Jana L. Keller

A comparison of Likert and visual analogue scales for measuring change in function

Many controlled trials rely on subjective measures of symptoms or quality of life as primary outcomes. The relative merits of different response options for these measures is an important, but largely unexplored, issue. Therefore, we compared the responsiveness of seven-point Likert vs visual analogue scales (VAS) in a questionnaire measuring quality of life in chronic lung disease. The VAS and seven-point scale versions of the questionnaire were administered to 28 patients before and after completing an inpatient respiratory rehabilitation program of known benefit. For all four dimensions of the questionnaire (dyspnea, fatigue, emotional function, and mastery) the VAS showed a larger improvement than the seven-point scale when both were standardized on a scale of 0-10. However, in each case the variability of the improvement was greater using the VAS. The difference in improvement between the two scales was not statistically significant. We conclude that the two methods of presenting response options show comparable responsiveness. The ease of administration and interpretation of the seven-point scale recommend its use in clinical trials.

Effectiveness of steroid therapy in acute exacerbations of asthma: A meta-analysis

The objective of this study was to determine the effect of steroid therapy on pulmonary function, admission rates, and relapse rates in patients presenting with acute exacerbations of asthma. Computerized MEDLINE and SCIENCE CITATION searches were combined with review of reference lists from book chapters and articles to identify published randomized trials on steroid interventions. Over 700 articles were reviewed by two independent reviewers who identified 30 relevant randomized controlled trials for analysis. Study validity was independently assessed by two reviewers and information regarding populations, interventions, and outcomes was abstracted. Binary outcomes were combined and reported as odds ratios (OR), using the Mantel-Haenszel method. Individual and pooled effect sizes (ES) were determined for pulmonary function data. The authors found that the use of steroids early in the treatment of asthmatic exacerbations reduces admissions in adults (common OR 0.47; 95% confidence interval (CI) 0.27, 0.79) and children (OR 0.06-0.42). They found steroids effective in preventing relapse in the outpatient treatment of asthmatic exacerbations (OR 0.15; CI 0.05, 0.44). Oral and intravenous steroids appear to have equivalent effects on pulmonary function in acute exacerbations (ES -0.07; CI -0.39, 0.25). The authors conclude that overall, steroid therapy provides important benefits to patients presenting to emergency departments with acute exacerbations of asthma. Further research into dosage, alternative routes of administration, and alternative outcome measures is needed.

N of 1 randomized trials for investigating new drugs

Presently, in the process of new drug development, large sample parallel group randomized trials are often begun without the detailed knowledge of optimal dose, most responsive patient group, and optimal outcomes which would be desirable. We propose that randomized trials in individual subjects (N of 1 RCTs) could be used to elucidate these issues at an early stage of drug development. In appropriate conditions N of 1 RCTs can be used to define the rapidity with which a drug begins and ceases its clinical action, the likely range of the optimal drug dose, and the optimal outcomes on which subsequent trials should focus. N of 1 RCTs can also generate initial estimates of the proportion of patients who respond to a new agent and for determining sample size, inclusion criteria, and dosage regimen(s) for subsequent parallel group trials. We provide an example of 14 N of 1 RCTs of amitriptyline in fibrositis that illustrate the ways in which N of 1 RCTs can elucidate these issues. The multiple uses of N of 1 RCTs suggest that the method has immense potential for use in the early phases of drug development programs.

An index of scientific quality for health reports in the lay press

Although the quality of health reporting has been criticized for being unscientific, evaluations of health care reporting have been limited by the lack of a reliable and credible measure of scientific quality. We developed an index of scientific quality (ISQ) for health-related news reports and tested its reliability and sensibility. Items were generated from a survey of the literature and experts in research methodology. Items that were unclear, confusing or discriminated poorly between articles of high and low scientific quality were revised or deleted in an iterative process wherein potential criteria were independently applied to samples of 5 to 15 articles by 6 raters. To test the reliability of the final criteria 60 articles were drawn from three sampling frames: newspapers, magazines, and professional journals. Articles were intentionally selected to obtain a wide range of quality and topics. Two categories of raters were used: research assistants and physicians with research training. All 6 raters assessed all 60 articles. The sensibility of the index was tested by a questionnaire with 13 items related to face validity and content validity as well as other aspects of sensibility. The questionnaire was completed by 20 researchers and 13 health and science writers. The final ISQ includes 7 items that address the extent to which a report allows readers to draw conclusions about the applicability, validity and practical importance of the information that is reported. Chance corrected agreement (kappa) among all 6 raters for overall scientific quality was 0.62 (SE 0.02). The index was found to be sensible with only one major problem, the need for judgment in making ratings. While some degree of subjectivity appears to be inevitable in rating the scientific quality of health reports, the ISQ is acceptable reliable and credible and should be useful for evaluating and improving the scientific quality of health reporting.

Acute Response to Bronchodilator: An Imperfect Guide for Bronchodilator Therapy in Chronic Airflow Limitation

We conducted a four-period cross-over randomized trial in which we found that patients with chronic airflow limitation demonstrated symptomatic improvement with both inhaled albuterol and oral theophylline. The response, however, was not uniform. We therefore tested the ability of acute change in forced expired volume in one second (FEV1) following inhaled beta agonist to predict long-term symptomatic response to albuterol and theophylline. We found that the reproducibility of acute change in FEV1 over three repetitions was poor (intraclass correlation 0.17). Furthermore, the mean improvement FEV1 following inhaled albuterol across the three repetitions did not relate closely to symptomatic response to either albuterol or theophylline. We conclude that acute response to inhaled beta agonist is not useful for identifying patients with chronic airflow limitation who are likely to benefit from bronchodilator treatment.

SHOULD STUDY SUBJECTS SEE THEIR PREVIOUS RESPONSES: DATA FROM A RANDOMIZED CONTROL TRIAL

To test the relative merits of administering questionnaires with previous responses available (the informed condition) or unavailable (the blind condition), we administered blind and informed versions of a quality of life questionnaire (the Chronic Respiratory Disease Questionnaire, or CRQ) in a randomized, double-blind trial of bronchodilators in chronic airflow limitation. The responsiveness of the two methods, as reflected in the p-values associated with salbutamol and theophylline effects were comparable for three of the four dimensions of the CRQ. The data suggested possible increased responsiveness of the informed method for the emotional function dimension of the questionnaire. Changes in the informed CRQ dyspnea and fatigue dimensions showed stronger correlations with changes in spirometry, 6 minute walk distance, and rating of dyspnea after the walk test than did blind administration. Further, changes in all four CRQ dimensions showed stronger correlations with corresponding global ratings using the informed questionnaire. These results suggest that by letting study subjects see their previous responses the validity of subjective measures of health status in clinical trials can be improved.

INTERPRETING CHANGES IN QUALITY-OF-LIFE SCORE IN N OF 1 RANDOMIZED TRIALS

To provide additional evidence regarding the plausible range of the differences in health-related quality of life (HRQL) questionnaire scores within which the minimal important difference (MID) falls, we reviewed the results of 32 randomized controlled trials in individual subjects (N of 1 RCTs) with chronic diseases. These trials had been conducted to establish whether a patient was obtaining more good than harm from a medication. Each N of 1 RCT included a series of pairs of treatment periods, one period on active drug, and one on placebo or alternative drug. We examined the relationship between small (MID), medium, and large differences between periods within pairs, as indicated by Global Ratings and differences between these same periods according to HRQL questionnaires. The results showed a mean difference of 0.29 points per question in HRQL questionnaire scores corresponded to the MID. Differences of approximately 0.66 points per question corresponded to a moderate difference as ranked by the Global Rating; difference of about 1.09 points per question represented marked difference.

An N of 1 Service: Applying the Scientific Method in Clinical Practice

The N of 1 service at our institution acts as a full referral service for clinicians who want a definitive answer to a difficult management question, and an instructional environment for clinicians who have more time and want to learn to run their own N of 1 RCT. The trial design is a double blind, randomized pair, multiple crossover. A number of methodologic issues are discussed, such as appropriateness of the patients problem to N of 1 trials, feasibility, types of measurement, such as clinical objective measurement and quality of life measurement, as well as timing of these measurements. The analysis issues include developing a reporting method which is statistically powerful and understandable to clinicians with little research background. Some of these issues have been well investigated and some have not.