Showing papers by "James F. Fries published in 2004"

Long-term randomized controlled trials of tailored-print and small-group arthritis self-management interventions.

Abstract: Objective:The objective of this study was to test the effectiveness of a mail-delivered, tailored self-management intervention (SMART) and to compare it with the classic Arthritis Self-Management Program (ASMP)Methods:We performed 2 randomized controlled trials: 1) a study of 1090 participants rand

Equipoise, design bias, and randomized controlled trials: the elusive ethics of new drug development

Abstract: The concept of 'equipoise', or the 'uncertainty principle', has been represented as a central ethical principle, and holds that a subject may be enrolled in a randomized controlled trial (RCT) only if there is true uncertainty about which of the trial arms is most likely to benefit the patient. We sought to estimate the frequency with which equipoise conditions were met in industry-sponsored RCTs in rheumatology, to explore the reasons for any deviations from equipoise, to examine the concept of 'design bias', and to consider alternative ethical formulations that might improve subject safety and autonomy. We studied abstracts accepted for the 2001 American College of Rheumatology meetings that reported RCTs, acknowledged industry sponsorship, and had clinical end-points (n = 45), and examined the proportion of studies that favored the registration or marketing of the sponsor's drug. In every trial (45/45) results were favorable to the sponsor, indicating that results could have been predicted in advance solely by knowledge of sponsorship (P < 0.0001). Equipoise clearly was being systematically violated. Publication bias appeared to be an incomplete explanation for this dramatic result; this bias occurs after a study is completed. Rather, we hypothesize that 'design bias', in which extensive preliminary data are used to design studies with a high likelihood of being positive, is the major cause of the asymmetric results. Design 'bias' occurs before the trial is begun and is inconsistent with the equipoise principle. However, design bias increases scientific efficiency, decreases drug development costs, and limits the number of subjects required, probably reducing aggregate risks to participants. Conceptual and ethical issues were found with the equipoise principle, which encourages performance of negative studies; ignores patient values, patient autonomy, and social benefits; is applied at a conceptually inappropriate decision point (after randomization rather than before); and is in conflict with the Belmont, Nuremberg, and other sets of ethical principles, as well as with US Food and Drug Administration procedures. We propose a principle of 'positive expected outcomes', which informs the assessment that a trial is ethical, together with a restatement of the priority of personal autonomy.

The rise and decline of nonsteroidal antiinflammatory drug-associated gastropathy in rheumatoid arthritis.

Abstract: Objective Nonsteroidal antiinflammatory drug (NSAID)–associated gastropathy is a major cause of hospitalization and death. This study was undertaken to examine whether recent preventive approaches have been associated with a declining incidence of NSAID gastropathy, and, if so, what measures may have caused the decline. Methods We studied 5,598 patients with rheumatoid arthritis (RA) over 31,262 patient-years at 8 sites. We obtained standardized longitudinal information on the patients that had been previously used to establish the incidence of NSAID gastropathy, and also information on patient risk factors and differences in toxicity between NSAIDs. Consecutive patients were followed up with biannual Health Assessment Questionnaires and medical record audits between 1981 and 2000. The major outcome measure was the annual rate of hospitalization involving bleeding, obstruction, or perforation of the gastrointestinal (GI) tract and related conditions. Results Rates of GI-related hospitalizations rose from 0.6% in 1981 to 1.5% in 1992 (P < 0.001), and then declined to 0.5% in 2000 (P < 0.001). The fitted spline curve fit the data well (R2 = 0.70). The period of rise was mainly associated with increasing patient age and the GI risk propensity score. The period of decline was associated with lower doses of ibuprofen and aspirin, a decline in the use of “more toxic” NSAIDs from 52% to 42% of patients, a rise in the use of “safer” NSAIDs from 19% to 48% of patients, and increasing use of proton-pump inhibitors, but not with change in age, NSAID exposure, or GI risk propensity score. Conclusion The risk of serious NSAID gastropathy has declined by 67% in these cohorts since 1992. We estimate that 24% of this decline was the result of lower doses of NSAIDs, while 18% was associated with the use of proton-pump inhibitors and 14% with the use of less toxic NSAIDs. These declines in the incidence of NSAID gastropathy are likely to continue.

Longitudinal comparison of the Health Assessment Questionnaire (HAQ) and the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC).

Abstract: Objective To compare the measurement properties of the generic Health Assessment Questionnaire (HAQ) and the disease-specific Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC). Methods Physical function, pain, and radiographic progression were assessed in knee or hip osteoarthritis patients (n = 271) who had 2 radiographs that were at least 6 months apart from 6 ARAMIS (Arthritis, Rheumatism, and Aging Medical Information System) databanks. Data were compared at baseline and after a mean of 3.2 (SE 0.10) years. Correlation coefficients and standardized effect sizes (SES) were used to assess their relationship and responsiveness. Results The majority of items in the 2 function and pain scales overlapped and were highly and significantly correlated with each other at baseline and last assessments (function at baseline rs = 0.71 and function at last assessment rs = 0.79, P < 0.0001; pain at baseline rs = 0.70 and pain at last assessment rs = 0.76, P < 0.0001). The HAQ disability index and total knee score were more sensitive to detection of disease progression than the WOMAC (SES for HAQ = 0.27; SES for WOMAC = −0.05). Conclusion Both instruments showed favorable measurement properties, with the HAQ having the advantages of being more sensitive to change and adaptable to a wide variety of diseases and conditions, which contribute to the generalizability of findings.

Percentile benchmarks in patients with rheumatoid arthritis: Health Assessment Questionnaire as a quality indicator (QI)

Abstract: Physicians are in need of a simple objective, standardized tool to compare their patients with rheumatoid arthritis, as a group and individually, with national standards. The Disability Index of the Health Assessment Questionnaire (HAQ-DI) is a simple, robust tool that can fulfill these needs. However, use of this tool as a quality indicator (QI) is hampered by the unavailability of national reference values or benchmarks based on large, multicentric, heterogenous longitudinal patient cohorts. We utilized the 20-year longitudinal prospective data from 11 data banks of Arthritis Rheumatism and Aging Medical Information to calculate reference values for HAQ-DI. Overall, 6436 patients with rheumatoid arthritis were longitudinally followed for 32,324 person-years over the 20 years from 1981 to 2000. There were 64,647 HAQ-DI measurements, with an average of 19 measurements per person. Overall, 75% of patients were women and 89% were Caucasian; the median baseline age was 58.4 years and the median baseline HAQ-DI was 1.13. Few patients were treated with biologics. The HAQ-DI values had a Gaussian distribution except for the approximately 10% of observations showing no disability. Percentile benchmarks allow disability outcomes to be compared and contrasted between different patient populations. Reference values for the HAQ-DI, presented here numerically and graphically, can be used in clinical practice as a QI measure to track functional disability outcomes and to measure response to therapy, and by arthritis patients in self-management programs.

Attrition bias in rheumatoid arthritis databanks: a case study of 6346 patients in 11 databanks and 65,649 administrations of the Health Assessment Questionnaire.

Abstract: OBJECTIVE: Patient dropout (attrition) can bias and threaten validity of databank-based studies. Although there are several databanks of rheumatoid arthritis (RA) in operation, this phenomenon has not been well studied. METHODS: We studied the attrition patterns of patients with RA in 11 long-running databanks where patients were followed using semiannual Health Assessment Questionnaires. Attrition rates were calculated as the proportion of living patients who were in active followup at the cutoff date. Mantel-Haenszel methods and Weibull regression were used to model the relationship between attrition and age, sex, race, education, disease duration, functional disability, and other characteristics. RESULTS: Overall, 6346 patients with RA were recruited into the study cohorts and followed for 32,823 person-years with 65,649 observations. The crude attrition rate was 3.8% per cycle. Rates were lowest in community-based databanks. Smaller size of the centers, inner-city location, and university clinic settings were associated with worse attrition. In multivariable analyses, younger age, lower levels of education, and non-Caucasian race predicted attrition. Level of disability and disease duration were not associated with attrition. Conclusion. In terms of person-years of followup and observation-points, this may be the largest study on attrition to date. While it is possible to have very high overall retention rates, certain types of databanks (smaller, inner-city-based, and university-based) are more likely to be biased due to selective retention of older, more educated Caucasian patients.

Measuring effectiveness of drugs in observational databanks: promises and perils

Abstract: Observational databanks have inherent strengths and shortcomings. As in randomized controlled trials, poor design of these databanks can either exaggerate or reduce estimates of drug effectiveness and can limit generalizability. This commentary highlights selected aspects of study design, data collection and statistical analysis that can help overcome many of these inadequacies. An international metaRegister and a formal mechanism for standardizing and sharing drug data could help improve the utility of databanks. Medical journals have a vital role in enforcing a quality checklist that improves reporting.

New Instruments for Assessing Disability: Not Quite Ready for Prime Time

Abstract: Measurement of patient-reported outcomes (recently termed “PROs”) has been on the ascendancy in medicine, and especially in rheumatology, for nearly 25 years. These outcomes, which can be recorded by patients on questionnaires, such as the Health Assessment Questionnaire (HAQ) from our group (1) and the Medical Outcomes Study 36-question Short Form (SF36) (2), offer important advantages over traditional physician-recorded “process” measures. These advantages include the measurement of outcomes using the values of patients, the focus on longer-term outcomes, and the emphasis on the quality as much as the quantity of life. PROs can include the assessment of physical function or disability, treatment side effects, medical care costs, pain, or other content areas. Furthermore, they are easier to administer and less expensive than physician-observed health status measures (3). As a group, PROs have been proven to be reliable, valid, and sensitive to change, displaying these attributes in greater abundance than do physicianreported measures. Accordingly, PROs have become standard in the evaluation and approval of new drugs as well as in observational studies and randomized trials. They have been translated and culturally adapted to scores of languages and cultures and validated in literally thousands of studies (4). Our increasing reliance on PROs represents a major change in the evaluation of treatments and services to patients. Rheumatologists have been among the leaders of these efforts. After a quarter of a century, it seems reasonable to ask whether a new generation of outcome assessment instruments should supersede the old. Surely, it may be time to move our efforts to the next plateau, to take a set of good ideas and to improve them further. There is no shortage of suggestions for change or for innovations promising improvements. The article by Wolfe et al in this issue of Arthritis & Rheumatism, analyzing the disability index (DI) of the HAQ, is an innovative step toward the refinement of PROs as they relate to rheumatic disease (5). The conclusions to be discussed in this editorial are as follows: 1) the time for improvement of the HAQ and other instruments has come; 2) Wolfe and colleagues have made the best approach yet toward these issues in rheumatology; and 3) the newer techniques of item response theory (IRT) and computerized adaptive testing (CAT) have great promise, especially when used in combination. Nevertheless, I suggest that 1) a potential limitation of these approaches is loss of face validity, which differentiated PROs from physician-based observations in the first place; 2) human insights cannot easily be delegated to machines; and 3) what has been proposed to this point may represent change but not improvement. First, it is important to consider the background, which may seem complex but is essential for understanding both the promise and the pitfalls of these new approaches. IRT uses statistical modeling techniques to analyze multiple items that attempt to assess the same outcome dimension, such as questionnaire items on the HAQ DI. Using data about the questionnaire items themselves, one can assess quantitative aspects of any resulting index and of alternative indices. A problem arises, however, because IRT approaches require unidimensionality (i.e., that all of the items on a scale measure the same concept, as, for example, with walking a short distance and walking a longer distance). When the content area is not unidimensional (as with disability), face validity can be lost, with unfortunate consequences. The simplest IRT statistical model, used by Wolfe et al, originates from the work of Rasch and focuses on the relative difficulty of items (6). More recent modeling approaches include a parameter for item discrimination that allows some items to be more James F. Fries, MD: Stanford University School of Medicine, Palo Alto, CA. Address correspondence and reprint requests to James F. Fries, MD, Stanford University, Division of Rheumatology and Immunology, Stanford University School of Medicine, 1000 Welch Road, Suite 203, Palo Alto, CA 94304-5755. E-mail: jff@stanford.edu. Submitted for publication April 22, 2004; accepted in revised form June 22, 2004.