S.D. Walter

Need for expertise based randomised controlled trials

Surgical procedures are less likely to be rigorously evidence based than drug treatments because of difficulties with randomisation. Expertise based trials could be the way forward Although conventional randomised controlled trials are widely recognised as the most reliable method to evaluate pharmacological interventions,1 2 scepticism about their role in nonpharmacological interventions (such as surgery) remains.3–6 Conventional randomised controlled trials typically randomise participants to one of two intervenions (A or B) and individual clinicians give intervention A to some participants and B to others. An alternative trial design, the expertise based randomised controlled trial, randomises participants to clinicians with expertise in intervention A or clinicians with expertise in intervention B, and the clinicians perform only the procedure they are expert in. We present evidence to support our argument that increased use of the expertise based design will enhance the validity, applicability, feasibility, and ethical integrity of randomised controlled trials in surgery, as well as in other areas. We focus on established surgical interventions rather than new surgical procedures in which clinicians have not established expertise. Investigators have used the expertise based design when conventional randomised controlled trials were impossible because different specialty groups provided the interventions under evaluation—for example, percutaneous transluminal coronary angioplasty versus coronary artery bypass graft surgery.7–9 In 1980, Van der Linden suggested randomising participants to clinicians committed to performing different interventions in an area in which a conventional randomised controlled trial was possible.10 Since that time, however, the expertise based design has been little used, even in areas where it has high potential (such as, surgery, physiotherapy, and chiropractic). ### Differential expertise between procedures Because it takes training and experience to develop expertise in surgical interventions, individual surgeons tend to solely or primarily use a single surgical approach to treat a specific problem.10 11 …

Dalteparin versus unfractionated heparin in critically ill patients.

BACKGROUND The effects of thromboprophylaxis with low-molecular-weight heparin, as compared with unfractionated heparin, on venous thromboembolism, bleeding, and other outcomes are uncertain in critically ill patients. METHODS In this multicenter trial, we tested the superiority of dalteparin over unfractionated heparin by randomly assigning 3764 patients to receive either subcutaneous dalteparin (at a dose of 5000 IU once daily) plus placebo once daily (for parallel-group twice-daily injections) or unfractionated heparin (at a dose of 5000 IU twice daily) while they were in the intensive care unit. The primary outcome, proximal leg deep-vein thrombosis, was diagnosed on compression ultrasonography performed within 2 days after admission, twice weekly, and as clinically indicated. Additional testing for venous thromboembolism was performed as clinically indicated. Data were analyzed according to the intention-to-treat principle. RESULTS There was no significant between-group difference in the rate of proximal leg deep-vein thrombosis, which occurred in 96 of 1873 patients (5.1%) receiving dalteparin versus 109 of 1873 patients (5.8%) receiving unfractionated heparin (hazard ratio in the dalteparin group, 0.92; 95% confidence interval [CI], 0.68 to 1.23; P=0.57). The proportion of patients with pulmonary emboli was significantly lower with dalteparin (24 patients, 1.3%) than with unfractionated heparin (43 patients, 2.3%) (hazard ratio, 0.51; 95% CI, 0.30 to 0.88; P=0.01). There was no significant between-group difference in the rates of major bleeding (hazard ratio, 1.00; 95% CI, 0.75 to 1.34; P=0.98) or death in the hospital (hazard ratio, 0.92; 95% CI, 0.80 to 1.05; P=0.21). In prespecified per-protocol analyses, the results were similar to those of the main analyses, but fewer patients receiving dalteparin had heparin-induced thrombocytopenia (hazard ratio, 0.27; 95% CI, 0.08 to 0.98; P=0.046). CONCLUSIONS Among critically ill patients, dalteparin was not superior to unfractionated heparin in decreasing the incidence of proximal deep-vein thrombosis. (Funded by the Canadian Institutes of Health Research and others; PROTECT ClinicalTrials.gov number, NCT00182143.).

Choice of effect measure for epidemiological data

The debate concerning the choice of effect measure for epidemiologic data has been renewed in the literature, and it suggests some continuing disagreement between the pertinent clinical and statistical criteria. In this article, some defining characteristics of the main choices of effect measure [risk difference (RD), relative risk (RR), and odds ratio (OR)] for binary data are presented and compared, with consideration of both the clinical and statistical perspectives. Relationships of these measures to the relative risk reduction (RRR) and number needed to treat (NNT) are also discussed. A numerical comparison of models of constant RD, RR, and OR is made to assess when and by how much they might differ in practice. Typically the models show only small numerical differences, unless extreme extrapolation is involved. The RD and RR models can predict impossible event rates, either less than zero or greater than 100%. Each measure has potential theoretical justification. RD and RR may enjoy some advantages for communication of risk, but OR may be preferred for data analysis. A clear distinction should be maintained between the objectives of data analysis and subsequent risk communication, and different effect measures may be needed for each.

Meta-analysis of diagnostic tests with imperfect reference standards.

We present a method to estimate the summary receiver operating characteristic (SROC) curve for combining information on a diagnostic test from several different studies. Unlike previous methods that assume the reference standard to be error free, our approach allows for the possibility of errors in the reference standard, through use of a latent class model. The model provides estimates of the sensitivity and specificity of the diagnostic test and the case prevalence in each study; these parameters can then be used in a meta-analysis, for example, using the regression method proposed by Moses et al., of a measure of test discrimination on a measure of the diagnostic threshold, to fit the SROC. The method is illustrated with an example on Pap smears that shows how adjusting for imperfection in the reference standard typically reduces the scatter of data in the SROC plot, and tends to indicate better performance of the test than otherwise.

Payments for care at private for-profit and private not-for-profit hospitals: a systematic review and meta-analysis

Background: It has been shown that patients cared for at private for-profit hospitals have higher risk-adjusted mortality rates than those cared for at private not-for-profit hospitals. Uncertainty remains, however, about the economic implications of these forms of health care delivery. Since some policy-makers might still consider for-profit health care if expenditure savings were sufficiently large, we undertook a systematic review and meta-analysis to compare payments for care at private for-profit and private not-for-profit hospitals. Methods: We used 6 search strategies to identify published and unpublished observational studies that directly compared the payments for care at private for-profit and private not-for-profit hospitals. We masked the study results before teams of 2 reviewers independently evaluated the eligibility of all studies. We confirmed data or obtained additional data from all but 1 author. For each study, we calculated the payments for care at private for-profit hospitals relative to private not-for-profit hospitals and pooled the results using a random effects model. Results: Eight observational studies, involving more than 350 000 patients altogether and a median of 324 hospitals each, fulfilled our eligibility criteria. In 5 of 6 studies showing higher payments for care at private for-profit hospitals, the difference was statistically significant; in 1 of 2 studies showing higher payments for care at private not-for-profit hospitals, the difference was statistically significant. The pooled estimate demonstrated that private for-profit hospitals were associated with higher payments for care (relative payments for care 1.19, 95% confidence interval 1.07–1.33, p = 0.001). Interpretation: Private for-profit hospitals result in higher payments for care than private not-for-profit hospitals. Evidence strongly supports a policy of not-for-profit health care delivery at the hospital level.

LOST to follow-up Information in Trials (LOST-IT): a protocol on the potential impact

BackgroundIncomplete ascertainment of outcomes in randomized controlled trials (RCTs) is likely to bias final study results if reasons for unavailability of patient data are associated with the outcome of interest. The primary objective of this study is to assess the potential impact of loss to follow-up on the estimates of treatment effect. The secondary objectives are to describe, for published RCTs, (1) the reporting of loss to follow-up information, (2) the analytic methods used for handling loss to follow-up information, and (3) the extent of reported loss to follow-up.MethodsWe will conduct a systematic review of reports of RCTs recently published in five top general medical journals. Eligible RCTs will demonstrate statistically significant effect estimates with respect to primary outcomes that are patient-important and expressed as binary data. Teams of 2 reviewers will independently determine eligibility and extract relevant information from each eligible trial using standardized, pre-piloted forms. To assess the potential impact of loss to follow-up on the estimates of treatment effect we will, for varying assumptions about the outcomes of participants lost to follow-up (LTFU), calculate (1) the percentage of RCTs that lose statistical significance and (2) the mean change in effect estimate across RCTs. The different assumptions we will test are the following: (1) none of the LTFU participants had the event; (2) all LTFU participants had the event; (3) all LTFU participants in the treatment group had the event; none of those in the control group had it (worst case scenario); (4) the event incidence among LTFU participants (relative to observed participants) increased, with a higher relative increase in the intervention group; and (5) the event incidence among LTFU participants (relative to observed participants) increased in the intervention group and decreased in the control group.DiscussionWe aim to make our objectives and methods transparent. The results of this study may have important implications for both clinical trialists and users of the medical literature.

The usefulness of the Denver Developmental Screening Test to predict kindergarten problems in a general community population.

The Denver Developmental Screening Test (DDST) was administered to 2,569 children five to seven months prior to starting kindergarten in September 1980 in a geographically well-defined community. The test was administered by trained public health nurses. At the end of the 1980-1981 school year, all 163 kindergarten teachers in the area completed a rating form for each child in their class. The rating form determined global ratings of: 1) learning abilities; 2) classroom behavior; 3) amount of special attention required; and 4) referrals to special education services outside the classroom. The specificity of the DDST in predicting kindergarten teacher ratings was 99 per cent for all areas. Test sensitivity varied from 5 per cent to 10 per cent in detecting problems in the four areas. The predictive values of an positive test varied from 31 per cent for behavior problems to 62 per cent for extra attention required in the classroom. Negative test predictive values varied from 79 per cent to 93 per cent. These results based on kindergarten teacher ratings suggest that, because of the low sensitivity and modest predictive value, the DDST may be relatively inefficient to use in a school entry screening program in a general community population of children.

PMI25: THE RELATION OF DISTRIBUTION-AND ANCHOR-BASED APPROACHES ON INTERPRETATION OF CHANGES IN HEALTH-RELATED QUALITY OF LIFE

and disability, issues often already assessed in trials and which may not be of direct concern to the respondent. Such assessments fail to account for interactions between health status and other influences on QoL, such as economic, social and environmental factors. The needs-based model argues that QoL is the extent to which individuals are able to meet their needs. Specific diseases influence different needs and interventions are effective where they allow more needs to be met. Relevant needs are identified from qualitative interviews with patients. The approach has been applied in the development of QoL instruments specific to depression, adult growth hormone deficiency, genital herpes, migraine, incontinence, urogenital atrophy, male erectile difficulties, rheumatoid arthritis, care givers of Alzheimer patients, ankylosing spondylitis, atopic dermatitis, psoriasis, lupus and psoriatic arthritis. All language versions of these instruments have excellent psychometric properties, are sensitive to clinical improvements and several are now the QoL instruments of choice for clinical trials. CONCLUSIONS: The needs-based model provides a true “QoL” measurement that complements clinical assessments of health status.

Statistical significance and fragility criteria for assessing a difference of two proportions

This paper compares the traditional methods of statistical inference on the data from biomedical studies with a proposed index of fragility in the results. In general, for any given study there are 8 possible combinations of conclusions regarding statistical significance, quantitative significance and fragility. The 8 possibilities are considered in turn with respect to how studies in each group might be interpreted. Numerical examples show that not all 8 possibilities need be attainable with a given study design, and that the relative likelihood of them occurring can vary widely. It is concluded that the fragility index may be a useful adjunct to conventional statistical inference, with certain intuitive appeal, but that more empirical experience is needed with the fragility method.

The impact of a public health nurse intervention on influenza vaccine acceptance.

Public health clients in an Ontario community 65 years of age or older were randomly allocated to receive an intervention by a public health nurse during a home visit promoting either influenza immunization or safety measures. There was no statistically significant differences in influenza immunization rates between these two groups (56.1% vs 56.6%). Men were significantly more likely to receive immunization.