Twenty-five years ago, I recall sitting in a journal club in which the collective minds tore apart the then recently published (and still much quoted) meta-analysis by Elizabeth Carlsen, 'Evidence for decreasing quality of semen during past 50 years'. This BMJ paper made the headlines in 1992 because it suggested that between the 1930s and early 1990s, sperm counts across the world had almost halved. Since then, several other similar meta-analyses were published (e.g. Swan et al., (2000)) with each one broadly making the same conclusion and often leading to equally scary and sensationalistic media headlines. However, many of us in the sperm-research world were troubled by these meta-analyses and the media headlines that accompanied them.
A meta-analysis is a statistical approach which combines the data from multiple (usually small) studies to increase the statistical power over and above what might be achieved from each study individually. This sounds fine in theory, but it only really works when each of the individual studies have been conducted in a similar manner – here is where the problem lies. If all the sperm count studies being compared across 50 years were performed differently, then the meta-analyses just end up comparing apples with oranges, with the danger of concluding that the answer is in fact banana! It's 'garbage in – garbage out' in other words.
Critics of the various sperm count meta-analyses published to date (of which I am one) have pointed to the fact that they have often uncritically included studies of different populations of men that are simply not comparable. For example, comparing the sperm count studies of healthy volunteers taking part in a drug trial is not the same as comparing studies of sperm donors, (who usually by definition are recruited because of their higher sperm quality) or studies of men attending fertility clinics (who you would expect to have poorer sperm quality). Therefore, if over time the mix of studies has changed (eg. more studies of men from fertility clinics in recent years), it might be concluded by the meta-analysis that sperm counts have declined, whereas in fact the conclusion just reflects the changing fashion of research publications.
However, my major criticism of the meta-analyses published so far is that none of them have considered changes in laboratory methods. Counting sperm (and calculating sperm concentration) sounds easy, but we have significantly changed our methods of doing it over the past 50 years, to the extent that it is impossible to compare with any certainty our modern measurements with historic ones (see Pacey, 22052). We know that historic (less accurate) techniques tend to over-estimate the sperm concentration. So, if more modern studies included in these meta-analyses were increasingly using better techniques, then an apparent decline in sperm concentration over time in the meta-analysis might simply reflect improvements in laboratory technique and reductions in measurement error.
For these reasons and more, I have never been particularly convinced by the conclusions that sperm counts were declining. Extraordinary claims need extraordinary evidence in my book, and the two fatal flaws I have outlined always seemed too significant to ignore. However, when I read the most recent iteration of the sperm-count meta-analysis published by Levine et al., (2017), I must admit that I changed my mind a little. This is because that authors: (a) excluded from the meta-analysis those studies which had been published on men according to their fertility status (eg. attending a fertility clinic); and perhaps more importantly for me (b) only included studies which had counted sperm according to the gold standard technique of haemocytometry. Effectively it addressed my two major concerns head on.
So, I now find myself much less sceptical than I was about the idea that the sperm counts of men may have declined in recent years. Whether this warrants the crazy media headlines about possible human extinction seems doubtful. There is a case-history here about how the media report and write headlines about science stories. But it does for me raise another important question.
If we had been serious about knowing once and for all whether sperm counts had declined or not, we wouldn't have relied on meta-analyses to answer the question. It would have been better to design a large prospective cross-sectional study many years ago, when we first became worried about a potential problem. For example, we could have each year invited men who had been randomly selected from the electoral register to donate a semen sample to the project, and measure their sperm concentration using the same reliable technique and all our modern methods of laboratory quality control. I know I am not the only person to have proposed this idea, but sadly none of the proposals were considered worthy of funding. However, had they done so 25 years ago, we'd have known the answer by now once and for all, and we wouldn't be even having this debate.
So perhaps the story is less about whether sperm counts have declined or not in recent past, but rather that we need to fund research into male fertility more seriously - and stop burying our heads in the sand.