TV Ad: Our drug cures 50% of patients!
When you unpack the data, the commercial’s drug showed a 2% success instead of 1%. Clinically, percents and statistics are often misleading and frustrating—sounding like miracle pills, they can be inflated by drug sellers and advertised in a newspaper or TV. The next day, patients show up at your front door asking for the drug their favorite actor took, and you know it showed poor consistency and low efficacy. But the healthcare industry continues to use statistics like relative risk reduction, which are heavily influenced and easily to manipulate.
The Medicare Access and CHIP Reauthorization Act (MACRA), effective beginning in 2017, tethers Medicare reimbursement to the efficient resource utilization of eligible providers. With this policy on the horizon, can current methods for comparing treatment options withstand the shift towards value-based care? Now more than ever, both physicians and administrators alike need a robust tool which can weigh multiple therapeutic options by representing treatment costs and efficacy in a single value. It needs to address how resources can be optimized while simultaneously allowing for comparison of treatment options. Finally, that metric should be intuitive enough to mean something to a patient. Number needed to treat (NNT) addresses each of these criteria.
NNT measures the number of patients who need to be administered a given intervention to avoid one adverse outcome.[2,3] NNT is calculated by taking the reciprocal of the absolute risk reduction (ARR) when comparing two therapies[2,3]. For example, if a new medication reduces the chance of death by 25% compared to an old medication that reduces it by 5%, the ARR is the difference between the two, or 20%. The NNT would be 1/0.2, or 5, meaning 1 death would be avoided for every 5 patients given the new medication. NNT can easily be configured to determine health care costs and resource allocation through its application. If that novel therapy, with an NNT of 5, costs $20 to administer, its effective cost to prevent 1 death is $100. The same idea can be extrapolated to compare a new treatment’s efficacy against hospitalization, surgery, or other adverse outcomes. Thus, NNT represents both a treatment’s efficacy and its relative cost against an adverse outcome.
Researchers have pointed out a few shortcomings of the NNT calculation.[2-4] The final result of an NNT calculation, for example, does not include a time interval for the treatment to take place. Therefore, the time interval for the treatment option needs to be reported clearly and considered when weighing treatment options. NNT, by definition, is binary in nature: it asks how many of A is needed to prevent B. Thus, NNTs of separate studies can be compared only when their two components are similar, such as hospitalized/non-hospitalized data for a beta blocker study and the same categories tested with ACE inhibitors. Lastly, since the NNT statistic compares two groups, it is affected by the baseline risk of a population, regardless of the etiology of the baseline risk. A higher baseline risk for an adverse outcome may produce a smaller ARR value and thereby an increased NNT. If each of these variables are considered when screening therapies, NNT quickly becomes a very powerful comparison tool.
What is a good NNT figure? Surely a time-proven drug like Aspirin would have good numbers. Aspirin’s NNT is 1667, assumed to be for one year. One thousand six hundred and sixty-seven patients would need to be given a daily Aspirin to prevent one adverse event (here, a cardiovascular problem). Moreover, 1 in 3333 of the patients had a major bleeding event—1 harm for every 2 benefits. Nevertheless, Aspirin remains a standard of care for patients with heart disease.
NNT should be used to judge the capability of any platform, even telehealth. Epharmix, through automated text messaging, has shown NNT values across multiple platforms in the single digits. With such outstanding NNT values, Epharmix suggests that well-designed systems, by taking advantage of widespread technology, can simultaneously improve clinical outcomes, optimize clinical efficiency, and increase patient satisfaction. Over 20 text message and phone-based interventions have improved the care of patients with COPD, depression, diabetes, and more. Even better, there are no contraindications complicating text messages.
With MACRA coming next year and more changes on the way, we cannot afford to waste valuable time using incompetent metrics. We need to move towards NNT as the metric by which we compare interventions. We must be able to identify and recommend robust interventions like Epharmix suites to positively affect the health of our patients.
- Haycock C, Edwards ML, Stanley CS. Unpacking MACRA: The Proposed Rule and Its Implications for Payment and Practice. Nurs Adm Q. Oct-Dec 2016; 40(4): 349-55.
- Suissa S. The Number Needed to Treat: 25 Years of Trials and Tribulations in Clinical Research. Rambam Maimonides Med J. Jul 2015; 6(3): e0033.
- McAlister FA. The “number needed to treat” turns 20 — and continue to be used and misused. CMAJ. Sep 9 2008; 179(6): 549-553.
- Stang A, Poole C, Bender R. Common problems related to the use of number needed to treat. J Clin Epidemiol. Aug 2010; 63(8): 820-5.
- Newman D. Aspirin to Prevent a First Heart Attack or Stroke. The NNT Group. Jan 8 2015. Accessed Nov 21 2016 from http://www.thennt.com/nnt/aspirin-to-prevent-a-first-heart-attack-or-stroke/.
This article is written by guest author Jacob Adney, a first year medical student at Saint Louis University Medical School and an independent researcher at the EPX Research Center @WUSTL.