Advertisement

The NNTnet metric is not new, not easy to use, and not routinely applied in medical research

  • Ralf Bender
    Correspondence
    Corresponding author. Department Medical Biometry, Institute for Quality and Efficiency in Health Care, Im Mediapark 8, D–50670 Cologne, Germany. Tel.: +49 221 35685 451; fax: +49 221 35685 10.
    Affiliations
    Department of Medical Biometry, Institute for Quality and Efficiency in Health Care (IQWiG), Cologne, Germany
    Faculty of Medicine, University of Cologne, Germany
    Search for articles by this author
      Editor:
      With interest, I read the article by Li et al. [
      • Li G.
      • Lip G.Y.H.
      • Marcucci M.
      • Thabane L.
      • Tian J.
      • Levine M.A.H.
      The number needed to treat for net effect (NNTnet) as a metric for measuring combined benefits and harms.
      ] regarding the NNTnet metric to apply the NNT concept to multiple endpoints. The authors claim that they “propose a new metric, the ‘NNT for net effect’ or NNTnet to present the combined benefit and harm effects of an intervention or therapy based on NNT-type information” and that “no studies have explicitly used the NNT information embedded in a net benefit (or harm) approach, which leaves an important research gap between the net benefit (or harm) concept and the NNT application [
      • Li G.
      • Lip G.Y.H.
      • Marcucci M.
      • Thabane L.
      • Tian J.
      • Levine M.A.H.
      The number needed to treat for net effect (NNTnet) as a metric for measuring combined benefits and harms.
      ].”
      However, the proposed basic metric NNTnet = 1/ARRnet = 1/(ARR−ARI), where ARR is the absolute risk reduction regarding an endpoint showing benefit and ARI is the absolute risk increase for another endpoint showing harm, has already been proposed by Riegelman and Schroth [
      • Riegelman R.
      • Schroth W.S.
      Adjusting the number needed to treat: incorporating adjustments for the utility and timing of benefits and harms.
      ] more than 25 years ago. Extensions of the basic metric such as incorporation of utilities and timing have also been considered by Riegelman and Schroth [
      • Riegelman R.
      • Schroth W.S.
      Adjusting the number needed to treat: incorporating adjustments for the utility and timing of benefits and harms.
      ]. Similar measures have been considered in detail in the benefit-risk analysis literature, including methods for visualization of results and application in regulatory decision-making [
      • Holden W.L.
      Benefit-risk analysis: a brief review and proposed quantitative approaches.
      ,
      • Najafzadeh M.
      • Schneeweiss S.
      • Choudhry N.
      • Bykov K.
      • Kahler K.H.
      • Martin D.P.
      • et al.
      A unified framework for classification of methods for benefit-risk assessment.
      ,
      • Nixon R.
      • Dierig C.
      • Mt-Isa S.
      • Stockert I.
      • Tong T.
      • Kuhls S.
      • et al.
      A case study using the PrOACT-URL and BRAT frameworks for structured benefit risk assessment.
      ].
      On the one hand, the combination of benefits and harms regarding multiple endpoints in one metric solves one limitation of the original NNT, namely that it refers only to one single endpoint. On the other hand, the intuitive meaning of the simple NNT is lost if multiple endpoints with incorporation of utilities are combined in one metric with an unclear scale. In addition, it is difficult to derive appropriate confidence intervals for combined NNTnet metrics. Maybe these are reasons why such extensions of the simple NNT are not routinely applied in medical research. However, these issues have already been discussed in the scientific literature [
      • Bender R.
      Number needed to treat: overview.
      ].
      In summary, the NNTnet metric is not new, not intuitive after adjustments for timing and utilities, and currently not routinely applied in medical research.

      Supplementary data

      References

        • Li G.
        • Lip G.Y.H.
        • Marcucci M.
        • Thabane L.
        • Tian J.
        • Levine M.A.H.
        The number needed to treat for net effect (NNTnet) as a metric for measuring combined benefits and harms.
        J Clin Epidemiol. 2020; 125: 100-107
        • Riegelman R.
        • Schroth W.S.
        Adjusting the number needed to treat: incorporating adjustments for the utility and timing of benefits and harms.
        Med Decis Making. 1993; 13: 247-252
        • Holden W.L.
        Benefit-risk analysis: a brief review and proposed quantitative approaches.
        Drug Saf. 2003; 26: 853-862
        • Najafzadeh M.
        • Schneeweiss S.
        • Choudhry N.
        • Bykov K.
        • Kahler K.H.
        • Martin D.P.
        • et al.
        A unified framework for classification of methods for benefit-risk assessment.
        Value Health. 2015; 18: 250-259
        • Nixon R.
        • Dierig C.
        • Mt-Isa S.
        • Stockert I.
        • Tong T.
        • Kuhls S.
        • et al.
        A case study using the PrOACT-URL and BRAT frameworks for structured benefit risk assessment.
        Biom J. 2016; 58: 8-27
        • Bender R.
        Number needed to treat: overview.
        in: Balakrishnan N. Brandimarte P. Everitt B.S. Molenberghs G. Ruggeri F. Piegorsch W. Wiley StatsRef: statistics reference online. Wiley, Chichester2017 (Available at)