Data from 22 RCTs (N=50,246) were included in the regression analysis.
Data from 11 RCTs were included in the calculation of the NNT per treatment delay period.
The number of patients randomised within 2 hours of symptom onset was 5,762; 10,435 patients were randomised within 2 to 3 hours of symptom onset.
The absolute mortality reductions per 1000 treated patients were, for patients who presented:
Within 1 hour of symptom onset, 65 lives saved (SD 14, 95% CI: 38, 93); NNT 15.
In second hour after symptom onset, 37 lives saved (SD 9, 95% CI: 20, 55); NNT 27.
Two to 3 hours after symptom onset, 26 lives saved (SD 6, 95% CI: 14, 37); NNT 38.
Three to 6 hours after symptom onset, 29 lives saved (SD 5, 95% CI: 19, 40); NNT 34.
Six to 12 hours after symptom onset, 18 lives saved (SD 6, 95% CI: 7, 29); NNT 56.
Twelve to 24 hours after symptom onset, 9 lives saved (SD 7, 95% CI: -5, +22); NNT no significant difference.
The ORs were significantly different over the 6 treatment delay periods (Breslow-Day test, P=0.001.
The proportionate mortality reductions were 48% (95% CI: 31, 61) for patients treated within 1 hour, 44% (95% CI: 32, 53) for those treated within 2 hours, and 20% (95% CI: 15, 25) for those treated later.
The best-fit regression model (correlation, R2=0.32) was found to be: absolute benefit per 1000 treated patients equals 19.4 minus 0.6 (treatment delay in hours) plus 29.3 (treatment delay in hours) minus 1.
The above findings were not changed after the sensitivity analysis.