Intended for healthcare professionals


Commentary: Importance of empirical evidence

BMJ 1994; 309 doi: (Published 02 July 1994) Cite this as: BMJ 1994;309:22
  1. N Keiding,
  2. A Giwercman,
  3. E Carlsen,
  4. N E Skakkebaek
  1. Statistical Research Unit, University of Copenhagen, Panum Institute, DK- 2200 Copenhagen, Denmark University Department of Growth and Reproduction, Rigshospitalet 5064, DK-2100 Copenhagen.

    Bromwich et al point out that the distribution of sperm count is skewed to the right and that if a differential selection of skewed distributions is applied over the years this will bias the observed time trends. Both of these assertions are correct; indeed, in all 16 of the 61 publications cited in our original overview for which both median and mean were given the median was smaller than the mean, confirming the skewness.1

    Bromwich et al present some elaborate, although rather elementary, statistical points about skewed distributions and differential selection from these, but they fail to give any empirical reference that might support their suggestion of differential selection. One possibility is that they believe that the lower reference values for sperm counts of 60x106/ml in the 1940s and 20x106/ml at present had been used as truncation values for the reported distributions over the years. If Bromwich et al had actually studied the reports they would have found that there were plenty of values under these limits in even the oldest articles. The article by MacLeod and Gold in 1951, based on 1000 men, is particularly important in this respect.2 This early paper is largely responsible for the high historical values and is thus responsible for a considerable part of the observed decline. However, the authors of this paper were surprised about the low values contained in it. This paper was presumably the first to explicitly mention that it is “obvious to many that this figure [60x106/ml] is too high.”

    There are many problems with historical overviews (meta-analyses), but the article by Bromwich et al amounts to discussing time trends detached from the relevant empirical evidence. Thus the most cautious conclusion that can be drawn from the existing data is still that semen quality has declined significantly between 1940 and 1990. After several years of published evidence being ignored, the increasing incidence of abnormalities of male genital organs (including a highly significant increase in incidence of testicular cancer3) has finally attracted the attention of the scientific world. We hope that the paper of Bromwich et al, which is apparently based on wrong assumptions, will not bring confusion or divert attention from the urgent need for more research into the threat to male reproductive functions.4


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