Ice Cream* 2. Do solid and liquid media yield the same. of ice cream? 3. Do the number of coliforms vary according to the flavor of the product, the

Transcription

1 A Survey of Coliform Organisms in Ice Cream* Commxittee on Research and Standards THE presence of coliform organisms in frozen desserts is generally believed to have the same sanitary significance as does their appearance in pasteurized milk.1 These organisms serve as indicators of poor sanitary practice. Experience has shown, however, that the average number of coliforms found in ice cream is greater than that found in pasteurized milk. No coliform standards for ice cream are suggested in Standard Methods. The manual does give directions for performing the test on ice cream, however. The original purpose of the present study was to compare the yield of coliform organisms by alternative methods of preparing the sample for analysis. namely, (A) by weighing it in the unmelted state, and then diluting it for plating as opposed to (B) allowing the ice cream to melt and then diluting it volumetrically, or measuring it volumetrically and plating it in the undiluted state. The first method is permitted in the Ninth Edition of Standard Methods for the Examination of Dairy Products (p. 205). Standard Methods does not approve of the volumetric measurement of melted ice cream for analysis except for screening tests (pp ). The data were also collected in an attempt to obtain answers to the following questions: I. What is the practical optimum quantity of ice cream which yie!ds a reliable finding on solid differential medium? This study was carried out by a Subcommittee on the Examination of Frozen Desserts of the Subcommittee on Standard Methods for the Examination of Dairy Products with the collaboration of laboratories in the United States and Canada. 2. Do solid and liquid media yield the same numerical estimates of coliform content of ice cream? 3. Do the number of coliforms vary according to the flavor of the product, the season of the year, or the type of laboratory, i.e., industrial, public health, or educational? 4. What is the usual numerical range of coliforms found in ice cream? METHODS The pertinent directions given to all laboratories were contained in an instruction sheet and are as follows: Two wide-mouthed vials (A and B) are filled with sample. A is kept frozen and B is iced or refrigerated (320 or 400 F.). Transport A and B in suitable containers under conditions met with in actual practice, such as from plant to laboratory. Plating is to be done as soon as practical after receipt of samples in the laboratory. In "Remarks" please state the transportation procedure and plating time used. In the laboratory test the samples according to Standard Methods, Ninth Edition, with modifications, as follows: A. Temper at room temperature 10 minutes; weigh gm. into a dry, sterile dilution bottle, using a spoon dipped in alcohol and flamed; make into 1-10 dilution by adding proper volume of water. Pipette into plates or tubes as is shown in the tabulation on the following page. Petri dishes are to be poured with ml. desoxycholate agar. Standard Methods, Ninth Edition is to be followed in all other respects. Where results in the plates or tubes are not distinct or clearly discerned enter tabulation as N.D. (not discernible). Count only dark red colonies measuring at least 0.5 mm. in diameter in the Petri dishes after hours at 350 C. If pink colonies are encountered, make completed test before tabulating coliforms. Note these resuits at base of column. Run completed tests when presumptive reactions are at variance between solid and liquid [761

2 Plate COLIFORM ORGANISMS IN or Tube Procedure 4 dishes, 1 ml. of diluted sample each 4 "22 4. " 5 " " " "' Br:lliant green bile lactose broth (2 per ccnt) fermentation tubes 10 tubes, 1.0 ml. of diluted sample each ICE CREAM Plate No. or Tube Letter 1, 2, 3, 4 5, 6, 7, 8 9, 10, 11, 12 y 77 Cell Al A2 A3 A4 B. Proceed as in 7.11 b. (p. 205). Pipette into plates or tubes as follows: Plate No. or Tube Letter I dish, 0.1 ml. direct (not diluted) 4 dishes, 212 ml. of dilution sample each (11 ml. added to 99 ml. water) 2 dishes, 1.0 ml. direct (not diluted) each Brilliant green bile lactose broth (2 per cent) fermentation tubes Cell Plate or Tube Proceduire B1 5, 6, 7, 8 B2 9, 10 B3 10 tubes, 0.1 ml. direct (not diluted) each x B4 media. Fermentation with gas in the tubes within 48 hours at 350 C. constitutes a positive result. Printed data sheets were also furnished. The latter were filled out and returned to the committee usually monthly and occasionally bimonthly. It was suggested that a separate sheet be used for each product (ice cream, or mix, or ingredient) checking the proper square at the head of the sheet. All entries were to be in ink or typed. Actual counts of Petri dish results were to be entered. The tube results were to be recorded as either P (positive) or N (negative) or N.D. (not discernible). The media used were arbitrarily limited to one solid medium and one liquid medium. Sodium desoxycholate lactose agar * (No. 114) in the dehydrated form was the solid medium used and brilliant * The Subcommittee is greatly indebted to the Baltimore Biological Laboratories, Inc., Baltimore, Md., which furnished all the sodium desoxycholate agar and to the Difco Laboratories, Inc., Detroit, Mich., which supplied all the brilliant green bile lactose broth. green bile lactose broth (2 per cent), also in the dehydrated form, was the liquid medium used. This agar medium, rather than the other commonly used medium, violet red bile agar, was used only as a matter of convenience. Likewise, the brilliant green bile lactose broth was employed because it is the most commonly used liquid medium for this purpose. International Business 'Machine equipment was used for punching data cards from data sheets and sorting and tabulating them. A. glossary of terms used in the analysis of the data follows: (a) Cell-each of the 8 subdivisions of technic described above. (b) Cell Total-sum of coliform colonies on all the Petri p'ates in an individual cell. (c) Punched Sample-sample which was (d) transferred to an IBM card. Nonpunched Sample-sample not transferred to an IBM card for one of the following reasons: 1. All 6 cells (A1, A2, A3, B,, B2, B3) yielded no coliform colonies.

3 78 YEAR BOOK 2. At least one cell of the sample had an average number of coliform colonies per Petri plate equal to or greater than At least one plate of a cell was marked T.N.C. (Too numerous to count). (e) Short Sample-a sample in which cells A2 and B2 were omitted by the laboratory performing the test. After the study had proceeded for some time, certain laboratories were instructed to omit the A2 versus B2 comparisons. (f) Summer Sample-a sample plated during the months of May through October. (g) Winter Sample-a sample plated during the months of November through April. (h) Product-one of the 6 categories of ice cream or ice cream mix used in the analysis. They are as follows: (1)-vanilla, (2)-chocolate, chocolate chip, chocolate frosty and fudge, (3)- fruit, fruit with flavor other than nut, (4)-nut, nut with flavor other than fruit, (5)-ice cream mix, (6)-all others-coffee, lemon, lime, orange, mixed fruit and nut, etc. FINDINGS A comparison was made first between cells A2 and B2. These 2 cells were chosen for comparison because all factors were identical except the method of preparing the sample. Included in each was the same number of plates (4) and the same quantities of ice cream per plate (0.25 ml., in one case by weight, and in the other by measure). It was first determined that the 4 plates of a cell did not differ significantly from each other (Appendix 1). Studies were then made to determine if the average difference in count between these 2 cells was significant (Appendixes 2a to 2d, et seq.). These were carried out by several statistical procedures on a varving number of samples ranging from a maximum of 1,511 to a minimum of 518, comprised of all flavors, and from all laboratories. Although the count on A2 exceeded that on B2 in more than half of the cases, a significant difference, nevertheless most of the comparisons of average differences were not significant. A possible exception was that the geometric mean of A2 was 7 per cent greater than that of B2. While this difference is of borderline significance from a practical viewpoint it is of questionable importance when critical low count results are considered. Next, A3 was compared with B3 in a manner similar to the above. The total quantity of ice cream in each cell was the same (2 ml.) but cell A3 had 4 plates and cell B3 only 2. It was found, in contrast with the A2 vs. B2 comparison, that A3 yielded greater counts than did B3, as shown by the average arithmetic difference and by the number of times A3 exceeded B3. Thus, there was a mean difference of 4 coliforms per ml. for those samples of which at least one of the cells had a count of 10 or less and A3 was greater than B3 83 per cent of the time (Appendix 3). The first 3 cells of the nonmelted ice cream (A1l A2, A3) were then compared against each other (Appendix 4) in order to determine which of the 3 quantities of ice cream tested yielded the greatest number of colonies per unit volume. It was found that A3 > A2 > A1, (i.e., that 5 ml. of the 1/10 dilution was better than 2½I2 ml. of the same dilution and that the latter was better than 1 ml.). The difference in both cases was statistically significant. The data from the two cells using liquid medium (A4 and B4) were then compared (Appendix 5). It was determined that A4 had a significantly greater proportion of positive tubes. A4 exceeded B4 in 58 per cent of the cases. The plate method giving the highest average count (A3) then was compared with the tube method giving the highest average MPN (A4) (Appendix 6). The plate count method gave significantly higher counts than did the tube method. For example, when comparison was limited to those samples where at least one of the cells had a count of 10 or

4 COLIFORM ORGANISMS IN TABLE 1 ICE CREAM 79 Summary of Findings on Methods of Analyzing Ice Cream for Coliform Organisms Unmelted Ice Cream (A) (Weighed and Diluted 1/10) Melted Ice Cream (B) (Measured) Ml. per Plate 1 ml. 2.5 ml. 5 mi. Volume per Tube 1 ml. No. of Plates No. of Tubes 10 Total Gm I 1..o RELATIVE PRODUCTIVITY Designation Al BI A I A2 > Ba A I As > BsI V A4 BI Tot6l Gm No. of Plates I No. of Tubes Diluted 1/10 Ml. per Plate Undiluted Ml. per Plate Volume per Tube 0.1 less it was found that the mean plate count was 8 per ml.; whereas the mean MPN was 2.4 per ml., a highly significant difference. The A3 findings with all samples were compared to determine whether the factors of flavor of product, season of year (winter or summer), and classification of the testing laboratory had any influence on the results (Appendixes 7 and 8). It was ascertained that none of these factors influenced the count.* Standard Methods, as was mentioned, does not suggest a standard for coliforms in ice cream. This is partly because there are no assembled data on coliform incidence upon which a standard might be based. The present study afforded the opportunity to analyze a group of data assembled from localities distrib- * Relatively few samples of certain fruit flavored ice creams, particularly bananas and peaches, were analyzed in this study. It is known, however, that these fruits are a common cause of high coliform counts in ice cream. They, thus, represent a special problem. Some manufacturers, nevertheless, have apparently satisfactorily solved this problem. uted over this country, from which an estimate of coliform incidence in ice cream might be derived. The A3 data (2.012 samples) were arranged for this purpose in a cumulative series according to estimated coliform numbers per ml. from 0 to too numerous to count. It was noted that 44 per cent of the samples revealed the presence of no coliform organisms per ml. in the quantity tested (2 ml.), that a little less than 23 (63.5 per cent) had a count of 5 or less, that 70 per cent had a count of 10 or less, 78 per cent a count of 20 or less, and 90 per cent 200 or less. These findings might be used as a basis for coliform standards. Thus, for example, if a system were adopted whereby it is required that 3 of 4 counts not exceed the standard, it might be set at a figure in the range of 10-20, preferably at 10. SUMMARY AND DISCUSSION Two methods of preparing ice cream for coliform analysis were compared, i.e., weighing it in the unmelted state and

5 80 YEAR BOOK then making a 1/10 dilution before plating (Al, A2, A3), or inoculating into liquid medium (A4), versus allowing it to melt and then making the same dilution by volumetric measurement before plating (B2) or plating in the undiluted state (Bl, B3) or inoculating in the same state into liquid medium (B4). See Table 1 for a summary of these findings. It was found in the (A2) versus (B2) comparison that, although the unmelted samples were more frequently higher in count than the melted ones, the actual difference in count was of dubious significance. However, when (A3) versus (B3) was compared, the variation was quite significant, the average arithmetic difference of those samples, in which at least one of the cells did not exceed 10 per ml., being 3.6 colonies. Likewise, when the 2 methods of preparing ice cream for inoculation into liquid medium were compared, it was found that (A4) vielded a higher average MPN. When the 3 different quantities of ice cream used for each coliform determination on unmelted ice cream were compared for relative productivity it was found that the greatest volume (5 ml. of a 1/10 dilution in each of 4 plates for a total of 2 gm.) yielded a significantly higher estimate than did '2 that volume (1 gm.). Likewise, the latter quantity gave more colonies per unit volume than did the smallest amount tested (0.4 gm.). Since the use of both solid and liquid media is permitted by Standard Methods for coliform analysis of dairy products, the comparison of 2 representative types of such media is of interest. It was found that the solid medium (sodium desoxycholate agar) was more productive than the liquid one (brilliant green bile lactose broth). Thus in the example cited, where the counts were in the lower range, the yield by plate count was more than 3 times that of the tube MPN estimate. A poll was taken of the cooperating laboratories to determine the opinion hed by the participants of the methods used. Most of the laboratories responded. There was an overwhelming preference for the solid as opposed to the liquid medium (20-2). When this fact is combined with the objective findings of greater productivity of the solid medium discussed above, it seems that the use of liquid medium should not be recommended. Likewise, the use of an unmelted sample was preferred to the use of a melted one by a vote of Ten of 23 experienced some difficulty in weighing the frozen sample and a few in adding the proper amount of water to make the 1/10 dilution. There apparently was little objection to plating the 1/10 dilution made from the unmelted ice cream in 1 ml. or 2.5 ml. quantities or reading those plates after incubation. There was, however, some opposition, to the use of the 5 ml. quantity because of difficulty in pouring the plates (about 1 in 3) and reading them afterward (8 of 19). Objections were raised, on the other hand, by about half of the laboratories to the direct pipetting of undiluted ice cream, particularly the 0.1 ml. quantity. There was similar disapproval caused by difficulty in mixing the undiluted ice cream with the agar. Furthermore, about half of the laboratories experienced difficulty in counting the plate inoculated with the 1 ml. quantity. Integration of the analysis of the data and the poll of participating laboratories leads to certain conclusions concerning the unmelted vs. melted, diluted vs. undiluted, and size of sample. There is preference for the use of unmelted ice cream and the latter method apparently yields higher coliform findings than does melted ice cream. The use of melted and undiluted ice cream for plating seems to have found little support. This is partly caused by difficulty in pipetting and also in distributing the ice cream evenly in the plate. It should be

6 COLIFORM ORGANISMS IN ICE CREAM 81 noted that this method has been used in the New York City Health Department Laboratory for several years. However, the use of a mechanical device (Boerner Rotator) was found necessary to secure uniform distribution of ice cream in the plate. The finding that the largest quantity of unmelted ice cream used (5 ml. 1/10 dilution) gave the greatest yield and hence the most reliable result conflicts with the stated opinion that 5 ml. is difficult to handle in a single Petri plate. It would seem impractical to decrease the volume per plate and increase the number of plates used. A possible solution would be to decrease the dilution from 1/10 to 1/5. If 2 ml. of ice cream is the preferred quantity for analysis, then many laboratories have not been using an adequate quantity in routine analysis. The finding that neither the flavor of the product nor the season of the year had a significant effect on the number of coliforms was surprising. It is known that certain products such as whole fruit ice cream sometimes tend to have higher coliform counts than other flavored ice cream because of the addition of some fresh fruit to the product after pasteurization. Yet, a very detailed breakdown of the data failed to reveal a significant difference among all the flavors tested. Similarly, in a study carried out several years ago in New York City, a difference in summer and winter counts on ice cream and on pasteurized milk was noted. The present data cover a diverse array of samples studied during all the seasons for a period of several years at laboratories throughout the United States. Because they thus offer some idea as to the coliform content of ice cream, it is suggested that they might serve as a reasonable guide to coliform standards. If the over-all findings on the incidence of coliforms in ice cream can be accepted as typical they support the general belief that coliform counts in ice cream are higher than those in pasteurized milk and suggest, perhaps, that efforts are in order to determine if it is possible to improve this record. CONCLUSIONS 1. Unmelted rather than melted ice cream is preferred for the coliform analysis of that product. 2. The use of undiluted ice cream was found to be unsatisfactory. 3. The validity of coliform analysis is related to the volume analyzed. The largest quantity tested, 2 gm., gave the highest count. 4. The one solid medium used (sodium desoxycholate agar) gave significantly higher coliform estimates than did the one liquid medium used (brilliant green bile lactose broth). 5. Neither season of year, flavor of product, nor type of laboratory, under the conditions of the study, seemed to influence the number of coliforms in ice cream. 6. It is suggested on the basis of the findings that a coliform standard might be established for ice cream in the region of 10 per ml. with a provision that 3 of every 4 samples comply. REFERENCE 1. Standard Methods for the Examination of Dairy Products. Ninth Edition. New York: American Public Health Association ACKNOWLEDGMENTS-The authors are grateful to Dr. John W. Fertig of the School of Public Health, Columbia University, who furnished invaluable guidance in the statistical analysis. The subcommittee is greatly indebted to the following persons and their sponsoring organizations for their gracious collaboration in the study: W. Ahlstrom, Carnation Co., Los Angeles, Calif.; Professor E. 0. Anderson, University of Connecticut, Storrs, Conn.; George L. Andrews, Breyer Ice Cream Co., Philadelphia, Pa.; Carleton J. Austin, Supple- Wills-Jones Milk Co., Philadelphia, Pa.; Dr. Franklin W. Barber, National Dairy Research Laboratory, Oakdale, Long Island, N. Y.; Dr. Fred W. Bennett, University of Georgia, Athens, Ga.; John D. Bowers, Borden's Moore and Ross, Columbus, Ohio; Hugh Butner, Florida State Board of Health, Jacksonville, Fla.; E. L. Byers, Bowman Dairies, Chicago, Ill.; Dr. W. R. Carroll, University of Florida, Gainesville, Fla.; E. K. Christman, Penn

7 YEAR BOOK Dairies, Lancaster, Pa.; Dr. E. S. Churchill, Michigan State College, East Lansing, Mich.; Dr. Samuel R. Damon and C. E. Schrock, Indiana State Board of Health, Indianapolis, Ind.; Dr. C. C. Flora, Virginia Polytechnic Institute, Blacksburg, Va.; Bert Giberson, Meadowsweet Dairies, Tacoma, Wash.; Dr. George H. Hauser, Louisiana State Board of Health, New Orleans, La.; R. K. Lawhorn, Abbotts Dairies, Inc., Philadelphia, Pa.; R. S. McKenzie, Providence Health Department, Providence, R. I.; N. A. Perry, Midwest Dairy Products Corp., Cape Girardeau, Mo.; R. W. Reeve, Swift and Co., Chicago, Ill.; Dr. H. B. Siegmund, Hendler Creamery Co., Baltimore, Md.; M. M. Simpson, Fairmont Foods. Omaha, Neb.; A. W. Strum, Arden Farms Co., Seattle, Wash.; L. Z. Szabo, Foremost Dairies, Jacksonville, Fla.; Dr. A. H. White, Department of Agriculture, Ottawa, Canada; J. R. Wiggins, Telling Ice Cream Co., Cleveland, Ohio; City Department of Health, New York, N. Y. LEON BUCHBINDER, PH.D. Bureau of Laboratories New York City Department of Health, 125 Worth St., New York 13, N. Y., Chairman of the Subcommittee on the Examination of Frozen Desserts WALTER C. BARTSCH W. A. CORDES LEO HABEL APPENDIX Purpose of Test Data Used Analysis Results. 1. Do the plate counts of 25 random samples none A XI value was computed The observed distribution a cell differ signifi- of which had a cell total from the 4 plates of each of X2 values was satiscantly from each other? of 0. A2 cell. The 25 X' factorily fitted by the values were fitted by theoretical distribution. the appropriate theoretical frequency distribution. Cells Ai, As, B2, Bs, were treated similarly. 2. Comparison of the A2 plate count with the B2 plate count. 2a. Do A2 and Ba differ All samples (both Hand tally of samples There is no significant significantly in their punched and non- showed the following: difference between A2 and distribution with re- punched). A2 B2 B2 with regard to this spect to number of Cell totals adistribution. cells too numerous to countable 1,511 1,511 count? T.N.C b. Is the average differ- Number of samples used 7 distributions of The average difference in ence between A2 and varied from 595 to 1,511 d=a2-b2 were obtained. favor of A2 varied from B2 greater than 0? depending on the disposi to 0.71 and in no tion of cells with counts case was statistically in excess of 30 or 10, significant. and of cells with both counts 0. 2c. Is the average differ- All punched samples d=log A2 - log B2 was d =.0293, ad = ence between log A2 where both A2 and B2 obtained. Significance is borderline. and log B2 greater had counts of 2 or than 0? more-518 samples. Antilog d = d. Does A2 exceed Ba All samples except where The proportion of times A2 exceeded Ba in 58 per in more than %2 the both A2 and B2>10 and As exceeds B2 was tested cent of the cells, which number of samples? where As = Ba. (523 for significance of differ- differs significantly from samples). ence from /2.-2/. 3. Do As and Bs give All nonshort samples ex- Test of average arithmetic As gave a significantly equal coliform counts? cept where both cells ex- difference and proportion higher count than did Be ceeded 10-1,188 samples. of times As exceeds Bs. by inspection. Average difference = 3.6.

8 COLIFORM ORGANISMS IN APPENDIX ICE CREAM 83 Purpose of Test Data Used 4. Are the colony yields All punched nonshort of Al, A2, and As, samples except where the equivalent? counts being compared were both 0. There were 536 samples in the Al and As comparison and 608 samples in the A2 and As comparison. 5. Does the productivity All punched samples of A4 exceed that of (1,171). B4? 6. Comparison of As plate count with A4 MPN 6a. Can negative sam- Nonpunched samples ples and those (842). marked T.N.C. be omitted from the comparison? 6b. Does the proportion of positive tubes from all samples with a specific As value yield an MPN estimate equivalent to that value? 7. Do the factors of type of laboratory or product form bases for classification? 8. Does the factor of the Same as 7. season of year form a significant basis for classification? i All punched samples (1,171). ii All punched samples (889) except those where A3 and A4 exceeded 20 are omitted. iii All punched samples (769) where AK20. All samples except those that were T.N.C. or where the cell average exceeded 400-1,926 samples. Analysis Weighted averages obtained by multiplying Ai by S and A2 by 2. A series of differences SAs- 2A2 and 2A2-As were obtained. A frequency distribution of the differences of number of positive tubes, i.e., A4-B4. The average number of positive tubes was also converted to MPN. estimate. Hand tally of A4 values. A distribution of the differences between the As and the MPN derived from the corresponding A4 values. A mean As count was obtained for each of the 36 subgroups, 6 flavors x 3 laboratory types x 2 seasons. The summer samples were arranged in an array of 6 rows and 3 columns and an analysis of variance performed. The same procedure was followed for winter samples. The mean summer count was compared with the mean winter count. Results The differences by mere inspection were found to be overwhelmingly in favor of A2 over Al, and A3 over A2. Average number of positive tubes: A4 = 3.81, B4 = The difference is statistically significant. The "all-zero" samples generally gave all tubes negative, and the T.N.C. samples all tubes positive. Hence, punched samples only need be employed. The average difference was always in favor of As and was significant. The average count did not vary significantly from laboratory to laboratory, or from product to product. The average count did not vary significantly from summer to winter.