For healthcare professionals only

Letters

Follow up care in general practice of patients with myocardial infarction and angina

BMJ 1999; 319 doi: https://doi.org/10.1136/bmj.319.7206.380 (Published 07 August 1999) Cite this as: BMJ 1999;319:380

Trial was underpowered

  1. John Robson, senior lecturer (j.robson{at}qmw.ac.uk)
  1. Department of General Practice and Primary Care, Queen Mary and Westfield College, London E1 4NS
  2. University of Oxford, Institute of Health Sciences, Oxford OX3 7LF

    EDITOR—The paper reporting the trial to improve follow up in general practice of patients with coronary heart disease after discharge from hospital is unduly pessimistic.1 At least one of the authors (Mant) has been in this situation before.

    Mant worked on the OXCHECK study, which found comparable differences of less than 5% between control and intervention groups and in which a strategy identifying patients at high risk was abandoned in favour of targeting those with established disease at yet higher risk.2 It was left to Field et al to show that the OXCHECK study was successful and that the incremental addition of assessment of multiple risk factors to a pre-existing programme targeting smoking and blood pressure was a cost effective way of improving risk factors.3 4

    The failure to show substantial differences in outcomes in both Jolly et al's trial and the OXCHECK study has resulted from flaws in the expectations and design of both trials. The trials are based on a prior assumption of the size of the effect of the intervention that is greater than that repeatedly found in similar trials aiming to improve professional organisation and behaviour—pragmatic trials testing well organised delivery versus less well organised delivery.

    It is unlikely that any trial would be feasible to show improvements in aspirin prophylaxis or outcomes among practices already reaching 80% of the target population. Both trials were underpowered to detect differences in physiological outcomes of 1-5%, which might more realistically have been predicted to result from improvements in prescribing or advice of the order of 2-6%.1 2

    What are the minimum outcome differences considered cost effective, and how large would the trial need to be and for how long would it need to be sustained to show these? These assumptions need to be considered within an incremental framework, as doing nothing is not a realistic option. Will these underpowered trials persuade anyone to invest in larger, better designed studies, or will we go on running trials designed to fail because of unrealistic prior assumptions?

    Improvements in professional organisation yield marginal though important benefits. Given the effort and commitment required to mount even a small study of this kind, time and pretrial modelling might usefully be spent in clarifying the smallest difference worth investing in and the feasibility of mounting a trial sufficiently large to test these small but important differences. Jolly et al's study confirms that improvements in delivery do not in the short term yield large changes in risk factors but do consistently yield small, important, improvements—a conclusion that the multiple risk factor intervention trial reached 20 years ago.5

    References

    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.

    Author's reply

    1. David Mant, professor of general practice (david.mant{at}public-health.oxford.ac.uk)
    1. Department of General Practice and Primary Care, Queen Mary and Westfield College, London E1 4NS
    2. University of Oxford, Institute of Health Sciences, Oxford OX3 7LF

      EDITOR—When colleagues and I designed the OXCHECK trial we had already confirmed that brief advice in general practice could increase smoking cessation rates.1 Our studies of dietary advice were less consistent but seemed to show that a well structured and supported intervention could achieve a reduction in serum cholesterol concentration of up to 5% in selected patients.2 Our interpretation of the results of the OXCHECK trial was in the context of the greater effectiveness of these interventions regarding single risk factors. It was also in the context of our experience that the likelihood of an invitation to attend being accepted was inversely related to cardiovascular risk.3

      Contrary to Robson's assertion, the OXCHECK trial was not underpowered; the difference in serum cholesterol of 2.3% was significant, and the precision of the estimate of the difference in smoking cessation rates was high (95% confidence interval −3.5 to 1.7%). We concluded that the public health importance of the difference in cholesterol concentration depended on the time for which it was sustained. Again, this reflected existing knowledge that the additional benefit of individual advice on lifestyle over community education is lost over a relatively short period.4

      The situation in the SHIP trial was in some ways similar. We started with the knowledge that both exercise rehabilitation and appropriate drug treatment can substantially reduce morbidity and mortality. We also had evidence that these interventions were not being implemented effectively in everyday practice in the trial area.

      The SHIP trial had less power than the OXCHECK study to detect physiological change (95% confidence interval for difference in total cholesterol −0.33 to 0.06 mmol/l), but the anticipated effect included increasing attendance for exercise rehabilitation and adherence to drug treatment—both of which are associated with a much larger effect size than lifestyle interventions Moreover, we incorporated a qualitative study within the SHIP trial's design and measured intermediate process outcome measures so that we could gain information to improve the quality of NHS care whatever the physiological outcome.5

      The cost effectiveness of care needs to be measured against alternative ways of providing that care and must take account of opportunity costs Showing absolute marginal benefit is not enough. In 1990 we concluded that doctors should be careful not to absolve the government of its public health obligations by substituting unproved preventive interventions aimed at individual patients.3 I still believe that this is true.

      References

      1. 1.
      2. 2.
      3. 3.
      4. 4.
      5. 5.
      View Abstract

      Log in

      Log in through your institution

      Subscribe

      * For online subscription