Comparison of sperm counts in two groups of men presenting for infertility investigations 20 years apart

Abstract

Aims:  To compare sperm counts for two groups of men who had presented for infertility investigations approximately 20 years apart.

Methods:  The study compared results for 309 men tested between 1977 and 1981 with those of 559 men tested between 1997 and 1998 using identical methodology. In order to approximate the normal population, only those men with counts above 5 million/mL were included in the final analysis. Bias, due to repeated testing after an initial abnormal result, was minimized by including only the patient's first test results. In addition, to allow for time‐dependent changes in the requirements for semen samples, results were included only if a complete sample was produced by masturbation after 3–5 days abstinence.

Results:  There was a small, but statistically significant drop in ejaculate volume (3.9–3.6 mL, P = 0.015) and a significant increase in the patient's mean age (32.18 vs 35.08, P < 0.001). Both groups had median abstinence of 3 days and no difference in sperm counts with a mean (median) count for the early group of 87.9 (75) versus 92.0 (76) for the recent group (P > 0.80). The significant drop in ejaculate volume was not reflected in a difference (P = 0.45) in total sperm numbers in the ejaculate with 320.7 (255) versus 313.1 (234).

Conclusion:  This study found no evidence of a decrease in sperm counts or total sperm output in men (excluding those with severe oligospermia) presenting for infertility investigations in Melbourne, Australia, over the last two decades of the twentieth century. (Reprod Med Biol 2004; 3: 211–216)

INTRODUCTION

THERE HAS BEEN a significant degree of controversy surrounding the hypothesis that human sperm counts may have been declining. ¹ Geographical variation in environmental factors, bias from self‐selected study groups and uncontrolled differences in sperm counting methodology, may all have contributed to the uncertainty. One study performed in Sydney, Australia, found no evidence of decreased sperm counts in men presenting as sperm donors or contraceptive trial volunteers between the years 1980 to 1995. ² The latter study also concluded that there was a very large magnitude of bias (estimated at up to 100%) when studying sperm counts on self‐selected groups of men such as sperm donors and contraceptive trial volunteers. It concluded that none of the studies published to that time would have been representative of the general population. ² It also argued strongly against the use of patient data because of the inherent bias contributed by male factor patients having reduced sperm output. ² Therefore, it is unlikely that any group of volunteers, including men whose partners are pregnant at the time of testing, will constitute a random sample of the general population. However, as pointed out by Jegou et al., ¹ it is more important that the study samples are as rigorously defined as possible and equivalent in terms of selection criteria and test performance, because it is not necessary to have samples which are fully representative of the general population in order to detect evidence of either change or stability over time or at different locations. ¹ Because semen results on ‘normal’ men are relatively scarce, whereas those on men undergoing infertility investigations are plentiful, it is worth making an effort to develop protocols for extracting useful data from patient databases. In order to develop criteria for doing so, it is necessary to take note of results from well controlled studies which have indicated that there is no significant difference in the distribution of sperm counts between groups of proven fertile men and those having infertility investigations, until counts of <5–10 million/mL are reached. ³ , ⁴ , ⁵ Therefore, the general statistical characteristics of groups of men undergoing infertility investigations, after exclusion of those men with counts ≤5 million/mL, should approximate those of the general population. The process of extracting putatively normal results from patient data in order to develop normal ranges is a valid approach. ⁶ , ⁷ Thus, the majority of men in a cohort having infertility investigations will have sperm counts equivalent to men in the general population, but have other reasons for their infertility/subfertility, including major female factors in approximately 50% of cases, low frequency of intercourse or impairment of sperm function associated with low motility, low sperm velocity, sperm auto‐immunity or an intrinsic defect in sperm fertilizing ability. Consequently, it is arguable that patient data extracted in this way from infertility‐related testing performed in specialist laboratories with tight quality control can be used to monitor changes in semen quality.

Having been in charge of the Andrology Laboratory at The Royal Women's Hospital since 1977, the first author has had the relatively unique opportunity of comparing sperm counts performed with exactly the same closely supervised methodology, on two selected groups of men presenting for infertility investigations, approximately 20 years apart. The rationale for the analysis was that if sperm counts in the general population have been falling, then this would be reflected in men presenting for infertility investigations. In the present study, a significant source of potential bias has been eliminated by including only the patient's first test result. Thus, male‐factor cases that would have been more likely to be quickly referred back to the laboratory for repeat testing were not over represented in the database.

MATERIALS AND METHODS

Sperm counting

IN BOTH THE 1977–1981 and the 1997–1998 study windows, the sperm concentration was determined using Improved Neubauer Haemocytometers (American Optical Corporation, Buffalo, NY, USA) of 0.1 mm depth. Initial examination of the semen sample was performed on a 7–10 µL drop of undiluted, fully liquefied semen evenly spread under a 22 × 22 mm coverslip and bright‐field or phase‐contrast microscopy. During this examination, the scientist estimated the semen dilution required for subsequent counting in the hemocytometer, so that a minimum of 100 sperm were counted (in the majority of counts, at least 150–200 sperm were counted). Therefore, the semen dilutions ranged from 1 in 2 for an extremely low count, to 1 in 40 for a high count. The dilutions were prepared in tap‐water using 0.2 mL of semen in order to minimize sampling error. Semen samples with increased viscosity or sperm agglutination were gently but repeatedly syringed through a 19 gauge needle to obtain even sperm distribution before dilutions were prepared. Any syringing was performed after determination of motility and immediately before making the dilution for sperm counting.

During both study periods, the laboratory was closely supervised by the first author, with careful attention paid to training of new staff and ongoing monitoring of fully trained staff. In particular, the procedure for semen dilution and hemocytometer chamber set‐up was identical during the two periods under comparison in this retrospective analysis. The two study periods were carefully chosen so that all sperm counts in each period were performed by hemocytometer; during part of the intervening period our laboratory used either photographic or computer‐aided sperm analysis methodology for performing semen analyses.

Statistical analysis

The prerequisites for inclusion of a patient's results in the study were very strict. Therefore, sperm counts were entered into the database only from the patient's first semen analysis in our laboratory, if he had indicated that a complete semen sample was produced by masturbation after 3–5 days sexual abstinence, and our records indicated that the couple had presented with a history of primary or secondary infertility and that there were no indications of major health problems in the male partner. In addition, the sample must have been analyzed within 3 h of ejaculation. In order to eliminate patients with severe reproductive tract pathologies, and therefore to more closely approximate the normal population with respect to sperm counts, all patients with counts of 5 million/mL and below were excluded from the database. Consequently, 33 patients were excluded from the 1977–1981 database (33/342, 9.6%) and 69 were excluded from the 1997–1998 database (69/628, 11.0%). Therefore, a total of 309 patients were included from tests performed between 1977 and 1981, and 559 patients from tests performed between 1997 and 1998.

For those variables, which were normally distributed (patient age, ejaculate volume), the means of the two groups were compared using the parametric Student's t‐test. In the case of abstinence, which was recorded as either 3, 4 or 5 days, it was not possible to determine the distribution, so median abstinence was used and the non‐parametric Mann–Whitney U‐test was used to compare the location of ranked variates. Sperm concentration and total sperm number per ejaculate, which are not normally distributed, were reported as mean (median) to allow comparison with previous reports which often gave mean values only. Sperm concentration and total sperm number in the two groups were compared by both the non‐parametric Mann–Whitney U‐test and the parametric Student's t‐test (after cube‐root transformation of variates to normalize the distribution).

Correlation analysis and stepwise regression (including group as a variable) were performed on the merged files (n = 868) to clarify the possible influence of the patient's age and abstinence period on the sperm counts under comparison.

RESULTS

Sperm counting precision

THE INTRALABORATORY COEFFICIENT of variation (CV) for sperm counts by hemocytometer has been determined 14 times within the 20‐year period by analysis of blind readings by three to five participating scientists or technicians. The CV estimates ranged from 2.7 to 17.5% with a median of 6.8%.

Comparison of study groups

A summary of results for the two study groups is presented in Table 1.

Table 1.

Mean (standard error) for each variable under comparison for the two groups of patients

Variable	1977–1981 data	1997–1998 data
Age	32.18 (0.35)	35.08 (0.26)
Abstinence	3.66 (0.04)	3.64 (0.03)
Volume	3.91 (0.11)	3.62 (0.07)
Count	87.93 (3.86)	92.03 (3.07)

There was a statistically significant drop in the mean seminal volume between the two groups (3.91 vs 3.62 mL, P = 0.015), but there was no difference in the median abstinence (both were 3 days, P = 0.81). The more recent group of patients were significantly older than the patients of 20 years earlier (mean 32.18 vs 35.08 years, P < 0.001). The mean (median) count for the early patient data was 87.9 (75) versus 92.0 (76) for the recent group of patients. The sperm counts were not significantly different by the non‐parametric Mann–Whitney U‐test (P = 0.77) or the parametric t‐test on cube‐root transformed variates (P = 0.80). The frequency distributions of actual sperm counts in the two groups and their respective distributions after cube‐root transformation are presented in Figure 1. The significant difference in ejaculate volumes presented above was not reflected in a statistically significant difference in total sperm numbers in the ejaculate with 320.67 (255) for the old data versus 313.07 (234) for the recent data (P = 0.45).

Figure 1.

Frequency distribution of sperm counts from (a,b) 1977 to 1981 patients and from (c,d) 1997 to 1998 patients. The original data is shown in the left panel and after cube‐root transformation, in the right panel.

Because the initial comparison showed that both the patient's mean age and ejaculate volume were significantly different between the two groups, a correlation analysis was performed on the merged files. As shown in Table 2, the ejaculate volume was negatively correlated with both the patient's age and the transformed sperm count (cube‐root) and positively correlated with abstinence. In addition, the sperm count was positively correlated with both abstinence and age, but there was no direct correlation between abstinence and age. Subsequent stepwise regression analysis by reverse mode including group as one variable in addition to count, volume, abstinence and age, indicated that age was not an independent correlate of count. The final model retained both abstinence and ejaculate volume (Table 3). However, these weak correlations do not affect the result of the comparison presented above because there was no significant difference in abstinence or in total sperm content (count × volume) in the ejaculate.

Table 2.

Spearman's correlation matrix between cube‐root transformed sperm counts, patient's age, period of sexual abstinence and ejaculate volume

	Sperm count	Patient age	Abstinence
Patient age	0.08 (P = 0.017)	–	–
Abstinence	0.14 (P < 0.001)	0.06 (P = 0.1)	–
Volume	−0.17 (P < 0.001)	−0.13 (P < 0.001)	0.11 (P = 0.001)

Table 3.

Model fitting results for sperm count (cube‐root) as the dependent variable obtained by stepwise multiple regression in reverse mode starting with independent variables of age, abstinence, volume and group

Independent variable	Coefficient	SE	t‐value	P
Constant	3.741981	0.191304	19.5604	<0.0001
Abstinence	0.238181	0.048865	4.8742	0.0001
Volume	−0.126353	0.021958	−5.7543	<0.0001

SE, standard error; r‐squared = 0.0538, n = 868.

Model: Count^.33 = 3.742 + 0.238 (abstinence) − 0.12 (volume).

DISCUSSION

THE ORIGINAL META‐ANALYSIS performed by Carlsen et al. ⁸ suggested that the mean sperm concentration of normal men had dropped by approximately 50% over the previous 50 years. The veracity of this conclusion has been widely debated, ¹ , ⁹ , ¹⁰ however, subsequent re‐analysis of the data ¹¹ , ¹² , ¹³ , ¹⁴ tended to confirm the main conclusion from the original study. In particular, the study by Swan et al. ¹⁴ was the most extensive meta‐analysis of sperm count data to date. Their meta‐analysis included 101 studies between 1934 and 1996 and it took account of geographic variations in sperm count, the subjects age and fertility history, their period of abstinence and method of semen collection. This study confirmed the sperm count decline in both Europe and North America, with an overall drop in sperm counts of around 50% over 60 years. However, there remains the possibility of other confounding factors which may have been overlooked, including circannual/seasonal variation in human sperm counts. ¹⁵ , ¹⁶ , ¹⁷ In the present study, any potential effect of seasonal variation was precluded by analyzing data spread over 2–4 complete years. More recent studies on data from different populations has tended to confirm that semen quality has deteriorated in some geographic localities, ¹⁸ but not in others ² and that this variation was one factor which confused the issue around the original meta‐analysis. ¹⁹ , ²⁰ , ²¹ It has been speculated that semen quality may have been affected by environmental factors such as steroid hormonal analogs, ²² , ²³ , ²⁴ however, there are many other environmental and life‐style factors which could potentially have detrimental effects on spermatogenesis. ²⁵ , ²⁶ , ²⁷ , ²⁸ The observation that semen quality has been deteriorating in some areas is a major concern which requires extensive and rigorous further investigation both retrospectively and prospectively on large samples from carefully selected specialist semen laboratories. The World Health Organization (WHO) has developed a prospective multicentre study called ‘Sentinel Surveillance of Semen Quality and Time to Pregnancy’ for prospective participating laboratories situated in Asia, Africa, South America, Russia and The Pacific. Similar, but independent studies will be conducted in Europe, Japan and the USA. The need for exceptionally rigorous quality control protocols, identical procedures for testing, rotation of some scientists between the different centres in the network and external quality assurance on identical samples with central analysis of results, are important aspects of the protocol for the WHO study. Unfortunately, this excellent project has not yet commenced.

The results of the present study, the third from Australia, concur with those of Handelsman ² and Costello et al. ²⁹ in that they have provided no evidence of decreasing sperm counts in Australia over the last two decades of the twentieth century. It can be argued that the use of first test patient data should result in minimal bias because the patients are not self‐selecting, as is the case for sperm donors or trial participants. Regardless, as mentioned in the Introduction, any significant trend towards lower (or higher) sperm counts in the general population should be reflected in men presenting for infertility investigations. As pointed out by Jegou et al., ¹ it is not essential to analyze a sample which is exactly representative of the general population in order to detect geographic or secular differences in semen quality. It is more important to analyze data from rigorously defined groups of men at each time point, including equivalent test performance (which is difficult to guarantee when comparing tests performed 20 years apart, in most laboratories), database veracity and valid statistical handling of the data. The finding of almost identical mean sperm counts in these patient groups, performed approximately 20 years apart, especially considering that the two groups had identical median abstinence times, lends no support to the hypothesis that counts are decreasing. It could be useful to analyze similar data from other centres in the southern hemisphere, in order to determine whether the findings show regional variation, as appears to be the case in the northern hemisphere.

In summary, this analysis, performed on a tightly controlled data‐set, indicates that there has not been a drop in sperm counts in Melbourne in a defined subset of men presenting for infertility investigations over the last two decades of the twentieth century. These results concur with two previous Australian studies on potential sperm donors in Sydney. It is of course possible that Melbourne and Sydney results are similar because of similar environmental/cultural homogeneity and stability in the south‐eastern corner of Australia, and that other regions of Australia may have experienced a decline in sperm counts over time. Rigorously controlled prospective studies are required in order to detect any further changes in human sperm counts and to obtain a global perspective with respect to environmental influences on the process of spermatogenesis.