Considerable variation in NNT - A study based on Monte Carlo simulations

T. Wisloff, O. O. Aalen, Ivar Sønbø Kristiansen

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

    Resumé

    Objective: The aim of this analysis was to explore the variation in measures of effect, such as the number-needed-to-treat (NNT) and the relative risk (RR). Study Design and Setting: We performed Monte Carlo simulations of therapies using binominal distributions based on different true absolute risk reductions (ARR), number of patients (n), and the baseline risk of adverse events (p(0)) as parameters and presented results in histograms with NNT and RR. We also estimated the probability of observing no or a negative treatment effect, given that the true effect is positive. Results: When RR is used to express treatment effectiveness, it has a regular distribution around the expected value for various values of true ARR, n, and p(0). The equivalent distribution of NNT is by definition nonconnected at zero and is also irregular. The probability that the observed treatment effectiveness is zero or negative when the true value is positive depends on n, p(0), and the true ARR. In some cases, this probability is even higher than 50%. Conclusion: For realistic values of true ARR, n, and p(0), the observed NNT varies much more than the observed ARR and RR. Clinicians should use NNT cautiously when expressing treatment benefits. (C) 2011 Elsevier Inc. All rights reserved.
    OriginalsprogEngelsk
    TidsskriftJournal of Clinical Epidemiology
    Vol/bind64
    Udgave nummer4
    Sider (fra-til)444-450
    Antal sider7
    ISSN0895-4356
    DOI
    StatusUdgivet - 2011

    Citer dette

    Wisloff, T. ; Aalen, O. O. ; Sønbø Kristiansen, Ivar . / Considerable variation in NNT - A study based on Monte Carlo simulations. I: Journal of Clinical Epidemiology. 2011 ; Bind 64, Nr. 4. s. 444-450.
    @article{b6d45f2b4b7a42f38e2ee667c1ece829,
    title = "Considerable variation in NNT - A study based on Monte Carlo simulations",
    abstract = "Objective: The aim of this analysis was to explore the variation in measures of effect, such as the number-needed-to-treat (NNT) and the relative risk (RR). Study Design and Setting: We performed Monte Carlo simulations of therapies using binominal distributions based on different true absolute risk reductions (ARR), number of patients (n), and the baseline risk of adverse events (p(0)) as parameters and presented results in histograms with NNT and RR. We also estimated the probability of observing no or a negative treatment effect, given that the true effect is positive. Results: When RR is used to express treatment effectiveness, it has a regular distribution around the expected value for various values of true ARR, n, and p(0). The equivalent distribution of NNT is by definition nonconnected at zero and is also irregular. The probability that the observed treatment effectiveness is zero or negative when the true value is positive depends on n, p(0), and the true ARR. In some cases, this probability is even higher than 50{\%}. Conclusion: For realistic values of true ARR, n, and p(0), the observed NNT varies much more than the observed ARR and RR. Clinicians should use NNT cautiously when expressing treatment benefits. (C) 2011 Elsevier Inc. All rights reserved.",
    author = "T. Wisloff and Aalen, {O. O.} and {S{\o}nb{\o} Kristiansen}, Ivar",
    year = "2011",
    doi = "10.1016/j.jclinepi.2010.04.025",
    language = "English",
    volume = "64",
    pages = "444--450",
    journal = "Journal of Clinical Epidemiology",
    issn = "0895-4356",
    publisher = "Elsevier",
    number = "4",

    }

    Considerable variation in NNT - A study based on Monte Carlo simulations. / Wisloff, T.; Aalen, O. O.; Sønbø Kristiansen, Ivar .

    I: Journal of Clinical Epidemiology, Bind 64, Nr. 4, 2011, s. 444-450.

    Publikation: Bidrag til tidsskriftTidsskriftartikelForskningpeer review

    TY - JOUR

    T1 - Considerable variation in NNT - A study based on Monte Carlo simulations

    AU - Wisloff, T.

    AU - Aalen, O. O.

    AU - Sønbø Kristiansen, Ivar

    PY - 2011

    Y1 - 2011

    N2 - Objective: The aim of this analysis was to explore the variation in measures of effect, such as the number-needed-to-treat (NNT) and the relative risk (RR). Study Design and Setting: We performed Monte Carlo simulations of therapies using binominal distributions based on different true absolute risk reductions (ARR), number of patients (n), and the baseline risk of adverse events (p(0)) as parameters and presented results in histograms with NNT and RR. We also estimated the probability of observing no or a negative treatment effect, given that the true effect is positive. Results: When RR is used to express treatment effectiveness, it has a regular distribution around the expected value for various values of true ARR, n, and p(0). The equivalent distribution of NNT is by definition nonconnected at zero and is also irregular. The probability that the observed treatment effectiveness is zero or negative when the true value is positive depends on n, p(0), and the true ARR. In some cases, this probability is even higher than 50%. Conclusion: For realistic values of true ARR, n, and p(0), the observed NNT varies much more than the observed ARR and RR. Clinicians should use NNT cautiously when expressing treatment benefits. (C) 2011 Elsevier Inc. All rights reserved.

    AB - Objective: The aim of this analysis was to explore the variation in measures of effect, such as the number-needed-to-treat (NNT) and the relative risk (RR). Study Design and Setting: We performed Monte Carlo simulations of therapies using binominal distributions based on different true absolute risk reductions (ARR), number of patients (n), and the baseline risk of adverse events (p(0)) as parameters and presented results in histograms with NNT and RR. We also estimated the probability of observing no or a negative treatment effect, given that the true effect is positive. Results: When RR is used to express treatment effectiveness, it has a regular distribution around the expected value for various values of true ARR, n, and p(0). The equivalent distribution of NNT is by definition nonconnected at zero and is also irregular. The probability that the observed treatment effectiveness is zero or negative when the true value is positive depends on n, p(0), and the true ARR. In some cases, this probability is even higher than 50%. Conclusion: For realistic values of true ARR, n, and p(0), the observed NNT varies much more than the observed ARR and RR. Clinicians should use NNT cautiously when expressing treatment benefits. (C) 2011 Elsevier Inc. All rights reserved.

    U2 - 10.1016/j.jclinepi.2010.04.025

    DO - 10.1016/j.jclinepi.2010.04.025

    M3 - Journal article

    VL - 64

    SP - 444

    EP - 450

    JO - Journal of Clinical Epidemiology

    JF - Journal of Clinical Epidemiology

    SN - 0895-4356

    IS - 4

    ER -