Evaluating a sleep-therapy abstract
A pre-dental student is reading a research abstract for a journal club. The abstract reports: 'In a randomized trial, 240 adults with chronic insomnia were assigned to 6 weeks of either guided sleep restriction therapy (SRT) or sleep hygiene education (SHE). The primary outcome was change in wake after sleep onset (WASO) measured by actigraphy. SRT reduced WASO by 22 minutes more than SHE (95% CI 8 to 36 minutes; p = 0.002). Secondary outcomes favored SRT for sleep efficiency and Insomnia Severity Index. However, 28% of SRT participants reported daytime sleepiness during the first two weeks, versus 9% of SHE participants; 12% of SRT participants dropped out, versus 5% of SHE participants. The authors conclude that SRT should be first-line treatment for chronic insomnia in primary care.' The student must evaluate whether the conclusion follows from the data, identify the most important threat to validity, and propose one additional analysis or control that would strengthen the inference.
1. Identify the main comparison and primary outcome, and evaluate whether the conclusion 'SRT should be first-line treatment' follows. Explain.
Read the worked answer
Main comparison: SRT versus SHE; primary outcome: actigraphy-measured change in WASO over 6 weeks. The trial supports a short-term advantage of SRT over SHE: the 22-minute difference has a 95% CI of 8 to 36 minutes, excluding no difference.
But 'first-line treatment in primary care' is a broader claim. It requires comparison with usual care and other active treatments, evidence beyond 6 weeks, and demonstration that benefits outweigh harms. The abstract also reports more daytime sleepiness and more dropout with SRT. So the conclusion is overreach: the data support efficacy versus SHE in this trial, not a policy recommendation.
2. What is the most important threat to internal validity, and how could it bias the WASO difference?
Read the worked answer
Differential attrition: 12% dropout with SRT versus 5% with SHE. If SRT dropouts were people who could not tolerate sleep restriction or whose sleep did not improve, a complete-case analysis would remove poor SRT outcomes and exaggerate SRT's benefit. Conversely, if SHE dropouts had worse outcomes, the bias could go the other way, but the asymmetry itself is the threat.
The analysis should be intention-to-treat with a prespecified imputation, such as multiple imputation or a worst-case sensitivity analysis. Actigraphy is objective, which reduces detection bias, but missing actigraphy data are still a problem.
3. Propose one additional analysis or control that would strengthen causal inference about SRT specifically, and explain what it would rule out.
Read the worked answer
Add an active attention-control arm matched for therapist contact, time, and expectancy, such as general sleep education with equal contact. This would rule out the possibility that the SRT advantage is due to extra attention or contact rather than sleep restriction itself. Alternatively, report intention-to-treat with multiple imputation and a dose-response analysis of adherence. The active control is stronger because it addresses a different causal explanation, not just missing data.