Irritable bowel syndrome, or IBS, is an uncomfortable and sometimes painful disorder affecting up to 1 in 5 persons. Effective treatment options have been somewhat limited, but researchers recently reported that two prevalent drug therapies — rifaximin and lubiprostone — may be preferred options, depending on the type of IBS. Also important, a unique but simple statistical approach is presented for calculating benefit-to-harm ratios for any therapy.
Patients with IBS often experience abdominal pain or cramps, excess gas or bloating, and visible abdominal distension. The disorder is usually classified as IBS with constipation or IBS with diarrhea, and the treatment choices for each type differ. However, many drug therapies carry risks of causing troubling side effects, including nausea, insomnia, palpitations, and decreased appetite. Some of those effects, or harms, are bothersome enough to cause treatment discontinuation.
A new research report — published online by the American Journal of Medicine, and set to appear in the April print edition — examines trials of subjects with IBS with diarrhea or with constipation who were receiving a drug intervention deemed of merit by the American College of Gastroenterology (ACG) task force [Shah et al. 2012, ref below]. The therapies are compared using a systematic review and meta-analysis approach for determining the NNH (number-needed-to-harm) and, when possible, the NNT (number-needed-to-treat) for each therapy.
NNT and NNH were previously discussed in Part 6 of our ongoing series, “Making Sense of Pain Research” [here], and this present study serves as a further example. Essentially, these two statistics address the question, “How many patients need to receive a treatment to either achieve one additional desired outcome (NNT, eg, pain relief) or for one additional undesired or harmful outcome to occur (NNH, eg, a side effect)? The smaller the NNT and the larger the NNH the better.
Each of these statistics, NNT or NNH, is calculated as the reciprocal of the absolute risk difference (ARD) between groups for the particular measure, or 1/ARD. For example, if the difference between a treatment and a control group on some measure of benefit or harm is 20%, the respective NNT or NNH would be 1/.20 or 5. As we shall also see, below, these two statistics can be used to calculate a unique benefit-to-harm ratio, suggesting the number of patients who might benefit from a treatment for each patient experiencing a significant adverse event or discontinuing treatment.
For their study, Shah and colleagues identified studies of IBS through an extensive literature search covering 1950 to 2011. A primary interest was studies reporting data for calculating the absolute risk difference (ARD) of patients experiencing adverse events requiring discontinuation of treatment, which was used to determine the NNH (number-needed-to-harm). Individual adverse events (side effects), themselves, were examined separately. In total, 26 clinical trials were included in the analysis:
Two groups of trials examined treatments for IBS with constipation…
> 4 assessing selective serotonin reuptake inhibitors (SSRIs),
> 3 examining lubiprostone, a type of laxative.
Three groups of trials assessed therapies for IBS with diarrhea…
> 6 evaluating tricyclic antidepressants for IBS with diarrhea,
> 8 with alosetron, a 5-HT3 serotonin receptor antagonist,
> 5 assessing rifaximin, a semisynthetic, non-systemic antibiotic.
The results indicated that there were insufficient data on SSRIs for a reliable meta-analysis to determine the NNH for study discontinuations, but these agents seemed to be relatively safe in terms of adverse event frequency. Similarly, lubiprostone was relatively safe with nonsignificant harms in a combined phase III trial.
More rigorous data were available for tricyclic antidepressants, alosetron, and rifaximin; the NNHs for discontinuation of therapy were 18.3, 19.4, and 8,971, respectively, and the respective NNTs for therapeutic response were 8, 7.5, and 10.6. The benefit-to-harm ratio — calculated as NNH/NNT, or the number of patients benefitting for each patient who discontinued therapy — for tricyclic antidepressants was 2.3, and for alosetron it was 2.6. However, for rifaximin, this ratio was a remarkable 846 patients benefitting for each one who discontinued therapy. Additionally, various individual side effects of therapy were more common with tricyclic antidepressants and alosetron.
The researchers conclude that, in IBS with diarrhea, tricyclic antidepressants and alosetron are associated with an unfavorable NNH compared with rifaximin, and rifaximin appears to be clearly the most advantageous agent. Likewise, serious individual adverse events are notable using tricyclic antidepressants and alosetron but not with rifaximin. For IBS with constipation, apart from lubiprostone, data on treatment with other agents are limited to small trials (with poor descriptions of side effects); on this basis, lubiprostone might be preferred, but both this agent and SSRIs appear to have favorable safety profiles.
COMMENTARY: Prior meta-analyses of studies investigating therapies for IBS had focused primarily on the number-needed-to-treat (NNT), whereas, Shah and colleagues also examined the number-needed-to-harm (NNH). This metric — in their study, portraying the relative risk of discontinuing treatment due to adverse effects — can be important to know with therapy for a chronic condition such as IBS that may require long-term therapy.
A most interesting feature of this study is its unique application of a benefit-to-harm ratio, calculated as NNH/NNT, used for comparing therapies. The data table for this, adapted from Shah et al., is at left.
As noted above, the ratio answers a vital clinical question for practitioners and patients, “How many patients might benefit from a therapy for each patient who experiences a harmful effect?” Of the 3 agents, rifaximin is the clear winner, with only 1 discontinuation of the drug due to adverse effects for every 846 patients who benefit. However, if only the NNTs were considered, rifaximin would not appear to be the most favorable among the 3 drugs.
In applying these principles to any study, the NNT needs to be defined as an outcome of significant benefit, whether it is pain relief, a measure of functionality or quality of life, or other. Similarly, the NNH might measure therapeutic discontinuation, as a surrogate measure of combined adverse effects, or individual side effects deemed to be of importance for patients, such as nausea, somnolence, or other adverse event (harm). The NNH/NNT ratio serves as an easy-to-calculate way of assessing the relative benefits vs harms of different therapies — provided that sufficient data are available for determining these measures.
With better reporting of research outcomes in the pain management field, this sort of approach might be applied to a variety of therapies. However, while NNT and NNH are clinically useful and intuitive measures of effect size — allowing assessments of both clinical efficacy and tolerability — there can be limitations, for example: [Citrome 2007 and 2008, refs below]
- NNT and NNH metrics should only be calculated from well-designed and well-conducted trials that enroll subjects similar to patients seen in everyday practice, and the doses of medications examined should be typical of those used clinically.
- If outcome results are drawn from different studies, without direct head-to-head comparisons of the treatments in question, event rates (and their respective NNTs or NNHs) are dependent on the particular study populations. The study populations must be comparable across studies or there may be significant differences (aside from drug effects) that could account for observed outcomes.
- Report authors should always provide 95% confidence intervals for NNT and/or NNH metrics as a measure of their precision and significance. (Unfortunately, Shah et al. do not provide these data in their study report.)
- Calculation of NNT or NNH is limited to binary or dichotomous (yes/no) types of outcome data representing mean (average) responses of the subjects. For example, either patients achieve a certain level of pain relief on average or they do not. Continuous data representing a broad range of individual responses cannot be portrayed as an NNT, unless such data are converted to a yes/no type of outcome (eg, pain scores below a certain point are denoted as favorable response and the rest as unfavorable). The same priniciples apply to data representing harms for calculating NNH.
- A benefit-to-harm ratio (NNH/NNT) much greater than 1.0 is obviously preferred, since it denotes that many patients benefit for each one harmed in some way. At a ratio of 1.0 there is an equal trade-off between benefits and harms/risks. Less than 1.0 would indicate that harm exceeds benefit.
However, it must be considered what the ratios represent. For example, a drug might have an unfavorably small ratio, say 5.0, when it comes to dry mouth, but this might be viewed as somewhat inconsequential since it is a minor side effect; conversely, even a fairly large benefit-to-harm ratio for a serious adverse effect, such as 50.0 for heart arrhythmia, might still be of great consequence. Note: this ratio also has been referred to recently as “Likelihood to be Helped or Harmed,” or LHH [see Citrome and Weiss-Citrome 2012, ref below].
- Interpreting the clinical relevance of an NNT, NNH, or NNH/NNT can be somewhat subjective and may be most helpful when comparing those metrics across multiple therapies for a disorder. This is clearly portrayed in the Shah et al. study, in which alosetron and tricyclic antidepressant appear comparable in terms of their benefit-to-harm ratios, but rifaximin is clearly superior on that measure.
Despite such limitations NNT, NNH, and NNH/NNT might be very useful for selecting therapies; however, this is not always possible with current research in pain management. For example, Shah and colleagues found that this sort of analytical approach was impractical in the case of treatments for IBS with constipation.
The pooled number of patients enrolled in SSRI studies was less than 200, so the lack of statistical power was problematic, and there also was a general paucity of adverse event reporting. The NNT for SSRIs appeared to be favorable, at 2.9, but the researchers say this number should be cautiously accepted as valid due to limitations of the data. For lubiprostone more ample adverse event data were available from large-scale trials and this agent did not seem to result in any excessive study dropouts compared with placebo. However, a valid NNT could not be calculated since data available for this purpose were inadequate.
> Citrome L. Compelling or irrelevant? Using number needed to treat can help decide. Acta Psychiatr Scand. 2008;117(6):412-419 [article here].
> Citrome L. Show me the evidence: Using number needed to treat. Southern Med J. 2007;100(9):881-884 [article here].
> Citrome L, Weiss-Citrome A. A Systematic Review of Duloxetine for Osteoarthritic Pain: What is the Number Needed to Treat, Number Needed to Harm, and Likelihood to be Helped or Harmed? Postgrad Med. 2012(Jan);124(1):83-93 [access here].
> Shah E, Kim S, Chong K, et al. .Evaluation of Harm in the Pharmacotherapy of Irritable Bowel Syndrome. Am J Med. 2012(Apr);125(4):381-393 [abstract here].
Don’t Miss Out. Stay Up-to-Date on Pain-Topics UPDATES!
Register [here] to receive a once-weekly e-Notification of new postings.