Enhancing clinical decision-making: The imperative of reporting number needed to treat in anaesthesia research

INTRODUCTION The number needed to treat (NNT) or the number needed to harm (NNH) is a metric in clinical research that represents the number of patients who need to receive a specific intervention in comparison to a control group to achieve one additional beneficial outcome or to prevent one additional adverse event, respectively. The NNT provides a direct measure of an intervention’s absolute effectiveness.[1] While P values denote statistical significance, NNT offers a concrete, practical impact of a treatment comparison and assists clinicians in weighing the benefits against potential harms.[2] We systematically reviewed randomised controlled trials (RCTs) with binary outcomes recently published in the Indian Journal of Anaesthesia (IJA) in 2024. The rationale of this study was to underscore the current state of NNT and NNH reporting in IJA, which could influence the current clinical practice in anaesthesia in a more meaningful way. In addition, we planned to calculate the NNT in these trials to see if decisions based on NNT would impact or change the original study conclusions. METHODS All the original articles published in IJA in the year 2024 were screened. The RCTs with a binary primary outcome were included, whereas RCTs that did not have a binary outcome and observational studies were excluded. Each of the authors independently searched the IJA issues in 2024 and populated the results of the studies using primary outcomes, which were binary/categorical. The included studies and the calculated NNTs of all included studies were confirmed by all the authors, and disagreements between any two authors were resolved by the third author. The research question, results, study conclusion, and the frequency of NNT reporting were collected. The NNT and 95% confidence interval (CI) were calculated from each study result. The absolute risk reduction (ARR) was calculated as the difference in event proportions (Control - Treatment). The 95% CI for ARR was derived using the standard error of the difference in proportions, computed as √[p1 (1–p1)/n1 + p2 (1–p2)/n2], and by applying the normal approximation. The NNT was computed as the inverse of ARR, with CIs derived from the reciprocal bounds of the ARR CI. RESULTS Fourteen RCTs fulfilled the inclusion criteria [Table 1]. None of the studies explicitly reported NNT or NNH values.Table 1: Summary of findingsDISCUSSION Our analysis reveals a significant underreporting of NNT. Categorical or binary outcomes can be represented through various statistical measures, including relative risk (RR), odds ratios (OR), ARR/absolute risk difference, and NNT/NNH. Among these effect measures, the NNT offers a more intuitive understanding for clinicians, translating statistical data into actionable clinical insights. The magnitude of NNT is interpreted as an effect size – the lower the NNT, the more potent the treatment’s benefit. As a general rule of thumb, an NNT below 10 is considered a clinically meaningful or ‘large’ effect size.[17] Ideally, an NNT of 1 means that every patient treated would benefit compared to none in the control group, indicating the highest effectiveness possible for a treatment. Generally, the closer the NNT is to 1, the more efficacious the new treatment is relative to the comparator. Conversely, when the difference between the treatments is negligible, the NNT will be large, potentially approaching infinity, indicating equal response rates or no response in both groups. Although single-digit NNTs are typically preferred, higher NNTs may be deemed acceptable if the objective is to avert a serious outcome.[18,19] We will explore how NNT could potentially change the conclusions in some of the included studies. For example, in the study by Thakur et al.,[8] when using the between-the-fingers grip on the endotracheal tube during intubation, the ARR for the requirement of external laryngeal manipulation was only 6% (1.8%–10.7%). However, the NNT of 17 (9.4, 55.2) suggests that implementing this practice, which involves no additional cost in 17 cases, would reduce the need for a second assistant in at least one case. This finding does have value in resource-limited settings. However, based on P > 0.05 for the co-primary outcome of frequency of single-attempt intubation success, the study concludes that the between-the-fingers grip proved to be an equally good method to hold the endotracheal tube for orotracheal intubation as compared to the standard method. A more nuanced conclusion would have been that the between-the-fingers grip, while comparable to the standard method in terms of first-attempt intubation success, may reduce the need for additional manpower to perform external laryngeal manipulation when the primary airway manager has more than 10 years of anaesthesia experience.[8] The RR and OR can exaggerate the perceived benefit or harm if the baseline risk is low. For example, the RR of requiring external laryngeal manipulation (BURP) in the between-the-fingers group would be 0.11, while the ARR is only 6%.[8] However, the NNT of 17 reframes this in the context of additional manpower requirements to provide BURP while keeping intubation success rates the same. This highlights how the addition of absolute measures such as ARR and NNT can enhance the interpretability and real-world applicability of study findings beyond conventional P value thresholds. Such evaluations are crucial for healthcare resource allocation and policy-making. The NNT, or any effect measure, is not all-encompassing. An effect measure needs to be interpreted in the context of the interventions being compared and the clinical relevance of the outcome measure. This can be better explained by an example from the study by Elsakka et al.;[6] the NNT of 3 suggests that if three parturients were kept seated for 2 minutes following subarachnoid block (SAB), at least one of them would not experience hypotension. This appears to be an excellent NNT for a clinically important outcome. However, the failure of SAB in 4/46 (9%) patients (who were excluded from the final analysis) in the intervention (sitting) group versus 0/46 in the control group suggests that sitting for 2 minutes following SAB may not be a good choice.[6] In this case, the NNH for SAB failure would be 11 (5.9, 180.8), meaning if 11 parturients were kept seated for 2 minutes following SAB, one of them would experience a failed spinal, which may necessitate conversion to a general anaesthetic. Thus, if both NNT and NNH are calculated, the notable incidence of failed spinal will not be missed in the conclusion of the study.[6] In the study by Misra et al.,[15] the NNT is 2 (1.5, 3.1) indicating that bronchoscopy would position the central venous catheter (CVC) at the ideal position where the distal tip is considered to be near the superior vena cava and right atrium junction, specifically 2 cm above and 1 cm distal to this junction. However, the invasiveness, availability, cost, expertise, and time taken in performing bronchoscopy to position the CVC at this junction probably do not justify its use in all patients requiring a CVC, despite the small NNT and P value. The above examples illustrate that acceptable NNT values become lower as a treatment’s risk, invasiveness, or cost increases – the intervention must help enough people justify exposing everyone to its downsides. It is feasible to calculate NNT and NNH in all studies using binary outcomes for treatment comparison. The methods to calculate NNT from summary measures and time-to-event outcomes with varying follow-up periods have been discussed elsewhere.[2,20–22] CONCLUSION The absence of reporting the NNT and NNH in studies constitutes a missed opportunity to enhance the clinical applicability of research findings. It is essential to provide baseline risks, clearly defined outcomes, time horizons, and CIs. Presenting an NNT in isolation, without its contextual framework, renders it ambiguous and less useful for decision-making. By adopting standardised reporting of these metrics, researchers can provide clinicians and policymakers with more precise, actionable data, ultimately improving patient care and resource utilisation. Author contribution details VG: Design, literature search, data collection, compilation of results, manuscript preparation KAG: Data collection, compilation of results, manuscript preparation NS: Concept, literature search, data collection, manuscript preparation. Financial support and sponsorship Nil. Conflicts of interest There are no conflicts of interest.