Using clinical evidence

BMJ 2001; 323 doi: https://doi.org/10.1136/bmj.323.7305.165 (Published 21 July 2001) Cite this as: BMJ 2001;323:165

Randomised controlled trials are not the only evidence

  1. Martin S Knapp (mknapp@optushome.com.au), locum consultant physician
  1. 12 Townsend Street, Ivanhoe, 3079, Australia
  2. Department of Public Health, Northumberland Health Authority, Morpeth, Northumberland NE61 2DL
  3. Department of Public Health, Gateshead and South Tyneside Health Authority, South Shields, Tyne and Wear NE33 3DP
  4. Maternity Unit, Chelsea and Westminster Hospital, London SW10 9NH
  5. BMJ Publishing Group, London WC1H 9JR

    EDITOR—Randomised controlled trials—the focus of evidence based medicine—will be given more momentum by the availability of Clinical Evidence on line.1 The criteria to be used to select evidence for this journal imply that only recommendations from randomised controlled trials can be used if medicine is to be considered evidence based.

    I urge the BMJ and Clinical Evidence to take a less conservative position on what can be regarded as clinical evidence. The exclusion from publication of anything other than randomised controlled trials (except where these do not exist on a particular topic) will dissuade clinicians and research workers from considering quantitative statistical analysis of sequential data from individuals as an alternative but meaningful form of clinical evidence. Statistical analyses of sequences of numerical data from individuals can provide evidence that is as robust and as statistically valid as data from the best randomised controlled trials.

    The probability of a real effect from an intervention on an individual may be high when there are sufficient data points for analysis before and after the intervention. Interventions used may be a drug, a placebo, a physical treatment (for example, physiotherapy), the introduction of a “natural” therapy, or any other intervention when the effect can be determined by its impact on relevant numerical variables that are collected sequentially. The variables may be laboratory measurements, clinical measurements, or patient assessments (for example, pain scores). Reports that contain statistically valid calculations of change as a result of an intervention in an individual or in a group should have as much right to inclusion in Clinical Evidence as evidence from randomised controlled trials.

    The statistical analysis of …

    View Full Text

    Log in

    Log in through your institution


    * For online subscription