Number needed to treat: A descriptor for weighing therapeutic options

Lonnie Wen, Robert Badgett, John Cornell

Research output: Contribution to journalReview articlepeer-review

41 Scopus citations


Purpose. The utility and limitations of the number needed to treat (NNT) art described and compared with those of alternative descriptors of therapeutic yield. Summary. Pharmacists, physicians, and patients are more likely to launch interventions based on relative measures of therapeutic yield. The NNT is a simple way to express the absolute therapeutic yield of an intervention. It is essential to take baseline risk and length of follow-up into account when calculating and interpreting absolute or relative measures of therapeutic yield. Patients may have difficulty understanding what the NNT tells them about the benefits and harms associated with an intervention, and confidence intervals for the NNT can be difficult to interpret. Relative measures of benefits are advantageous for comparing benefits across populations with different risks. The use of absolute measures to describe therapeutic options leads to more conservative decision-making and is potentially more useful. Among the absolute measures, the absolute risk reduction may be easier to understand than the NNT. These measures art most helpful when the best options are not clear or when patient input is needed. Personal digital assistants and Web-based tools can assist with calculations. Conclusion. The NNT is a useful measure but is influenced by numerous variables and may not be easily understood by patients.

Original languageEnglish (US)
Pages (from-to)2031-2036
Number of pages6
JournalAmerican Journal of Health-System Pharmacy
Issue number19
StatePublished - Oct 1 2005


  • Calculations
  • Computers
  • Decision-making
  • Methodology
  • Technology
  • Toxicity

ASJC Scopus subject areas

  • Pharmacology
  • Health Policy


Dive into the research topics of 'Number needed to treat: A descriptor for weighing therapeutic options'. Together they form a unique fingerprint.

Cite this