Pain research methods: NNT, trial designs and clinical importance

Read pain trials with three questions: what is the absolute difference (NNT), who was actually randomized (enrichment), and is the change clinically important (30% moderate, 50% substantial, judged within patients, not between groups).

1. NNT is the reciprocal of the absolute difference. Subtract the control (placebo) response rate from the treatment rate, then take 1 divided by that difference; always state the outcome and time frame ¹.
2. NNH mirrors NNT for harm. NNH is 1 divided by the absolute risk increase; in neuropathic pain meta-analyses it is often the number treated for one dropout from adverse effects ¹².
3. Neuropathic pain NNTs are modest. In the 2025 NeuPSIG meta-analysis of 313 trials, NNTs for treatment response were 4.6 for TCAs, 7.4 for SNRIs and 8.9 for α2δ ligands ⁷.
4. Enriched withdrawal trials select responders. In enriched enrollment randomized withdrawal, only patients who tolerate and respond to open-label drug are randomized, so results describe continuing benefit in early responders; about a third of entrants in one such trial were never randomized ³.
5. IMMPACT benchmarks are within-patient. A pain reduction of at least 30% is moderately important and at least 50% substantial; Farrar linked about 2 points or 30% to "much improved" ⁴⁵.
6. Between-group differences need case-by-case judgment. Applying within-subject thresholds to active-versus-placebo differences is a common misinterpretation; because large samples make trivial differences statistically significant, magnitude must be judged separately ⁶.

A confidence interval for NNT that includes infinity means the result is not statistically significant, and wide intervals signal imprecision; report the interval with the point estimate, not the NNT alone ¹.

• In the 2015 NeuPSIG review, publication bias analysis suggested about a 10% overstatement of neuropathic pain treatment effects ².
• Peer-reviewed publications reported larger effects than trial results posted online ².
• In the tapentadol withdrawal trial, 395 of 588 open-label entrants were randomized ³.
• IMMPACT recommends using two or more methods to judge the clinical importance of change ⁴.
• IMMPACT core domains: pain intensity, physical functioning, emotional functioning and participant global impression of change ⁴.
• Farrar pooled 2724 patients from 10 pregabalin trials; percentage change related consistently to global change regardless of baseline pain ⁵.
• Higher baseline pain requires larger raw point changes to be clinically important, so percentage change travels better ⁵.
• When baseline risk is high, odds ratios look more extreme than relative risks; for rare outcomes they converge ¹.

• Computing NNT from the treatment response rate alone, ignoring placebo ¹.
• Generalizing enriched-withdrawal effects to all patients starting a drug ³.
• Equating statistical significance with clinical importance ⁶.
• Calling a between-group difference unimportant because it is under 30% ⁶.