ACP J Club. 1997 Sep-Oct;127:A-16. doi:10.7326/ACPJC-1997-127-2-A16
To the Editor
A report in ACP Journal Club and Evidence-Based Medicine is an endorsement of the selected article and endows it with the status of "truth." Health care planners, managers, and clinicians, too busy to read the literature, may be overly influenced by these abstracts, for example, with the recent report on the effect of psychosocial interventions in cardiac rehabilitation programs (1).
There are potentially quite serious problems with meta-analysis, particularly if the studies being pooled lack rigor or are an incomplete set of studies (2). Several writers have pointed out limitations of this new approach to reviewing the literature (3, 4), but this example suggests that it may be important to remind readers of these limitations again.
A near halving of 2-year mortality is too good to be true and should evoke an even more critical approach than usual, even if the abstract writer and commentator were unaware of the findings of the largest trial in this area (with as many patients as all previous trials combined) (5).
Finally, the title and the conclusion for the abstract add further emphasis to an interpretation by the authors of the pooling, which is not justified in the original trial data and not claimed by authors of the original trials.
It is doubtful whether this report or the article on which it is based will, in the longer term, serve the best interests of the cardiac patient.
Robert West, PhD
University of Wales College of Medicine
Cardiff, Wales, UK
Linden has responded elsewhere (6).
These concerns center on estimating the effect of psychosocial interventions on mortality. The odds ratio for death in Linden's review was 1.70 (95% CI 1.06 to 2.64), favoring addition of a psychosocial intervention to usual cardiac rehabilitation (1). Of note, in their abstract (5), Jones and West reported a relative risk of 0.68 (95% CI 0.44 to 1.04) for 6-month mortality, favoring psychosocial intervention, which is consistent with the results of Linden's meta-analysis. The psychosocial interventions and the comparison groups in different studies varied in the details of the care they received, and the studies were too small to determine whether the effects of some interventions differed in mortality or other events.
In their letter to the editor about the original Linden report (7), West and Jones cited 4 studies that were not included in Linden's overview. 2 did not report mortality by group, and 1 lacked an appropriate comparison group. Their recently published study (8) was previously available only in abstract form (7). Publication bias is a concern in meta-analysis. Linden and colleagues stated that they used only published peer-reviewed reports and did not estimate the number of unpublished reports necessary to negate the result.
In the full report by Jones and West of 2328 patients discharged from hospital after myocardial infarction (8), who were unselected for psychosocial distress and followed for 12 months, 6-month mortality appeared to be less in the treatment group and survival curves rejoined between 6 to 12 months, but only 75% of the patients in the experimental group completed the 7-week intervention. This report adds to the knowledge of post-MI rehabilitation but is not the final answer any more than Linden's overview. Moreover, Linden and colleagues question whether psychosocial intervention is cost-effective, so any differences in the findings may not be of great clinical importance.
Further investigation is warranted given the conflicting results of the overview (1) and the recent trial (8). Identifying subgroups of patients most likely to benefit and determining the most cost-effective and efficient means of delivering psychosocial interventions are 2 goals that may be achieved through further study.
Craig Redfern, DO
Providence Portland Medical Center
Portland, Oregon, USA
1. Redfern C.Meta-analysis: Psychosocial interventions augment cardiac rehabilitation programs. ACP J Club. 1996 Sept-Oct;125:34. Evidence-Based Medicine. 1996 Sept-Oct;1:178. Comment on: Linden W, Stossel C, Maurice J. Arch Intern Med. 1996;156:745-52.