Baskadia

The determination ofsperm concentration by a single haemocytometer count is the main basis for deciding whether a human semen specimen falls within normal limits and for predicting whether the donor is fertile. Furthermore, a single haemocytometer count has been the standard method for measurement of the sperm concentration of the specimen in most of the research studies on semen production in man. Recently, Freund (1962, 1963) has studied the among-donor and specimens-within-donor variation in human semen speci¬ mens and the effects of factors, such as frequency of emission, age of the specimen, and method ofspecimen collection, on semen characteristics. In these studies, the sperm concentration of a specimen was determined by a single haemocytometer count and it became apparent that very little was known about the accuracy and repeatability of the haemocytometer technique as applied to the counting ofhuman spermatozoa.

Berkson, Magath & Hum (1940) and Biggs & Macmillan (1948) studied the sources of error involved in erythrocyte counting by haemocytometer and found marked variation among observers, among duplicate pipettes, and among duplicate chambers. Bane (1952), in an investigation of the haemocytometer technique as applied to bull semen, found that 55 % of the variation between duplicate determinations by the same technician was due to sampling and pipetting error and 45 % to chamber and counting error. He reported that the 99% confidence interval, in a single determination of 1000 spermatozoa per mm3, was ±229 spermatozoa. Campbell, Hancock & Rothschild (1953) and Campbell, Dott & Glover (1956), counting live and dead ram, boar and bull spermatozoa, reported a significant difference between the estimates made by two experimental workers and set up tables for the determination of 95 % fiducial limits of a single count by one observer. The general plan ofthis investigation was to determine what part ofthe total variance associated with a study of the sperm concentration of human semen specimens is due to the haemocytometer technique, to analyse thevarianceofthe haemocytometer technique itself in order to assess the relative importance of each step in the procedure, and to estimate the confidence interval of a single haemocytometer count in the measurement of sperm concentration. A prelimi¬ nary report on this work has been published (Freund & Carol, 1963).

Semen specimens, collected by masturbation into clean dry sputum bottles, were received from regular donors who were young white men and students at this medical college. The three research technicians who made the counts in this study ranged in experience with haemocytometer work from less than 1 year (Technician 1) to more than 3 years (Technicians 2 and 3). Each of the three technicians made two pipettings from every specimen in the study and filled two chambers of a haemocytometer with each pipette. Standard white cell pipettes (Clay-Adams, A-2450) and haemocytometers (Levy chamber with double Neubauer ruling, Clay-Adams, A-2900) were used. The specimen was mixed by swirling, semen was drawn up to the 1 graduation on the pipette (0-1 ml), and the pipette was filled to the top mark with a sperm diluting solution (16 g sodium bicarbonate and 4 g phenol in 400 ml distilled water). The pipette was shaken in an electrical shaker for 30 sees, the first few drops were discarded, and both chambers of a haemocytometer were filled from the pipette. The cells were allowed to settle and were counted under low power (x 150) of the microscope. The number of spermatozoa in each of the four large corner squares of the chamber were counted and recorded separately. The technicians had no knowledge of each other's counts during the course of the determinations. An analysis of variance was made on thirty-six specimens (the first six specimens from Donors 1, 2, 3 and 5, 6, 7, Table 1) in order to avoid the difficulties inherent in an analysis of data with unequal subclass numbers. A preliminary examination of the data confirmed that they were distributed in Poisson fashion with variance proportional to the mean. Therefore, the data were transformed to the square roots of the counts. The statistical analyses followed the procedures outlined by Snedecor (1956).

These data demonstrate that there is a large amount of variation among technicians and within technicians in the determination ofsperm concentration by counting in the haemocytometer. This is particularly striking in view of the fact that the three technicians who participated in this study were trained in the same laboratory and used the same methods and equipment. The techni¬ cians were also aware ofthe fact that they were participating in a study oftheir technique and it is most probable that they made every effort to achieve the most accurate results possible, within the limitations of their training and abilities. The possibility that an unconscious bias was present cannot be dis¬ counted, since each technician was aware that she was counting the same specimen four times and there might have been a tendency to make the counts closer to each other than might have been the case if the technician was not aware that the four counts were being made on the same specimen. While such a bias might affect the variability of the counts of the same specimen by each technician, there would be no such effect on the variability of counts among technicians. When the technician error is converted to a confidence statement, it may be stated that the 95 % confidence interval for haemocytometer counts, among technicians, on a single semen specimen which is equal to the mean of this study, is ±52 %. In other words, when two technicians count the same semen specimen, 95 % of the time the second count will be within ±52 % of the first, and 5 % ofthe time the second count will be more than ±52 % ofthe first. It is suggested that the differences in haemocytometer counts among technicians working in different laboratories would be ofthis magnitude or greater. It is apparent from these data that there is need for extreme caution in com¬ paring the experimental results from two laboratories in studies of sperm pro¬ duction, based on haemocytometer counts made in the different laboratories, since a considerable part of such apparent differences may be due to variation among technicians in counts. This variation among technicians could become a serious problem in co-operative studies among two or more laboratories and every attempt should be made to standardize the haemocytometer technique before such studies are started. It is also of importance to consider the variance among technicians within the same laboratory when more than one technician works on the same study. It would be preferable to have one technician do all of the haemocytometer counts for one study, but, ifthis is impractical, care should be taken so that the specimens are randomly assigned to the technicians for counting. If this is not done, a greater systematic error among technicians could be built into the study due to the differences among technicians in haemocytometer technique. The variance among technicians in the same laboratory or among technicians in different laboratories is large and of evident importance and confirms the report of Campbell et al. (1956), who found that a significant difference existed between the estimates of live-dead counts of bull spermatozoa made by two workers. This variance must be considered in the design ofthe experiment when it is planned to test treatment differences in the statistical analyses of such studies. Not so evident, however, is the variance among duplicate determinations by the same technician on the same specimen. These data indicate that mean differences of the order of 20 % may be expected between duplicate determin¬ ations by the same technician. The variance, among duplicate determinations by the same technician, encountered when counting human spermatozoa is very similar to that reported by Berkson et al. (1940) for leukocytes and by Bane (1952) for bull spermatozoa. This variability must be considered as part ofthe experimental error, since all the steps in the technique are involved (Table 3) and since significant technician-specimen and technician-donor effects exist. These data support the earlier reports (Freund, 1962, 1963) that a single semen specimen cannot be used to estimate the level of a donor's semen production with an accuracy suitable for research purposes. This is particularly true when the sperm concentration has been determined by a single count in the haemo¬ cytometer, since the variance ofthe technique is incorporated in the estimate. In studies where a very accurate determination of the sperm concentration ofeach semen specimen is critical, e.g. in the calibration ofa photometer for bull sperm concentration determinations (Salisbury, Beck, Elliot & Willett, 1943) or in the development of a method for the estimation of bull sperm concentration by packed cell volume (Foote, 1958) at least two well-trained technicians should do the haemocytometer counts and duplicate pipettes and chambers should be used by each technician. Since the repeatability of duplicate photometric determinations of sperm concentration is so high, it is not usually emphasized that the calibration ofthe photometer is based on haemocytometer counts and that the variance associated with such counts is incorporated into the photo¬ meter method. In a study where the treatment effect on sperm concentration is expected to be small, the variance ofthe technique might be critical. The most practical single step to decrease the variance associated with the haemocytometer technique would be to load and count duplicate chambers from each pipette. In any case, the variance associated with the technique must either be reduced to a negligible level or else it must be accounted for in the statistical analysis of the data.