Understanding Systematic Reviews

Disclaimer: I have NOT looked at this data…
…but it seems as though there’s a lack of understanding about systematic reviews. A friend of mine is a rep for the company that makes nesiritide – his take (presumably the company line on the issue) is that the JAMA authors simply did not include four of the seven articles in their systematic review – therefore the conclusions could not possibly be reliable. There seems to be no general understanding of the ability of lower quality trials to overestimate benefit. Certainly argues for more EBM literacy…

Perspective on the News About Nesiritide

Quote from Medscape/Journal Watch:
On April 20, 2005, the FDA approved a revised package insert for Natrecor (nesiritide), which is used widely as first-line therapy in the management of acute decompensated heart failure. This insert includes pooled mortality data from randomized trials of nesiritide versus standard therapy alone.

For 30-day mortality, data from seven randomized trials (involving 1717 patients) were analyzed. The overall 30-day rates were 5.3% with nesiritide and 4.3% with control therapy — a nonsignificant difference. For 180-day mortality, data from four randomized trials (involving 1167 patients) were analyzed. The overall 180-day rates were 21.7% with nesiritide and 21.5% with control therapy — also a nonsignificant difference.

The 30-day mortality data from this insert are at odds with 30-day mortality data from a meta-analysis recently published in JAMA (Journal Watch Cardiology May 6 2005). In that analysis, the overall 30-day mortality rates — based on three randomized trials (involving 862 patients) — were 6.5% with nesiritide and 4.0% with control therapy (hazard ratio, 1.86; P =0.04). This difference dropped just below statistical significance after adjustment for treatment and study characteristics (HR, 1.80; P =0.057).

The difference between the mortality findings published in JAMA and those published in the new package insert can be explained by the former’s exclusion of four trials that did not meet the JAMA authors’ inclusion criteria (e.g., open-label design, differences in drug-administration protocol, etc.)