Methods to estimate number needed to treat (NNT) in recent event-driven clinical trials comparing combination therapy versus monotherapy in pulmonary arterial hypertension

Study characteristics
 Duration weeks104+85+796971§64§
 Patients at risk154154253247462458
 Number of events50684677124183
 Cumulative proportion %32.544.
 Event rates per patient-year0.1620.2690.1200.2350.1970.344
NNT calculations
 Cumulative incidence8.6≈97.7≈86.5≈7
 Incidence rate# patient-years9.4≈108.7≈96.8≈7ƒ
 HR (estimated at month 24)7.4≈86.1≈76.7≈7
 At week 1610.5≈1133.3≈3412.2≈13
 At month 67.9≈812.0≈1210.4≈11
 At month 126.6≈77.1≈88.0≈8
 At month 246.0≈66.1≈77.5≈8

Using the Kaplan–Meier method, the NNT appears to be somewhat comparable across studies, especially at 12 months and beyond, where the NNT is relatively constant, implying a relatively constant absolute risk reduction. Note that in the AMBITION trial, the NNT estimation at week 16 is probably distorted because clinical worsening did not occur randomly, clustering around study visits. Per consensus, the NNT is most frequently rounded up to the nearest whole number. Moreover, although only mean NNT are presented, the 95% confidence intervals around the NNT, specifying the “limits” within which we can confidently state the true NNT lies, should be presented in index articles when the NNT is presented [16]. Finally, the results of the COMPASS trial [17] are not computed in the table, since NNT calculation is controversial in the absence of statistically significant results. CombiTx: combination therapy; monoTx: monotherapy; HR: hazard ratio. #: the incidence rate of the outcome may be used as a measure of outcome frequency. It is computed as the number of patients with the outcome divided by the total amount of person-time or patient-time, to account for varying follow-up times. This NNT represents the number of patient-years to prevent one clinical worsening event. : the event probability in the active treatment group at a fixed time can be computed from the event probability in the control group at that time and the HR. If at some specified time, t, the survival probability in the control group is Sc(t) then the survival probability in the active group is [Sc(t)]HR [14]. The NNT is estimated as 1/{[Sc(t)]HR–Sc(t)}. As example, the AMBITION trial provides an HR of 0.50. At 104 weeks, ≈63% of patients did not meet the primary end-point (figure 2). Then, NTT = 1/[0.630.50−0.63] = 6.1≈7. +: mean duration for the overall study population, whereas event rates and NNT are estimated for the subgroup of patients on background therapy only (i.e. combination therapy versus monotherapy). §: median duration of follow-up. ƒ: assuming that median follow-up duration approaches mean follow-up duration.