Fig 1.
Estimated survival curves, numbers needed to treat (NNTs), and the average life gain (ALG) based on the reconstructed data from the radical prostatectomy study.
(A) Kaplan-Meier curves for the radical prostatectomy and watchful waiting groups. (B) The NNTs based on the difference in restricted mean survival times (NNTRMST) or the absolute risk reduction (NNTARR). The rescaled y-axis accommodating infinity to distinguish NNT to harm (NNTH) and NNT to benefit (NNTB). (C) The ALG for the NNTARR and NNTRMST during the follow-up time.
Table 1.
Comparison of the NNTARR and NNTRMST based on three real trials examples.
Fig 2.
Estimated survival curves, numbers needed to treat (NNTs), and the average life gain (ALG) based on the reconstructed data from the atezolizumab and nab-paclitaxel trial.
(A) Kaplan-Meier curves for the atezolizumab plus nab-paclitaxel and the placebo plus nab-paclitaxel groups. (B) The NNTs based on the difference in restricted mean survival times (NNTRMST) or the absolute risk reduction (NNTARR). The rescaled y-axis accommodating infinity to distinguish NNT to harm (NNTH) and NNT to benefit (NNTB). (C) The ALG for the NNTARR and NNTRMST during the follow-up time.
Fig 3.
Estimated survival curves, numbers needed to treat (NNTs), and the average life gain (ALG) based on the reconstructed data from the caplacizumab trial.
(A) Kaplan-Meier curves for the caplacizumab and placebo groups. (B) The NNTs based on the difference in restricted mean survival times (NNTRMST) or the absolute risk reduction (NNTARR). The rescaled y-axis accommodating infinity to distinguish NNT to harm (NNTH) and NNT to benefit (NNTB). (C) The ALG for the NNTARR and NNTRMST during the follow-up time.
Fig 4.
Hypothetical examples representing four different scenarios.
(A) Scenario 1: Survival curves with an increasing treatment effect over time. (B) Scenario 2: Survival curves converge at the end of the follow-up. (C) Scenario 3: Survival curves cross during the follow-up period. (D) Scenario 4: Survival curves with a mixture of short- and long- term effects. The area up to time t under each scenario between treatment and control groups represent the average life gain (ALG).
Fig 5.
Simulation results of NNTRMST and NNTARR under four hypothetical scenarios.
The upper panel presents the bias of NNT at different time points, and the lower panel shows the bias of the ALG at different time points. (A1-2) Scenario 1: Survival curves with an increasing treatment effect over time; (B1-2) Scenario 2: Survival curves converge at the end of the follow-up; (C1-2) Scenario 3: Survival curves cross during the follow-up period; (D1-2) Scenario 4: Survival curves with a mixture of short- and long- term effects.