Endgames Statistical Question

What is per protocol analysis?

BMJ 2013; 346 doi: https://doi.org/10.1136/bmj.f3748 (Published 14 June 2013) Cite this as: BMJ 2013;346:f3748
  1. Philip Sedgwick, reader in medical statistics and medical education
  1. 1Centre for Medical and Healthcare Education, St George’s, University of London, Tooting, London, UK
  1. p.sedgwick{at}sgul.ac.uk

Researchers evaluated the efficacy of 4% dimeticone lotion for the treatment of head louse infestation. A randomised controlled trial was performed. Control treatment was 0.5% phenothrin liquid, the standard treatment. Treatments were applied twice, seven days apart, with dimeticone lotion for eight hours or overnight and phenothrin liquid for 12 hours or overnight. One of the main outcome measures was cure of infestation after the second application, regardless of whether reinfestation occurred later.1

Participants were 214 young people (aged 4-18 years) and 39 adults with active head louse infestation; 127 were allocated to 4% dimeticone lotion and 125 to 0.5% phenothrin liquid. Per protocol analysis indicated that a greater proportion of patients were cured with phenothrin—84 of 121 (69%) participants were cured with dimeticone versus 90 of 116 (78%) with phenothrin, a difference of −8% (95% confidence interval −19% to 3%).

Which of the following, if any, would result in a trial participant being included in the per protocol analysis?

  • a) Withdrew from trial after informed consent and treatment allocation, but before starting treatment

  • b) Withdrew from trial having started the allocated treatment

  • c) Started the allocated treatment, but did not adhere to treatment protocol

  • d) Completed the protocol for the treatment originally allocated

  • e) Changed to alternative treatment to one originally allocated and adhered to treatment protocol


Answer d would result in a trial participant being included in the per protocol analysis, whereas a, b, c, and e would not.

The effectiveness of 4% dimeticone lotion compared with 0.5% phenothrin liquid in treating head louse infestation was investigated using a randomised controlled trial. A per protocol analysis found that a greater, but not significant, proportion of patients were cured with penothrin. The per protocol analysis, sometimes known as an “on treatment” analysis, included only those participants who complied with the trial protocol. Specifically, it included only those who adhered to the treatment protocol that they were allocated to and provided sufficient data for assessment of the outcome measures. For both treatments the protocol was two applications, seven days apart, with dimeticone lotion for eight hours or overnight and phenothrin liquid for 12 hours or overnight.

By including only those participants who complied with the trial protocol, the per protocol analysis reflected the maximum potential benefits in the treatment comparison for head louse infestation. However, the per protocol analysis may be biased. Because participants were excluded if they did not complete the protocol of the allocated treatment, the original comparability of the treatment groups in their baseline characteristics achieved after randomisation may not have been maintained. This would have resulted in confounding. Confounding factors are those that influence treatment and outcome measures. They include demographics, prognostic factors, and other characteristics that influence someone to participate in, or withdraw from, a trial. Therefore, any differences between treatment groups at the end of the study may not have been due to differences in treatment received but a result of differences between treatment groups in their baseline characteristics.

The results of a per protocol analysis are typically compared with those of an intention to treat approach. Described in a previous question,2 intention to treat analysis compares two groups of participants based on the treatment originally allocated. This is regardless of whether the patient began the treatment allocated, subsequently withdrew from the trial, did not adhere to the protocol of the allocated treatment, or received a different treatment from that originally allocated. In the above trial, an intention to treat analysis indicated that 89 of 127 (70%) participants treated with dimeticone were cured compared with 94 of 125 (75%) treated with phenothrin (difference −5%, 95% confidence interval −16% to 6%). Intention to treat analysis provided a pragmatic comparison of the treatments for head louse infestation; it resembled clinical practice, where patients may not adhere to treatment protocols.

By comparing the results from the intention to treat and per protocol analyses, the potential bias of confounding in the estimate of treatment effectiveness indicated by per protocol analysis can be estimated. The results of the two analyses were similar, however, which suggests that non-adherence to treatment was limited, and that in clinical practice patients would adhere to the treatment protocol of their care programme.


Cite this as: BMJ 2013;346:f3748


  • Competing interests: None declared.


View Abstract

Log in

Log in through your institution


* For online subscription