteroids Figure 2, panel F) (P = 0.05 after 9 years [13] and P = 0.15 after 15 years), although this finding was explained by greater severity of disease in these patients [13].

Significant differences in survival over 15 years were seen according to functional status measures (Figure 3). Survival was less than 50% at 5 years and less than 30% at 15 years in patients whose modified walking time was more than 30 seconds, compared with more than 90% at 5 years and 70% at 15 years in those with baseline values of less than 10 seconds [P 0.001] Figure 3, panel A). In patients whose button test values were greater than 120 seconds, survival was 50% at 5 years and 30% at 15 years, compared with patients whose baseline values were less than 40 seconds; all of these patients survived for 5 years, and 90% of them survived for 15 years [P 0.001] Figure 3, panel B). Differences in survival rates according to grip strength were seen, statistically significant in women [P = 0.04] Figure 3, panel C) but not in men [P > 0.2] Figure 3, panel D); however, the numbers in each of the two groups were small.

Survival rates differed according to responses on an activities-of-daily-living questionnaire [P < 0.001] Figure 3, panel E). Among patients with the highest level of functional status, 90% survived for 5 years and 70% survived for 15 years, whereas fewer than 50% of patients with the poorest functional status survived for 5 years, and none survived for 15 years. Survival rates also differed according to a three-item disease adjustment scale [P = 0.04] Figure 3, panel F). Survival rates were greater than 90% at 5 years and greater than 70% at 15 years in patients with adjustment scores of 3, but rates were only 70% at 5 years and 40% at 15 years in patients with scores of 0 or 1 or with a missing score Figure 3, panel F).

Multiple Variable Analyses of Survival for 15 Years

Multiple independent variables were studied in Cox proportional-hazards models to assess 5-, 10-, and 15-year survival according to questionnaires for activities of daily living, joint count, and other variables. In these models, the clinical variables remained statistically significant and were independent predictors of mortality, when age, duration of disease, and formal education level were included in models. For example Table 2, the activities-of-daily-living score was an independent predictor of mortality at 5, 10, and 15 years (P = 0.001). At 5 years, no other variable, including age, was entered into the model. At 10 years and 15 years, age (P = 0.10 and P = 0.037, respectively) and formal education (P = 0.015 and P = 0.010, respectively) were entered, whereas duration of disease was not entered into any model.

Actuarial Survival according to Activities-of-Daily-Living Questionnaires and Joint Counts with Decreased Items

Two of the most effective predictors of mortality were the activities-of-daily-living questionnaire and the joint count. These variables are regarded as primary outcome measures for clinical trials in rheumatoid arthritis [42]. Reduction in the number of items included in these quantitative measures has been studied to facilitate their use in routine clinical care. Survival rates according to activities-of-daily-living questionnaires Figure 4 were similar when the number of activities was 74 to 87 [P = 0.01] Figure 4, panel A) as in the original questionnaire, or was 20 activities [P = 0.03] Figure 4, panel B) as in the Stanford Health Assessment Questionnaire [25], or was 8 activities [P = 0.01] Figure 4, panel C) as in the modified version of the Stanford Questionnaire [26].

Mortality patterns also were similar according to joint counts that included 50, 36, 28, 12, or even only 6 joints (shoulders, hips, and knees) (all P 0.03) (Figure 5). Involvement of 2 knees, rather than 1 or none, was predictive of mortality of 35% at 5 years and of 60% at 15 years [P = 0.02] Figure 5, panel F). These data indicate that reducing the number of items included in these quantitative baseline measures does not compromise their predictive capacity for mortality in rheumatoid arthritis for 15 years.

Relative Risks for Mortality over 5, 10, and 15 Years according to Clinical Severity

Relative risks for mortality over 5, 10, and 15 years were computed for patients with values above and below the median values for each variable, controlled for age, duration of disease, gender, and formal education level (Table 3). Most variables, other than duration of disease and morning stiffness, identified a greater risk for mortality over 5, 10, and 15 years (P < 0.05), adjusted for age, duration of disease, gender, and formal education level.

Relative risks computed according to whether patients were in the lowest quartile for various clinical status measures (data not shown) were similar or higher than when computed according to median values. Relative risks for clinical measures were greater after 5 years compared with 15 years after baseline, reflecting findings in Cox regression analyses (see Table 2). In contrast, relative risks for age, duration of disease, gender, and formal education level, were similar after 5, 10, and 15 years and thereby assumed greater relative predictive value with time compared with the clinical measures.

Discussion

Top Methods Results Discussion Author & Article Info References

These data extend previous studies [6, 13, 16-1823, 43-48] to document that quantitative measures of clinical severity, including a higher joint count (an increased number of involved joints), poorer status in questionnaire scores and in physical measures of functional status, comorbid cardiovascular disease, as well as age and formal education level, predict higher mortality rates over 15 years in rheumatoid arthritis. The predictive power of the clinical variables is not explained by age, duration of disease, or other variables. Evidence of dose-response relations among all substantial predictors of mortality suggest their clinical as well as their statistical significance. In this study, simplified questionnaires and joint counts with fewer items, which can be ascertained in clinical practice in 10 to 15 minutes, are as effective prognostic measures for mortality as more elaborate traditional measures.

We offer several important caveats about this study. The number of patients, 75, is small, although the observation of statistically significant results, with dose-response relations, is reassuring. The study does not involve an inception cohort but rather a clinical cohort whose average duration of disease was 12 years in 1973, when an extensive quantitative evaluation was done. However, most patients with rheumatoid arthritis in previous studies were not seen until after at least several years of disease, and the mean duration of disease of patients in most clinical trials in rheumatoid arthritis was 5 to 10 years [49]. Further, the markers identified as predictive of mortality in this study have been identified as predictive of long-term morbidity [13, 14, 30, 43-46, 50-55], including in patients within the first 2 years of disease [44, 47, 50, 53, 54]. Further prospective studies in inception cohorts appear of value to determine whether the variables we studied are useful in the prognosis of patients early in disease.

Questionnaires with fewer items and reduced joint counts [25-28] are as useful in predicting mortality in patients with rheumatoid arthritis as are more traditional measures, extending evidence of the clinical utility of these measures. Self-report questionnaires with 20 or 8 items are correlated with traditional measures, such as the joint count, radiographic scores, erythrocyte sedimentation rate, and physical measures of functional status [56]. Questionnaire data are as sensitive as traditional measures to changes in status in clinical studies [57] and in clinical trials in rheumatoid arthritis [58, 59], as well as in osteoarthritis [60]. Therefore, functional status questionnaires have been included in a core set of measures to be used in clinical trials [42]. Joint counts that include 36 [27] or 28 [28] joints are correlated with other clinical measures as effectively as a traditional index of 50 joints and are easier to do. Joint counts that include only 12 joints have effective predictive value, and 5-year survival rates were only 65% in patients with involvement of 2 knees, regardless of other clinical features.

The degree of prognostic value of a simple questionnaire or joint count for mortality (or morbidity) in rheumatoid arthritis has not yet been documented for any imaging or laboratory data. This may be of particular interest to internists, in an era of emphasis on cost containment in health care. If the predictive power documented here for clinical measures were documented for a radiographic or laboratory variable, it would likely be incorporated into routine care, at considerable expense, with further diversion of resources from clinical care. Most patients with rheumatoid arthritis incur considerably higher costs for laboratory and radiographic data than for clinical care. The possibility of reimbursement formal clinical questionnaire or joint count data, comparable with reimbursement for laboratory and radiographic data, might be a reasonable consideration in view of the prognostic value documented here.

The observation that formal education level is associated with poor outcome in patients with rheumatoid arthritis is similar to studies of cardiovascular disease [61, 62], cancer [63] and of the general population [64, 65]. Formal education level is a marker for socioeconomic status, which appears to be more explanatory of clinical status than age or duration of disease in patients with rheumatoid arthritis [66], as well as osteoarthritis, systemic lupus erythematosus, scleroderma, and fibromyalgia [67]. We have suggested that formal education level may be a marker for behavioral variables that may influence the prevalence, morbidity, and mortality of many chronic diseases [68]. Regardless of specific mechanisms of action, formal education level may be a marker for the influence of the patient, in contrast to treatments and health care system variables, on the development and outcome of chronic disease.

This study does not analyze the "natural history" of rheumatoid arthritis, which is not possible with modern medical treatment. Nonetheless, evidence of substantial mortality, work disability, and increased mortality rates in patients treated using the standard treatment of the 1970s and early 1980s [6, 20, 69] suggests that the traditional approach of waiting for evidence of radiographic erosion before beginning aggressive therapies [70] might be revised [71-76]. Recognition that rheumatoid arthritis may progress to levels at which clinical markers predict mortality rates of 30% to 50% during the next 5 years, comparable to three-vessel coronary artery disease [21] and stage 3 or 4 Hodgkin disease [22], may provide a rationale for new clinical approaches early in disease to prevent progression to these levels.

The situation is analogous to hypertension, in which the goal is not only to control the blood pressure (analogous to joint pain and swelling in rheumatoid arthritis) but also to prevent retinopathy or cardiomegaly (analogous to end-organ damage such as joint deformity and radiographic destruction in rheumatoid arthritis). It is not known whether earlier interventions might result in improved clinical outcomes, but approaches that include evaluation of more aggressive interventions according to quantitative measures should advance treatment of rheumatoid arthritis [71-75].