Saturday, November 19, 2005
BMJ has just published a metaanalysis evaluating the risks-benefits of sedatives for treating insomnia in the elderly. Sedatives here means benzodiazepines and benzo receptor agonists (zolpidem/Ambien, zaleplon/Sonata, etc). It aggregates data from 24 (total n=2400) randomized trials of sedative agents in people >60 years old. Most of the studies were for 2 weeks or less, although a couple were longer. They found that the number-needed-to-treat for any measurable sleep "improvement" (which was measured heterogeneously across the studies) was 13, & the number-needed-to-harm was 6 (this number catches any measurable side effect...from benign--GI upset--to severe--falls, etc). The average increase in sleep length was 25 minutes vs. placebo. The head-to-head studies of benzo's vs benzodiazepine receptor agonists essentially found no difference in effectiveness or adverse effects, including cognitive ones. Morning or daytime fatigue was significantly worse with active treatment vs. placebo, with quite a strong odds ratio (3.8, 1.88 to 7.8, p<.001). They did not mention that they were able to compare benzos to non-benzos specifically for this outcome.
All of this has me asking why do I prescribe these drugs at all and why do patients like them so much? What's the point if you feel worse the next day? This is naive of course, and insomnia is an unpleasant symptom that may have nothing to do with actual sleep/daytime energy/etc. for many patients, and these agents are effective (somewhat) for this symptom. What disappoints me is how relatively high the NNT is: 13 is great when you're saving lives through cardiovascular drugs but not to treat an unpleasant symptom. In addition, the NNH is not helpful in deciding, for a given patient, if the benefits outweigh the risks. GI upset or headache are qualitatively different from falls/severe psychomotor side effects and aggregating a global NNH is clinically meaningless.
And, I must add that in the middle of the article sits this sentence:
A significant improvement in sleep quality improved significantly[.] It actually makes sense in context but there must be a better way to say this...