Summary The results of this statistical analysis of counts made by the direct microscopic and agar plate methods show a high degree of variability in both methods, as applied comparatively to rating raw market milk supplies. In other words, there is such a relatively large experimental error in both methods that sanitarians interested in the bacteriological control of milk supplies must avoid the danger of making too fine distinctions in establishing the maximum numerical limits for the various, so-called, grades of milk. Variations of 200 to 300 per cent have been frequently reported in counts made in well controlled laboratories. While it is true that some technicians, under exceptionally well controlled conditions, have been able to obtain smaller percentage variations, it should not be forgotten that we must be governed by limitations arising through practical application where many samples of milk are daily prepared for counting by technicians who do not have time to exercise that extreme care which is possible only in research laboratories where there always exists ample time in which to plate the half dozen or so samples. The direct microscopic count of individual cells per cubic centimeter varied from 4.8 to 6.2 times greater than the count of the groups of bacteria under the microscope, and from 3 to 4.5 times greater than the agar plate colony count. The plate count on either plain or lactose agar was greater than the corresponding microscopic group counts. This relationship is to be expected in most samples due to the breaking up of the groups into smaller components during the process of plating the milk. The individual cell count averaged 4.5 times greater than the plain agar plate count but where lactose was added this difference was reduced to 3 times. There was likewise a reduction in the variability as well as a noticeable increase in the correlation. Undoubtedly the addition of lactose to the medium would materially increase the accuracy of the plating method. Although the microscopic group count was more variable than either the count of individual cells by the microscopic method or the agar plate count, yet the probable error of the coefficients of variability showed little or no significant difference. There was a better correlation between the counts of the individual bacterial cells and groups of cells per cubic centimeter by the direct microscopic method than between the individual cell count and the agar plate count. The poorest correlation obtained was that between the microscopic group counts and the agar plate counts. This is undoubtedly largely due to the variations introduced in the former count by the arbitrary manner in which separate groups are determined and by the counting of dead organisms and in the latter by variations in the adaptability of different organisms to growth in cultural media, and also by variations in the breaking up of colonies during plating. A few excessively high counts may have a marked effect upon the degree of correlation. For example, the elimination from 439 counts in all, of nine which were excessively high and which represented over-crowded microscopic fields and agar plates, resulted in a markedly lower correlation throughout (see table 7). It is conceivable that with raw milk uniformly low in bacterial content that there might be very little if any correlation between the two methods. Although, while showing more or less lack of correlation as applied to individual samples of milk, yet a close agreement may, in general, be expected in the results obtained by the two methods in rating the whole supply. Arithmetic averages of bacteria counts are apt to be misleading as a measure of the central tendency. One high count among several that were uniformly low would give an erroneous idea of the numerical bacterial quality of a given milk supply. The median, which is the middle number of a series of counts arranged in sequence, is a much fairer measure. One method gives as reliable a picture of the bacterial condition of any given raw milk supply as does the other. There is no reason for believing that one method is significantly more accurate than the other; nor that one lends itself more advantageously to the dividing of a raw milk supply into classes than the other. The plating method, however, is more applicable to the counting of bacteria in milk where there are only a few thousand present per cubic centimeter. Irrespective of the wide variations and possible sources of error that may be pointed out, both methods, if properly administered, are sufficiently accurate to insure a marked reduction in the amount of carelessly handled milk, without inflicting an undue hardship upon anyone.
Read full abstract