Original article
Number needed to treat: easily understood and intuitively meaningful?: Theoretical considerations and a randomized trial

https://doi.org/10.1016/S0895-4356(02)00432-8Get rights and content

Abstract

Graphic representation was used to explore to what extent the number needed to treat (NNT) conveys the appropriate notion of benefit for the individual patient in interventions aimed at delaying adverse events. A sample of the Danish population (n = 675) was interviewed face to face, and asked whether they would consent to a hypothetical drug that reduces the risk of heart attack. The benefit of the drug was expressed in terms of NNT and was randomly set at 10, 25, 50, 100, 200, and 400. NNT does not convey information on the proportion of patients being helped by an intervention or the size of the delay of the adverse event intended to be prevented. The proportion of people consenting to the hypothetical drug was about 80%, irrespective of NNT, and some of those who rejected the drug misinterpreted the meaning of NNT. Lay people may have difficulties in understanding the meaning of NNT, and clinicians may do well to use the NNT with caution until more is known about how patients comprehend it.

Introduction

Considerable proportions of health care resources are devoted first to detect and second to ameliorate risk factors such as hypertension, dyslipidaemia, osteoporosis, or certain types of life style. An important issue is then how to inform clinicians and patients about the benefits of interventions that delay or prevent chronic diseases. These benefits can be expressed in several ways such as absolute risk reduction (ARR), its reciprocal, the number needed to be treated (NNT), relative risk reduction (RRR), differences in event-free patients, the number of avoided events within a population, or as the gain in life expectancy.

First proposed by Laupacis and coworkers [1], the NNT has been recommended for communicating benefits from risk reductions 2, 3, 4, 5, 6. Advocates of evidence-based medicine claim that the “number needed to be treated (NNT) to prevent one event is the most useful measure of clinical effort … patients must expend in order to help them avoid bad outcomes” [7]. It is further claimed that NNT is “a currency for making decisions” [8], that it is “easily understood by clinicians” [1], and that it “has intuitive meaning” [9]. The use of NNT has been expanded to encompass harm (“number needed to harm”—NNH), screening (“number needed to screen”) [10], and education (“number needed to educate”) [11].

There is ample evidence that doctors 12, 13, 14, health care managers [15], and patients [16] are more likely to accept interventions where the benefits are presented in terms of RRR rather than ARR. Some authors claim that the RRR “may overstate the effectiveness of the treatment” [17], but the basis for judging what is “overstating” is unclear.

Although NNT is often used in medical publications and possibly also in clinical practice, there is little evidence that patients or doctors make more appropriate decisions when they use NNT rather than other measures of benefit [18]. By “appropriate” we mean decisions that are in line with those that patients and doctors would make when they have received and comprehended all relevant information.

In the following we will examine some theoretical properties of NNT, develop hypotheses about how lay people comprehend NNT, and test them empirically.

Section snippets

Theoretical properties of NNT

If the risk of an adverse event is Dc in a control group and Di in an intervention group at some point in time, the proportion (Dc − Di)/Dc is RRR while the difference (Dc − Di) is ARR. The reciprocal of ARR is NNT. In the following, NNT denotes the concept and nnt a specific value of it. The nnt expresses how many patients on average have been treated per event avoided at a specific point in time. When NNT sometimes is interpreted as ‘the number of patients needed to be treated for x years in

Methods

A random sample (n = 675) of noninstitutionalised Danes aged 20–74 was interviewed face to face through a so-called “Computer Assisted Personal Interviewing” by Gallup Inc, Denmark. When sampled individuals were not at home, up to two repeat calls were made. The interview questions for this paper were clustered with those for other customers of Gallup, but our items were placed at the beginning of the interview due to their complexity. For each week of interviewing, items from different

Results

In total, 675 subjects were interviewed. Their mean age was 44 years, and 51% were female, the same as for the Danish population of the same age range. The income distribution was similar to that of the Danish population, while the study sample had a lower proportion of individuals with elementary school only (21% vs. 37% in the general population). The health status was reported excellent among 29% of the respondents and less or poor among 10%. Heart disease was reported among 3.3% of the

Discussion

NNT has important shortcomings as a single measure of health benefit in that NNT measures benefit at an arbitrary point in time and fails to capture effects over time. Contrary to previous claims, NNT does not indicate the individual patient's probability of having benefit from a treatment when the aim of it is to delay an adverse event. The empirical study presented here lends little support to the claim that NNT is easily understood in that nnt had little influence on the proportions that

Acknowledgements

Financial support: Danish Heart Foundation. Comments from Gavin Mooney on the manuscript are gratefully acknowledged. Conflict of interest: none.

References (26)

  • D.L Sackett et al.

    Evidence-based medicine. How to Practice and teach EBM

    (2000)
  • H.J McQuay et al.

    Using numerical results from systematic reviews in clinical practice

    Ann Intern Med

    (1997)
  • R Riegelman et al.

    Adjusting the number needed to treatincorporating adjustments for the utility and timing of benefits and harms

    Med Decis Making

    (1993)
  • Cited by (70)

    • A review of the use of the number needed to treat to evaluate the efficacy of analgesics

      2015, Journal of Pain
      Citation Excerpt :

      Therefore, to calculate the NNT of an active drug in a trial, one needs to select 1) the derived outcome measure (eg, % responders); 2) the cutoff for the definition of response (eg, 30% reduction in pain); and 3) the time at which the observation of response/nonresponse is made (time point). Several articles have shown that the NNT is dependent on the outcome measured,11,28 the time point of measurement,6,11,44,45 and the cutoff value of response.31 Comparison of the NNT for total mortality, mortality, and all cardiovascular events in various statin trials showed that the NNT value was different for each of these outcome measures.11

    • What are effective strategies to communicate cardiovascular risk information to patients? A systematic review

      2011, Patient Education and Counseling
      Citation Excerpt :

      56 full papers were retrieved for further assessment and 29 were excluded. Of the 27 studies included for detailed data extraction, four were subsequently excluded because they only varied the degree of risk rather than the presentation or communication of risk formats [26–29]. Another five were excluded because the risk communication elements were embedded in a decision aid that had other components, and therefore, the outcomes could not be attributed to the risk communication strategy alone [30–34].

    View all citing articles on Scopus
    View full text