Errors and Omissions in Experimental Trials - 1d
THE BRANTFORD STUDIES [cont.]
Delayed eruption or "weighting"? These results could have arisen by there being a delay in the eruption of these teeth in the Newburgh children, for it is unlikely that the eruption rate altered in the Kingston subjects. However, Ast et al., (195 1) said that "there does not seem to be any change in the eruption pattern among the children in Newburgh, the study city, as compared with those in Kingston, the control city." No definition of an "erupted tooth" was given, but it is presumed that the authors of this study did not adopt the odd method used in Evanston, where "Only teeth which were 50 per cent or more erupted were considered. A carious or filled tooth was, of course, considered regardless of its stage of eruption" (Hill et al., 1955).
The conclusion of Ast et al. that has just been mentioned was reached only four years after the commencement of fluoridation, and as the teeth considered were partially formed prior to the commencement of that process, they would not show effects which the ingestion of fluoridated water may produce on the early stages of tooth development.
If delay in eruption did not occur in Newburgh, the lower number of permanent teeth present at those ages in that city compared with that present in Kingston was due to a "weighting" effect; there having been, in proportion, more young children in each of these age groups in Newburgh than there were in Kingston. If this is the case, as it is reasonable to assume that the mean DMF rates of the younger children were lower than those of the older ones, it would appear that in these age groups the contrast between the DMF rates in Newburgh and those in the control city was exaggerated in the final report.
"Smoothing" of initial rates. In 1951 Ast et al. reported that the "initial clinical examinations made in Newburgh and Kingston in 1944-1946 were made by one examiner, at which time the DMF rates were the same." However, reference to Table 2 in that paper shows that the DMF rates per 100 erupted permanent teeth were, in Newburgh and Kingston respectively, at age six years, 8.5, 7.2; age seven years 11.7, 12.0; age eight years 17.1, 17.3; age nine years 21.2, 18.9; age ten years 21.9, 21.3; age eleven years 21.8, 21.8, and age twelve years 25.3, 25.4. Also, Table 5. which shows the DF rates per 100 deciduous teeth present, gives the rates in Newburgh and Kingston respectively as 27.2, 21.5 at age five years; 34.2, 32.1 at age six years; 42.3, 43.3 at age seven years, and 48.0, 47.2 for the eight-year-old children. Data for the DF rates of the deciduous teeth of older children were not provided.
In the first report of this study (Ast et al., 1950) no results were given for the deciduous teeth, and the results for the permanent ones were presented in three age groups, six to seven, eight to nine and ten to twelve years. It can be seen that by adding the data from children aged six years, in whom the DMF rate in Newburgh was higher than that in Kingston, to those of the seven-year-old children, in whom the reverse situation was present, the divergence between the rates prevalent in the two cities was reduced. In Table I (Ast et al., 1950) the combined rate was shown as 10.7 in Newburgh and 10.8 in Kingston. In a similar manner, the addition of the data for eight and nine-year old children and those of children who were ten, eleven and twelve years of age produced a levelling effect between the rates of the two cities in these two combined age groups. This process of combining data >from children of different ages, when reporting the DMF rate per 100 erupted permanent teeth, although it was employed in only the first and the last dental reports, may have been used in order to simplify the presentation of the data; but it had the unfortunate effect of disguising differences between the DMF rates in the two cities at the time of the basic examinations. In the next report the situation was stated more accurately, Ast and Chase (1953) saying that "the DMF rates in both cities were approximately the same at the start of the study".
Fluctuations in the control city. In this, as in other studies, it is found that the comments made in the text tend to underrate the changes that took place in the dental caries attack rates in the control city. In the summary of the paper by Ast et al. (195 1) it was stated that "the DMF rates in the control city of Kingston show no changes." In that paper, Table 2 shows the DMF rates per 100 erupted permanent teeth; in Kingston the "per cent change" in the rates of the four age categories six, seven, eight and nine years were 30.5, 7.5, 0.6 and 9.5 respectively. Small changes were shown for ages ten, eleven and twelve years. It should be noted that the six, seven and eight-year-old children all showed decreased rates between 1945-6 and 1949. No attempt was made to explain these decreases, and the water of Kingston "remained fluoride deficient throughout the study period" (Ast et al., 1956).
Fluctuations disguised. The method used by Ast et al. in 1951 was to compute the mean DMF rate per 100 teeth in all the children aged six to twelve years; the Kingston rate for this combined age group declining slightly from 20.2 to 19.9 between 1945-6 and 1949. However, by adjusting to the "permanent tooth population in Kingston 1945-6 examinations", the authors showed that the rate of 19.9 became 20.2. On this basis it could be claimed that the "rate" in Kingston had not changed, but the incorrect statement was made that the DMF "rates" in the control city of Kingston showed no "changes". These rates of 19.9 and 20.2 were produced by combining the data of young children - that had few erupted permanent teeth and relatively low DMF rates per 100 teeth with data of older children that had most of their teeth erupted, and considerably higher DMF rates per 100 teeth. The rate obtained in 1949 was then adjusted. This procedure, no doubt unintentionally, disguised the fluctuations in the rates in the control city.
In Table 2 (Ast et al., 195 1) the "per cent change" in the Kingston children aged nine years was shown as 9.5, but if the figures 18.9 and 19.1 are the correct ones for the years 1945-46 and 1949, the "per cent change" should have been stated as 1.1, not 9.5.
Variability of caries rates. Unfortunately, the variability of even the mean caries rates cannot be studied, for the rates of yearly age groups were not published in the first and the last reports (Ast et al., 1950, 1956), and the only results shown in the 1955 report of Ast et al. were based on a combined clinical and X-ray examination.
The meagre data supplied for deciduous teeth. Data regarding the caries rates of the permanent teeth were shown in each report of this study; however, only very meagre data were published for the deciduous ones. None were made available in the first report (Ast et al., 1950). In the following year (Ast et al., 195 1) the DF rates per 100 deciduous teeth were given, but only for children aged five, six, seven and eight years, and in each age group the rates had decreased both in the test and in the control cities.
Unexplained marked decreases in the control. No explanation was given by Ast et al. (1951) for the decreases in the DF rates in Kingston, where the greatest relative decrease, from 32.1 per cent DF to 24.8 per cent DF, was seen in the teeth of the six year-old children. It would have been of great interest to see whether this trend was maintained in later years, but DF rates were not stated in the tables contained in any of the later reports. However, in the following one (Ast and Chase, 1953) the situation in regard to the deciduous teeth of children five, six, seven and eight years old (now termed "def teeth per 100 deciduous teeth present") was depicted diagrammatically by means of a histogram, these unexplained decreases in the def rates in the control city being clearly seen, a small one at age five years, and considerable ones at the ages of six, seven and eight years.
The increase in caries-free teeth in the control. The only other information published regarding the deciduous teeth was expressed in terms of "Children with caries free deciduous cuspids, first and second molars". This type of table appeared first in the 1951 report of Ast et al., and the results were given for only those children who were five or six years of age. In both age groups in Kingston the figures suggest an increase in these caries free teeth, the six-year-old children changing, between 1945-6 and 1949, from 17.2 per cent to 25.5 per cent free from caries. These changes were mentioned, but no attempt was made to explain them. In the next report (Ast and Chase, 1953) data for children aged seven years were also included. This report showed that, between 1945-6 and 1951-2, the percentage of children in Kingston who had these deciduous teeth free from caries showed a slight decrease at age five years (28.2 per cent to 26.4 per cent); but in the six-year old children the percentage increased from 17.2 to 26.3; and in those who were seven years of age, it practically doubled (8.3 to 16.5). On this occasion, these changes in the control city were not even mentioned.
"Analysis" of findings. In the 1955 report of Ast et al. it was stated that "As an indication of the benefits of water fluoridation to deciduous teeth, a previous report [in 1953] analyzed the findings among the 6 to 7 year old children in each city after six to seven years following the initiation of water fluoridation." Actually the report showed findings for the children aged five, six and seven years who had caries-free deciduous cuspids, first and second deciduous molars. At the ages of six and seven years, in both the test and the control cities, there were increases in the percentages of these teeth that were free from caries; therefore, although these increases were greater in Newburgh, they should not have been attributed solely to water fluoridation. In any case, the publication of one table showing, in this selected group of deciduous teeth, the percentage changes that have just been mentioned, a histogram depicting the def rates, and twenty lines of comment in the text on the results displayed, can hardly be said to indicate the benefits of fluoridation to deciduous teeth, or even to constitute an adequate analysis of the findings in regard to the deciduous teeth present in children aged six and seven years.
Changes in caries-free teeth in the control. In the 1955 report of Ast et al. the age range was changed by not publishing the results for the five-year old children, but showing, for the first time, the results for caries-free deciduous cuspids first and second deciduous molars, for eight and nine-year old children. However, these results cannot be compared with those of the previous years, as they were based on a combined clinical and X-ray examination. Nevertheless, a comparison can be made with the rates shown in the final report (Ast et al., 1956). In the year between the 1953-4 and 1954-5 examinations, the rates in Kingston for the ages six, seven, eight and nine years changed from 10.6, 7.0, 7.9 and 0.0 to 11.1, 4.7, 1.8 and 1.6 for the respective ages. Such changes are not unexpected, for marked variations were seen in Evanston, where, also, the examinations were made by a clinical plus X-ray procedure. For instance, the percentages of children aged seven years who were drinking fluoridated water and who had caries free deciduous teeth were, in successive examinations, 11.33 (pre-fluoridation), 8.71, 3.87, 10,66, 13.01 and 17.86 (Hill et al., 1956). It would seem that assessments made on the basis of caries-free groups of deciduous teeth are not very reliable.
Changes in the sampling method. Consideration of these five dental reports shows that the sampling method changed from time to time, and that the method used in the control city was sometimes the same and sometimes different from that used in the test one. In the first report (Ast et al., 1950) it was stated:
"we are considering only those children age 6-12 who were in the original base study and who have had each successive examination until they reach age 12. Also included are new school children who entered the study at age 6 subsequent to the first examination and were present at each of the successive examinations. Thus, this study group will have only those children who we are assuming have had continuous residence in their respective cities."
Continuous residence only assumed. It can be seen that the "continuous residence" of each subject was based on assumption only, and not on statements made in a questionnaire, such as was used in Evanston (Blayney and Tucker, 1948). Therefore, it is possible that children could have been absent from the city for considerable periods between the times of successive examinations. Also, there is no assurance that the six-year-old children entering the study in any of the post-fluoridation examinations had not come to live in the area since the commencement of the study. Therefore, it is doubtful whether the objective of having "reasonable assurance that the children studied had had continuous residence in their respective cities" (Ast et al., 195 1) can be said to have been attained.
Population changes in Newburgh. "Early in 1950 questionnaires were given to more than 3,200 children in the Newburgh schools for completion by their parents" (Ast et al., 195 1). The questions asked were not stated, nor was the number of replies received, but it was said that:
An analysis of the answers to those questionnaires shows that the Newburgh population is a relatively stable one and that the inclusion of the small migrant groups does not alter the caries picture to any significant degree. Consequently, in this report there are included all 5 to 12 year old children present in the schools in Newburgh and Kingston on the days the examinations were made" (Ast et al., 195 1).
Since information in regard to the caries attack rates in these migrant groups could not have been obtained directly from the questionnaires, it is presumed that the dental record cards of those children were grouped and that the cards of the children who were judged from the answers not to be migrants were also grouped, and the data contained in the two groups in regard to the caries attack rates were compared. If that process was carried out, it was not mentioned, nor were data published which would enable the reader to assess the situation. If no differences were found between the two groups, it must be considered to be strange because by that time it was said that "The DMF rates among permanent teeth of 6 to 12 year old children in Newburgh show a consistent downward trend" (Ast et al., 195 1). The United Kingdom Mission (1953) reported that the authors of this study had "found that the proportion of immigrants in Newburgh and Kingston was too small to affect the comparison." However, although the Newburgh population was said to be "relatively stable", in the 1954-5 examinations in that city 24 per cent of the children were excluded because they failed to fulfil the residence qualifications (Ast et al., 1956).
The workers who conducted the paediatric study in these cities, Schlesinger et al., in 1950 said that in each city "An effort was made to select... children from families which might reasonably be expected to remain for the duration of the study." In spite of that precaution, they found that 29.9 per cent of their subjects in Newburgh moved from the city during the period of the study (Schlesinger, Overton, Chase and Cantwell, 1956).
Population movement in Kingston. No mention was made of the issue of a questionnaire to children in the control city; apparently it was assumed that migrants to that city would have come >from areas with "fluoride-free" water supplies. Schlesinger et al. (1956) found that 22.2 per cent of the children included in the paediatric examinations moved from Kingston during the period of the study; presumably a similar number of new residents settled in the city.
It may be considered that in moving from one locality to another, interruptions could occur to regular conservative and prophylactic treatment of the children, so that their dental health may not have been as good as that of children who lived for many years in the same city. It is possible also that regular dental examinations, by stimulating interest in the teeth, may improve eating habits and oral hygiene measures.
Considerable alterations in populations. In Table I of Ast et al. (1950) the number of permanent teeth erupted is shown. The numbers given for Newburgh in the examination of 1944-5 for the three age groups six to seven, eight to nine and ten to twelve years are respectively 3,579, 7,937 and 24,586. However, by adding in Table I of Ast and Chase (1953), the number of erupted teeth - for the same age groups, and in the same examination - are 5,379, 10,033 and 27,186. It was stated in the former report that "we are considering only those children age 6-12 who were in the original base study and who have had each successive examination until they reach age 12." It therefore appears that to meet those requirements, it was necessary to exclude, for the three age groups, 33 per cent, 21 per cent and 10 per cent of the number of erupted teeth, and, presumably, similar percentages of children. A like situation was seen in regard to the Kingston data, the percentages of teeth excluded being 24, 26 and 12. After only four years, it was apparently necessary to omit these large proportions of the data in order to consider only those children who were "continuous residents", no other explanation being evident for the different numbers of erupted teeth that were stated in the two papers. Although the population of Newburgh may have been "relatively stable" when compared with some unnamed population, it is obvious that the number of migrants was so great that they should have been excluded from the study.
Data of migrants excluded only in Newburgh. The necessity for excluding the data of migrants was later realized, and the method of including in the study all the children present in the schools on the day of the examination - although it was continued in Kingston - was abandoned in Newburgh. Ast et al. (1955) stated: "Based on residence histories, the Newburgh study group was limited to those who had used Newburgh water since the introduction of sodium fluoride on May 2, 1945." In the final report, also, only those children who had lived continuously in Newburgh were included, but "All the Kingston children examined are included in this report" (Ast et al., 1956).
Alterations in sample size. The sample size and the age distribution of the children were altered during the course of this study. The data included in the first three dental reports were obtained from the "entire elementary school populations" (Ast and Chase, 1953), except that in some years some of the children were excluded in Newburgh on residential grounds, and that in 1951-2, owing to the loss of an examiner, only half of the children in each city were included. However, in the 1953-4 series the age range was restricted to six to ten years, and the number of children examined was only a small fraction of those inspected in the same age groups during other examinations. Ast et al. (1956) said that the preceding report "dealt with rather small groups of children (about 375 children ages six to ten in each city), and there was considerable difference in age distribution."
Sampling by selection. The method of sampling used in the 1953-4 examination must be considered to be unorthodox, and was described by Ast et al. (1955) in these words:
"The current series includes a limited number of schools which were chosen because of the availability of X-ray facilities. From previous data on DMF rates by school, it was determined that the selected Kingston school had a caries rate which was among the lowest in the city, while the rates for the three Newburgh schools were distributed through the range of rates for that city. This has the effect of minimizing the difference in the DMF rates between the two cities."
A decrease in the "per cent difference". In the final report (Ast et al., 1956, Table 1) the "per cent difference" between the DMF rate per 100 erupted teeth of children aged six to nine years in Newburgh and Kingston was given as 56.7. This is a smaller difference than any of those shown for the ages six, seven, eight and nine years (74.7, 68.3, 58.1 and 66.0 respectively), in the previous (1955) report, despite the fact that it was stated in that report, that the sampling method used had minimized the difference between the DMF rates in the two cities. A trial period of ten to twelve years was suggested by Ast (1943), and was mentioned in the authors' first report (Ast et al., 1950). In view of the decrease in the "per cent difference" between the test and the control cities, which was revealed in the final report, it is unfortunate that the trial was stopped as soon as the minimum period proposed by the authors had elapsed.
The Expert Committee on Water Fluoridation of the World Health Organization (1958) stated that "Hundreds of controlled fluoridation programmes are now in operation in many countries. Some have been in progress for the past 12 years, so that conclusions are based on experience." This statement suggests that there is a large amount of experimental evidence in regard to the process of artificial fluoridation. It is very doubtful whether this is the case. If hundreds of fluoridation programmes have been conducted with experimental controls, it is strange, and very unfortunate. that such a large body of data has not been published; for, except in the cases of the trials which have been considered here, published data concerning fluoridation trials are very meagre. It would seem, therefore, that the Expert Committee did not use the term "controlled" in its experimental sense, but in that of regulated measurement of the fluoride salt, such as in its statement that "The precision of fluoride application should be carefully controlled."
The United Kingdom Mission (1953) which visited North America in 1952, in its report referred to "the Fluoridation Studies", and enumerated only six study centres; and Jenkins (1955) mentioned "the six study centres on the American continent". In addition to the four test cities which have been considered, the Mission referred to Sheboygan, Wisconsin, and to Marshall, Texas. No control city for Sheboygan was mentioned; and the Mission stated that in the latter study "The neighbouring town of Jacksonville with a fluoride-free water supply was selected as control, but although caries experience in the two areas was compared after 2 1/2 years of fluoridation, the most valuable basis for comparison is the baseline data of Marshall itself." The Mission quoted two unpublished reports as the source of its information in regard to the latter study. An indication of the minor importance of these two trials is the fact that in the 240-page report of the New Zealand Commission of Inquiry (1957) no data from them were presented, the former city being mentioned twice and the latter only once. Furthermore, the directors of these studies, Doctors F A. Bull and E. Taylor, were not named in the report, and the extensive bibliography did not include any papers published by them.
The crucial importance, even at the present time, of the trials conducted in Newburgh, Grand Rapids, Brantford and Evanston was demonstraled by the report made in 1957 by the New Zealand Commission, the hearings of which did not conclude until April of that year, and by the report of the Expert Committee of the W.H.O. (1958), which met during August 1957. The only evidence mentioned by the Commission with regard to the dental results of the addition of fluorides to water supplies was that obtained in those four cities. The Expert Committee referred to only the first three of those cities in the few lines of its report which mentioned dental results of fluoridation.
In discussing the general design used in fluoridation studies, the United Kingdom Mission (1953) said:
"In a fluoridation study, two nearby towns, comparable in all respects, are chosen, both having an almost fluoride-free domestic water supply, preferably from the same source. The water of one town is fluoridated while that of the other remains untreated, this town serving as the control. Before fluoridation is started the teeth of the children in both towns are examined in detail to ascertain if caries experience is similar and to determine its prevalence in the various age groups. Further examinations are carried out at yearly intervals and the dental condition of the children in the fluoridated town is compared with that of similar groups in the control town. The prefluoridation data also serve as a basis for comparison. The caries incidence may also be compared with that in a town where a similar concentration of fluoride occurs in the water naturally. In practice it is often difficult to obtain all these conditions and in some studies there is no independent control."
The term "comparable in all respects" describes a theoretical ideal for a test and a control town rather than a practical possibility. In regard to the other matters mentioned in the design these studies exhibited numerous deficiencies. No control was employed in the City Council study in Brantford, and the Grand Rapids study lost its control in 1951 as a result of the fluoridation of the Muskegon water supply. In the extremely important matter of the water supplies, both the source and the composition of the Newburgh water is considerably different from that of Kingston. Further examinations were "carried out at yearly intervals" only in Grand Rapids-Muskegon, and in Newburgh-Kingston until 1952; if yearly examinations were made in the latter study after that year, the results for all years were not published. In the Evanston trial, only two examinations were made in the control city, and few data from it have been published; in the test city only one age group was examined each year. In Brantford and in Evanston, and in the first and the last dental reports from the Newburgh-Kingston study, data from children of different yearly ages were added, thus introducing the possibility of "weighting". In some instances, at least, the degree of "weighting" found indicated that the comparisons were not being made between similar groups in the test and the control cities. No pre-fluoridation data were gathered in Brantford by the Canadian Department of National Health and Welfare, for that study was not commenced until over two and a half years after the fluoridation of the city water supply.
In all of the studies that have been considered, it has been seen that fluoridation of the water supply of the test city was initiated before the initial caries rates in the control city were known. This late examination of the control cities, on first thought, may not seem to be of much consequence. However, it means that, in all of these studies, a matter of fundamental importance was disregarded-it could not have been established that the children of similar ages in the test and the control cities, prior to the commencement of the experiment, had reasonably comparable caries attack rates. Therefore, the statement of the United Kingdom Mission (1953) that "Before fluoridation is started the teeth of the children in both towns are examined in detail to ascertain if caries experience is similar and to determine its prevalence in the various age groups" appears to have been based on assumptions only.
Caries attack rates may be expressed as decayed, missing and filled teeth per 100 erupted teeth, or expressed as per 100 children or per child. The former method was preferred by the authors of the Newburgh trial "because individual teeth may be subjected independently to the hazard of caries" (Ast et al., 1956). In the Evanston study, the rate per 100 erupted teeth was given, but, curiously, only for children aged twelve to fourteen years. All other cases (the younger children in Evanston, and the other studies) in which the caries attack rates per 100 children or per child were given are based on the assumption, unsupported by published evidence, that in each age group the mean numbers of each category of erupted teeth per 100 children is very similar in the test and the control cities, and that little variation occurs from year to year. If this is not the case, comparisons between the rates prevalent in the test and the control cities, and those seen in different years, are not valid. Feltman (1956) gave fluorides in tablet form td-pregnant women and young children, and reported that "Many children in the study group showed a marked delay in the eruption of the deciduous teeth. This delay is in some instances a cause for alarm by the parents. The second incisor, second molars, and cuspids are the most frequently delayed, in many cases by as much as a year from the accepted average eruption dates." Of course, if fluoridation results in the eruption rate of teeth being retarded, a decrease in caries experience would be expected due to the shorter time of exposure of the teeth to the risk of caries. It will be recalled that data were published in the Evanston study which were compatible with a continuous and marked decline in the rate of eruption of first permanent molars during the first four to five years of fluoridation, but that further comparisons could not be made because this type of data was not published for younger children in later reports.
In order to decrease the chance of misinterpretation, extensive use has been made of direct quotation from the original reports, and to avoid unnecessary repetition, consideration of the comments made on the results reported from these control cities, apart from those made by the authors of these studies, has been restricted to the statements of only a few writers.
It is felt that it is not necessary to discuss further the matters which have been noted above, for they are self-explanatory. It has been shown that the reports of the controls used in these fluoridation trials contain arithmetical and statistical errors, and that results and relevant data were omitted. Also, misleading statements were made which denied, ignored, or underrated the unexplained changes in caries attack rates which took place in the control cities, and which suggested that the pre-fluoridation data from the test cities, and those obtained during the basic examinations in control ones, were more closely comparable than was the case. Jean R. Forrest, the Senior Dental Officer, Ministry of Health, who was a member of the United Kingdom Mission and of the Expert Committee on Water Fluoridation of the World Health Organization, in 1957 contrasted "the emotional type of opposition" to fluoridation, to "the precise correct statements of scientists," However, the situation which has been encountered is more aptly described by the words of Wade Hampton Frost "an outstanding American epidemiologist" (Bews, 1951). More than thirty years ago Frost (1925) said:
"It is frequently easy to exhibit some figures which, though not really to the point, will nevertheless serve to impress an uncritical public, and the temptation may be great to give them, at least by implication, an unduly favourable interpretation. It is more difficult and more tedious to present the full argument, based on all the facts, and it is perhaps a little humiliating to admit that the statistical evidence is deficient because we have failed to collect it; but to do this is not only more scientific, it is in the end more convincing, and after all there is no free choice, because it is the only honest method, whether it be convenient or not. Finally, it is the only way of progress, for the first step towards collecting better evidence is to recognize the deficiencies of that which is at hand."
More than eleven years after the initiation of the last of these five trials, the deficiencies of their controls still remain unrecognized. The endorsements of fluoridation by medical and dental associations, by public health authorities, and even the recent one by the Expert Committee on Water Fluoridation of the World Health Organization (1958), appear to have been based mainly on the opinions of the authors and of others. Indeed, in the report of that Committee, under the heading "Results of fluoridation", instead of results being considered, comment was confined to: "Reports of the results after 10 years of controlled fluoridation in three cities". Examinations of the data obtained in these trials, which have been published by other endorsing bodies, are also inadequate or absent. It is an understatement to term this failure regrettable.
In 1951, Appleton stated that in any future fluoridation trial: "The experiment should be genuine, and not one in name only. In designing such an experiment, a careful and competent analysis of those now in progress should first be made, in order to see how they might be improved or extended."
In the early part of this paper some basic experimental considerations were mentioned. In cities in which it is intended to compare the caries attack rates of the children in a fluoridation trial, the three main factors which should be as closely comparable as is practically convenient are the composition of the water supply, the climate and the dental caries attack rates. Four trials having "fluoride-free" control cities have been considered. The composition of Newburgh's water supply is considerably different from that of its control city. There were gross differences between Evanston and its control city regarding the initial caries attack rates in the younger children. In the Department of National Health and Welfare study in Brantford, as the first examination was made over two and a half years after the commencement of fluoridation, it cannot be known what the pre-fluoridation rates in that city would have been, if assessed by those examiners; thus, it cannot be determined how closely the (1945) Brantford rates resembled those of Sarnia. In the Grand Rapids study, the fluoridation of the water supply of Muskegon in 1951 severely limited its usefulness as the control city.
In each trial both the test city and its control were selected. For instance, "Oak Park graciously offered to serve as the control community" for Evanston (Blayney and Tucker, 1948). Two cities which had agreed to participate in the experiment, after having been found suitable for comparison, should have been allotted at random to be test or control. It will be recalled that in at least two of the cities selected as test ones, Evanston and Brantford, "dental care was outstandingly good" (United Kingdom Mission, 1953)
It has been pointed out in Part One that in all these trials no attempt was made to devise a randomization procedure in the clinical examinations which would have eliminated examiner bias, nor were estimates made of examiner variability.
Two statements made by authors of these studies may be recalled. In 1950 Hill et al. said: "It is to be expected that the rate of caries in all teeth varies from year to year due to chance. A significant reduction of caries prevalence can therefore be assumed to exist only when the statistical analysis of the data provides almost absolute certainty that the observed differences are not due to chance." However, as was mentioned in Part One this very important matter of random variation has been ignored in all these studies. Blayney and Tucker (1948) stated that: "A study of this nature must have an adequate control." It has been seen that the controls used in these trials cannot be considered to be adequate.
It would appear that these shortcomings have not been recognized, for those who conducted these studies, and other writers, have expressed their satisfaction with the methods used. For instance, Ast and Chase, the authors of the 1953 report on the Newburgh-Kingston study, referred to "the carefully controlled studies such as the Newburgh-Kingston, Grand Rapids-Muskegon, and the Evanston studies"; and Mather (1957) said: "This study at Brantford was most carefully set up and has been under the strictest control."
Approval of the methods used in these studies was also expressed by the New Zealand Commission of Inquiry (1957), for it considered that: "All these investigations" in Brantford, Newburgh and Grand Rapids "were designed and executed with great thoroughness." The Commission also said: "We have examined the statistical evidence brought forward by the advocates of fluoridation, and the conclusions they have drawn from that material ... We have found nothing to invalidate the statistics or cast doubt on their reliability." It will be realized that many of the deficiencies of these studies can be noted only when different reports >from the same study are compared. It seems that the Commission was handicapped in this regard, for although its "List of exhibits produced at public hearings" mentions over 250 items, such as papers, books, charts and letters, it includes only the final report, or the one which was the most recently published at that time, of the numerous reports showing the dental caries attack rates which were published in each of these studies. It would appear that none of the earlier accounts of these trials were shown to the Commission, nor were they mentioned in the "bibliography" of 144 references. None of the reports from the City Health Department trial in Brantford were listed as exhibits. The paper by Brown, Kohli, Macdonald and McLaren (1954a) which is mentioned deals only with gingival results. Although the Commission had the assistance of legal counsel in gathering the evidence, no mention was made of the employment of a statistician to assist its members in evaluating the numerical data.
The Expert Committee on Water Fluoridation of the World Health Organization (1958) also expressed its satisfaction with the methods used in these trials. Out of the hundreds of controlled fluoridation programmes which it stated have been set up, it mentioned only the Newburgh, Grand Rapids and Brantford (City Health Department) studies in the sixteen lines which allotted to the mention, one cannot say consideration, of the results of fluoridation on dental caries prevalence. Presumably these three trials were cited because the Committee considered that they were the most important and reliable studies, and it said that they were "carefully planned and controlled". As this opinion of the Committee was made in referring to the three studies which it cited in mentioning results of fluoridation, it is reasonable to assume, at least in this instance, that the term "controlled" was used in its experimental sense.
If this is the case, the inaccuracy of that statement of the Expert Committee is astonishing, for it will be recalled that, of the three studies which the Committee quoted, the Brantford (City Health Department) study, far from being carefully controlled, was not controlled at all. Furthermore, the control for the Grand Rapids study was abandoned after only six years, at the crucial stage of the trial when the first of the permanent teeth were erupting in the children of the test city who had ingested fluoridated water throughout their lives. Therefore the control , was abandoned before any assessment of caries activity in those teeth could be made. In regard to the remaining study mentioned by the Expert Committee, the Newburgh trial, after the unexplained decreases in the DF rates for deciduous teeth, which were shown as having occurred in Kingston, the control city, between 19456 and 1949, no further caries rates for deciduous teeth were published. Also, the erratic changes which were made in the methods used in this trial are not consistent with careful planning, nor is the choice as the test and control areas of two cities with water supplies which were of considerably different composition.
It has been acknowledged for many years that one of the fundamental procedures in planning an experiment is the establishment of a statistical design for the procedures before work is commenced. The deficiencies in the basic statistical requirements of a good experimental design are only too obvious in all these studies. Therefore, it is surprising that the Expert Committee did not point out these deficiencies, but, on the contrary, described the three studies which it mentioned as carefully planned ones. The importance of these matters is emphasized by the authoritative statement of Fisher (1951) that: "If the design of an experiment is faulty, any method of interpretation which makes it out to be decisive must be faulty too."
This investigation of reports of these fluoridation trials was instituted when a preliminary examination of the methods used revealed disturbing facts, and solely because it was felt that, as Appleton (195 1) expressed it: "Professionals and specialists have the duty of insisting upon a scientific demonstration of a high probability that a proposed method will be useful and safe, before it is recommended for general adoption. The maintenance of this attitude is of paramount importance."
The deficiencies of these trials not having been recognized, many cities have already fluoridated their water supplies on advice which is based largely on the results that have been considered. It is, therefore, an important and urgent matter that a more accurate assessment of the efficacy of this process should be obtained, but, unfortunately, it appears that little long-term experimental evidence is available. Therefore, despite the limitations imposed by the methods used in these studies, consideration should be given to a careful and competent examination of the whole of the original data obtained in them. The findings resulting from such an examination would be of assistance in designing future fluoridation trials, and would provide a far more adequate assessment of the results reported from these studies than it is possible to obtain from an examination of the very limited data that have been published.
At least until such a report is available for examination, it would be wise to maintain an open mind in regard to the efficacy of artificial fluoridation.
SUMMARY1. Endorsements of the process of the mechanical fluoridation of public water supplies rely mainly on five experimental trials.
2. The controls used in these studies are considered.
3. The reliability of the results reported is affected by:
(a) odd experimental and statistical methods;
(b) failure to consider random variation and examiner variability, and to eliminate examiner bias;
(c) omission of relevant data;
(d) arithmetical errors;
(e) misleading comments.
4. Controls were either doubtful or inadequate.
5. No control was employed in one trial.
6. The published data do not justify the statement that caries rates remained the same in control cities.
7. The sound basis on which the efficacy of a public health measure must be assessed is not provided by these five crucial trials.