General Practice

Commentary: Replication of results

BMJ 1996; 312 doi: (Published 29 June 1996) Cite this as: BMJ 1996;312:1653
  1. M D Buckley-Sharp, consultant in chemical pathologya,
  2. D A Gardner, consultant in chemical pathologya
  1. a Department of Chemical Pathology, University College London Hospitals, London W1P 6DB

    The phenomenon of high ambient temperature causing a factitious hypokalaemia is not described in the usual textbooks and seems to be largely unknown. If it is so easily demonstrated, even in a temperate climate, then it should be equally easily repeatable.

    We recovered all plasma potassium results from 1 January to 30 September 1995, divided them into two groups according to their origin (general practice and other), and calculated the mean and standard deviation for each group for each day. There were no results from general practice for Saturdays, Sundays, or holidays, and samples with these dates were removed for both groups. Over the 189 working days, there remained 6191 results from general practice and 81 496 results from other sources. After reviewing the data, we thought that the clearest presentation would be given by graphing the five day rolling average of the daily means, and these are shown in figure 1.

    Fig 1
    Fig 1

    Daily mean potassium results for GP (*) and other sources (*) samples, omitting Saturdays, Sundays, and holidays, shown as five day rolling averages, 1 January to 30 September 1995

    The important features of these results are all in agreement with Masters et al. Results from general practice were usually higher than results from other sources. From late June there was a steady fall in the mean of results from general practice. By late July, mean results from general practice were lower than mean results from other sources. As our data series is longer than that of Masters et al, we observed a return towards the more usual comparison by late September. The difference in spring, and the amount of the summer reduction in results from general practice, are both quantitatively similar to the results of Masters et al. Our results were obtained with three separate ISP1000 analysers, and the changes cannot be explained by standardisation or quality variation.

    It is supposed that the lowered plasma potassium concentration is caused by high temperature incubation of the samples, especially in a closed van on a hot day. Some of our samples in the general practice group did not arrive by van, while some samples from other sources did. This may explain some correlation between the two data series—as seen in August, for example.

    View Abstract