Tuberculosis in rural South India. A study of possible trends and the potential impact of antituberculosis programmes*

Bull. World Health Organ. |1974 51 263-271Bull. Organ. mond. Sante J 1 2 Tuberculosis in rural South India. A study of possible trends and the potential impact of antituberculosis programmes* HANS T. WAALER,1 G. D. GOTHI,2 G. V. J. BAILY,3 & S. S. NAIR 4 Data on the dynamics of the tuberculosis situation in rural South India, obtained by the National Tuberculosis Institute, Bangalore, were fed into a mathematical model. By this means predictions about the future tuberculosis situation have been made under a wide range of hypothetical assumptions. The relative and the absolute emphases to be given to the specific antituberculosis measures available under Indian conditions are under continuous debate. One reason for this debate may be inade- quate knowledge of the epidemiological situa- tion-e.g., as regards the present level and trend of morbidity and of the risk of infection. Another reason may be differences in interpreting published data on technical efficacies-e.g., the clinical efficacy of treatment under field conditions and the pro- tective effect of BCG vaccination. These uncertain- ties will logically lead to differences of opinion on the epidemiological impact of various programmes. An additional source of disagreement lies in differ- ences in the approach to decision-making. Thus, opinions may differ regarding what is rational; how economic constraints interfere; and the time-pre- ference a values of cases treated or prevented in the future. In 1961 a longitudinal epidemiological study was started by the National Tuberculosis Institute, Ban- galore, with the main object of obtaining informa- tion on the natural course of tuberculosis in rural South India, expressed as values of important vari- * From Det sentrale tuberkuloseregister, P. B. 8155, Oslo-Dep., Norway. 1 Forsker. 2Epidemiologist, National Tuberculosis Institute, Ban- galore, India. 3Tuberculosis Specialist, National Tuberculosis Insti- tute, Bangalore, India. ' Senior Statistical Officer, National Tuberculosis Insti- tute, Bangalore, India. a The concept of time preference is defined and discussed by M. S. Feldstein et al. (1). ables and parameters. It was thought that, by feed- ing such information into an epidemetric model, it would be possible to make predictions about the future situation of tuberculosis in that part of India and that, by adding information on the operational and technical qualities of various programmes, the total epidemiological impact of those programmes might be estimated. In this paper, an attempt is made to define the epidemiological situation in rural South India, to describe the situation by means of an epidemetric model, and to estimate the potential epidemiological impact of a few selected programmes. The con- clusions are purely epidemiological and there is no discussion of recommended programmes. THE MODEL The epidemetric model applied in this paper is that described by Waaler (2). It consists of a sub- grouping of a total population by age and epi- demiological group, with equations for flows be- tween the groups. The model also offers considerable possibilities of evaluating interference by various programmes, which are defined in terms of opera- tional and technical parameters. The final outputs selected are incidence (in both absolute and relative terms) and cumulative future prevalences of disease. The latter term is referred to as " the tuberculosis problem ", in accordance with the definition adopted by the WHO Expert Com- mittee on Tuberculosis (3). This definition includes present and future cases. Since there exists a time preference in the judgement of most people, the 3264 -263 H. T. WAALER ET AL. Table 1. Study population by age and epidemiological group Infected Active cases Total infected Age Non- Protected infectious, Previous group infected < 5 years > 5 years by BCG non- bacterio- cases Total infectious logically No. confirmed 1 2 3 4 5 6 7-8 0-4 151 222 3878 0 0 0 0 0 3878 2.5 155100 5-9 121 645 8092 2532 0 531 0 0 11 155 8.4 132800 10-14 105818 11 706 8972 0 1 278 26 0 21 982 17.2 127800 15-19 59122 8394 12700 0 811 73 0 21 978 27.1 81 100 20-24 54462 7718 21 461 0 847 212 0 30238 35.7 84700 25-29 48676 6 912 28761 0 1 290 361 0 37 324 43.4 86000 30-34 32684 4634 26472 0 1 310 400 0 32816 50.1 65500 35-39 24860 3524 26212 0 1 412 492 0 31 640 56.0 56500 40-44 16 641 2365 22274 0 1 290 430 0 26359 61.3 43000 45-49 15311 2172 25351 0 1 572 494 0 29589 65.9 44900 50-54 10294 1 458 20687 0 1 368 393 0 23906 69.9 34200 55-59 9 116 1 294 21 874 0 1 720 396 0 25 284 73.5 34 400 60-64 3 670 522 10 432 0 945 181 0 12 080 76.7 15 750 65-69 2 690 380 8 991 0 849 150 0 10 370 79.4 13 060 70-74 1 866 265 7338 0 722 119 0 8444 81.9 10310 75-79 1 166 165 5328 0 547 84 0 6124 84.0 7290 80-84 601 85 3184 0 341 49 0 3 659 85.9 4 260 85-89 260 37 1 601 0 178 24 0 1 840 87.6 2100 90-94 99 14 705 0 82 10 0 811 89.1 910 95-99 31 4 251 0 30 4 0 289 90.3 320 Total 660234 63619 255126 0 17123 3898 0 339766 34.0 1 000000 problem is estimated with the various discount rates described and discussed by Waaler & Piot (4). By varying the technical or operational para- meters representing the programmes, the model can be made to produce different incidence patterns over time and age as well as different estimates of the reduction in the tuberculosis problem. Variations in the epidemiological assumptions can also be simu- lated. MATERIAL The report of the longitudinal study referred to above has not yet been published, and in the present paper only preliminary estimates of the parameters have been applied. However, simulated variations in these input values have indicated that the main conclusions based on the use of preliminary esti- mates will not be affected even by the use of more precise input information. The longitudinal study consisted of a baseline survey and three further surveys at intervals of 11/2, 11/2, and 2 years-a total observation period of 5 years. The study was conducted in 119 villages with about 65 000 inhabitants. They were examined by a standard tuberculin test, X-ray, and bacterio- logical examination in the case of an abnormal X- ray. Details of the study have been given by Raj Narain et al. (5). ASSUMPTIONS The assumptions made can be divided into 4 groups: demographic, epidemiological, technical, and operational (programme). 264 TUBERCULOSIS 265 8.106 -+_ 6.10o66r ___ ___ ___ __CONSTANT FERTILITY 4106 - __0 net106 yersa0atentie etiit ats II06 8105 68105 _ _ 4-105 215 0 10 20 30 40 50 60 70 80 90 100 TIME IN YEARS Fig. 1. Development of the total population over the next 100 years at alternative fertility rates. Demographic assumptions For simplification, a population of 1 million is considered. The age distribution (Table 1) is typical of communities with a high fertility rate and rela- tively high mortality rate, 42% of the population being under the age of 15 years (United Kingdom, 1970: 240%). The entry of the newborn into the population is provided for in the model by a sim- plified fertility rate presuming a fixed ratio between the age group 20-34 years and the newborn. The cohort 20-34 years (representing initially 23.62% of the population) is multiplied by a factor (0.67) to give the size of the new 0-4-year age group-158 254 (15.8% of the population). This corresponds to an annual average input of new generations of 3.16%. In the present study we have assumed a constant reduction of the factor 0.67 by 1 % per year until it has reached 50% of its original value (0.5 x 0.67 = 0.335). Thereafter (i.e., after 70 years) it remains constant. The epidemiological implications of this assumption have also been examined. The mortality conditions are indicated by the age- - a: Initial population b:After 45yeors with reduced fertility - ...... c: After 45 years with constant fertility 12-1 0 10 20 30 40 50 60 70 80 90 00 AGE IN YEARS Fig. 2. Total population according to age. specific survival rates (Table 2 a). Thus, over the first 5 years, there are 42 672 deaths and an annual death rate of 0.00853. Fertility minus mortality over 5 years will then be 158254-42672 = 115 582, corresponding to an initial natural increase of the population of 2.3% per year. The development of the total population is given in Appendix table 1 a and Fig. 1. The assumption of a reduction in future fertility, which in a way reflects the goal of current family planning programmes, is seen to have a considerable impact on the size of the population in the long run. Fig. 2 gives the initial age distribution in comparison with the situation after 45 years. Assuming a future reduction of fertility, the age distribution will change considerably, with potential impact on the epidemiological situation. Further demographic assumptions are: excess a Space did not permit the inclusion of Tables 2-6 and Appendix tables 1-7. These tables have been deposited in the WHO Library and single photocopies may be obtained by professionally interested persons on request to: Chief Librarian, World Health Organization, 1211 Geneva 27, Switzerland. 266 H. T. WAALER ET AL. mortality applied to groups of active cases, and fatality among untreated cases (Table 3 a). Epidemiological assumptions The subdivision of the population into epidemio- logical groups is shown in Table 1. It is assumed that, as in the study population of the longitudinal study, no BCG vaccinations have been given initially. This is not so for the entire Indian population. The groups of previous cases are given zero values as there are no satisfactory estimates available. This means that the relapse rates cannot be given separate values, but will be confounded with the group of persons infected for more than 5 years. The break- down rate from this group is thus an average including new cases and relapses. The groups of active cases are included. In the present study, group 6 (infectious cases) consists of confirmed bacillary cases, whereas group 5 is defined as cases diagnosed by X-ray without detection of bacilli. Only the persons in group 6 participate in the model as infectors. The division of the population into infected and noninfected persons is shown in Table 1, and Fig. 3 gives the age-specific prevalence rates of infection. Actual observations of tuberculin sensitivity are used as a basis for the division up to the age of about 30 years. The slower decrease in sensitivity above this age is not supposed to reflect the prevalence of infection, and the figures have been adjusted accord- ingly. The estimation of group 2 (infected for less than 5 years) posed problems as this group was not yet available from the longitudinal study. The risk of infection is assumed to be known from actual obser- vations. In a population with a known fraction of noninfected persons, the number of new infections over the last 5 years can be estimated (assuming constant rates, or applying an average risk, for these 5 years). The new infections are then distributed among the various age groups according to the number of noninfected in the age group and the age- specific pattern of infection, with the lowest rate for age group 0-4 years, increasing by 40% to the age group 5-9 years and again by 70% to older age groups (6). The initial annual age-specific risks of infection are assumed to take the following values: age group 0-4 years, 1% per year; 5-9 years; 1.4%; and 10 years or more, 2.5 %. These figures agree closely with a See footnote on page 265. the observations in the longitudinal study except that in the age group 5-9 years the risk of infection is slightly higher in the model. Fig. 3 indicates close agreement between the ex- pected and the observed prevalence of infection under these assumptions. It should be stressed that in the model the future risk of infection at any point of time is adjusted to the force of infection, which is a function of the prevalence of infectious cases. When a case is detected and successfully treated, the contribution of that case to the force of infection is assumed to be reduced to 1/7. Morbidity rates include transfers from groups 2 and 3 to groups 5 and 6. The 5-year rates of transfer from group 2 to groups 5 and 6 are 0.15 and 0.34, respectively, whereas those from group 3 to groups 5 and 6 are 0.03 and 0.014, respectively. Transfers from groups 2 and 3 to group 6-i.e., the incidence of new cases of infectious tuberculosis-are age- dependent, and the relevant rates are shown in Table 4.a The incidence resulting from these trans- 100-- 90 ESTIMATED aoF 10000 80 / = 70 -O 20_ 30_ 40 5 7 9 60 K'~~ ~~~~....OBSERVED 50 40 30 0 10 20 30 40 50 60 70 80 90 100 AGE IN YEARS Fig. 3. Observed and estimated prevalence rates of tuberculous infection. Initial situation. TUBERCULOSIS 267 fers will show rates corresponding closely to the observations in the longitudinal study. Technical assumptions Modern drug therapy gives cure-rates approach- ing 100% under experimental conditions. Field con- ditions yield lower results owing to practical, and sometimes economic, reasons. In the present study an efficacy of 80% is assumed. This implies that, of 100 cases put on treatment, 80 will become abacil- lary and will remain so until the end of the 5-year period if they survive. The percentage thus repre- sents the sum of the spontaneous and of the clinical effects. Successfully treated cases are still subject to an excess mortality, whereas others are subject to a spontaneous healing rate of 50%. The applied clin- ical efficacy of 80% may be considered as reasonable under the field condition prevailing in India. For BCG vaccination two technical parameters are introduced. The protective effect of BCG is given three values: 30%, 50%, and 80%. The last-men- tioned percentage reflects the standard value. It is possible that, for some reason or other, only a lower protection by BCG is attainable in India. To account for such a possibility, lower values have also been considered. The authors do not thereby express any opinion about the most likely value. The other parameter is related to the duration of protection and expresses the annual reduction of protection. Whereas the precise value of the parameter is not known, the British Medical Research Council trial (7) indicates some reduction. In the present study a uniform annual reduction of 1% is assumed. The technical parameter values applied are also given in Table 3. Operational (programme) assumptions Two programme elements are visualized: a case- finding and treatment programme (CF/T) and a BCG vaccination programme (BCG). Simulations have also been carried out with a combination of the two. The CF/T programme consists of a continuous effort to detect and treat infectious cases. The method of detection is not described in the model. All age groups are eligible and a coverage of 66% is selected. The ability to detect and treat 66% of all cases existing at a given point of time reflects in fact a very intensive programme. Doubts might be raised a See footnote on page 265. whether such efforts are feasible in India. Therefore, simulations with a lower coverage of 20% were also carried out. The BCG programme defined operates with cover- ages of 66% and 30% but is limited to the age groups 0-20 years. If an eligible person is not covered in one period, he gets a chance in the next. This will mean that at a 66% coverage a cumulative coverage of 96% will be achieved for a cohort over three 5-year periods (i.e., 15 years). Most of the efforts will be put into maintaining BCG coverage in the age group 04 years. COMPUTER SIMULATION OUTPUT Once the initial conditions and parameters (demo- graphic, epidemiological, technical, and operational) have been defined, the future tuberculosis situation is predicted through model simulations in a computer. (The model is programmed in ALGOL-60 and is run by a UNIVAC 1109 at Computas Ltd, Oslo.) The incidence (5-year rates per 1000) under condi- tions of non-interference -i.e., without any active programme-is given in Fig. 4 (curve 0-0) and Appendix table 2.a As the model is deterministic, no expression for the precision is available. It is clear, however, that the projections for the latter half of the 100-year period are less precise. Whereas the absolute number of new cases in- creases considerably, the incidence rates do not warrant firm conclusions about any long-term trend. At first, there might be a downward tendency. The above-mentioned non-interference situation is altered by introducing programme parameters for case-finding and treatment and for BCG vaccina- tion. BCG vaccination is simulated to have three different protective effects-30%, 50%, and 80%, whereas treatment has only one value for efficacy. All programmes are observed to have a considerable impact (Appendix tables 2-5 a and, for selected combinations, Fig. 4 and 5). After 50 years (i.e., ten 5-year periods) the BCG programme with 80% protection reduces the rate by about 2/3 (70%) whereas, after 25 years, the reduction is seen to be about 1/3 (38 %). The case-finding programme with 66% coverage would also have a considerable impact. Fig. 4 shows that this competes with a mass BCG vaccination programme operating with a pro- tective value of between 30% and 50%. After 25 years, the combined programme (with BCG giving 80% protection) will yield incidence rates of H. T. WAALER ET AL. 0z 0=~ TIME IN YEARS Fig. 4. Predicted 5-year incidence rates of bacillary pulmonary tuberculosis for the next 100 years under various assumptions on the coverage of the programmes. about 1/2 (570%) that in the non-interference situa- tion. Even the BCG vaccination programme with only 300% protection is seen to have a considerable epidemiological impact. Whereas the coverage of 66% reflects levels that are obtainable in practice, actual programmes do not have such a coverage level today. More realistic coverages would be 20% for the case-finding and treatment programme and 30% for BCG vaccina- tion. Fig. 5 gives predicted values of incidence corresponding to these coverages. The combined programme is seen to be able to reduce incidence by 26% after 25 years. The case-finding and treatment programme alone, at the assumed coverage level, will be able to reduce the incidence after 25 years by only 12%. BCG alone gives a reduction of 17% after 25 years. However, Fig. 5 shows very clearly that the impact of BCG is increasing. The trend in incidence observed as a result of CF/T is in general seen to be TIME IN YEARS Fig. 5. Predicted 5-year incidence rates of bacillary pulmonary tuberculosis for the next 100 years under various assumptions on the coverage of the programmes. different from that brought about by BCG. The effect of CF/T will be more immediate, whereas that of BCG will be seen much later. This is in accor- dance with what could be expected on the basis of general knowledge about the underlying mechanism. The simulations referred to above have all been made under the assumption of a gradual reduction in the fertility of the population. If such a reduction does not occur, the population will increase (Fig. 1), as will the absolute number of new cases (Fig. 6). However, the incidence rates show a trend with lower figures (Fig. 7). (Simulations have been carried out only under the assumption of 50% protection by BCG.) Differences in programme impact can also be seen, but they are not important in relation to the total impact. The use of incidence as an indicator of the tuber- culosis situation has certain drawbacks. The most important is that it does not distinguish between 268 TUBERCULOSIS 269 100000- 80000- - - 60000--- 40000 - - - - 20cY-C_OO- CONSTANT FERTILITY 1I000C - ro600C0 __4 REDUCED FERTILITY 4000- - - - - 2000- - - 0 10 20 30 40 50 60 70 80 90 too TIME IN YEARS Fig. 6. Predicted 5-year incidence (absolute numbers) of bacillary pulmonary tuberculosis for the next 100 years. Effect of a changed assumption on the fertility rate. short and long duration of the morbid status. The cumulated future prevalence (i.e., the prevalence of tuberculosis where present and future values are added, thus reducing a time-trend to one single figure reflecting the area under the curve) does not have this drawback. This index is based upon the absolute number of cases and is referred to as " the tuberculosis problem" (P). Reduction in this index, resulting from interference with a programme, is referred to as " the tuberculosis problem reduction " (A P). The present significance of future cases as a part of the actual problem might be reduced-e.g., by a certain discount rate (r). A P-values have been calculated as percentages and are presented in Tables 5 and 6.a This has important implications, as has been pointed out by Waaler & Piot (4). The results are therefore presented here with various r-values. A rate of r = 0 signifies that all cases now and in the a See footnote on page 265. 6 II 4 - -_- ~1- CF/Tcoveroge- BCGcoverage 4 0-0 2A __ 0.8 T _ = T I 0.6 L = 1 °0 lV _ 0. L_6-66 (0.5) ------- Reduced fertility 0 _-Constont fertility 0.08 0.00 _ 0.04 t X 0.020 10 20 30 40 50 60 70 80 90 100 TIME IN YEARS Fig. 7. Predicted 5-year incidence rates of bacillary pulmonary tuberculosis for the next 100 years. Effect of a changed assumption on the fertility rate. future are given the same present value. With larger values of r, the value of a case becomes negligible earlier. The extreme value of r = 16 implies that cases occurring 20 years or more from now are reduced to about 5%-i.e., are almost negligible. The CF/T programme and the BCG programme with 50% protection are able to produce a 69% problem reduction (with r = 0). With increasing r- values, an advantage is given to CF/T over BCG, as would be expected on the basis of the different time patterns demonstrated above. This advantage of CF/T over BCG diminishes as the value of the protective effect of BCG increases. The fact that the A P-values are in general reduced with increasing r- values is of course natural, as any programme takes time to exert its effect. For r = 16, the problem reduction achieved by a CF/T programme is 3 times greater than that resulting from a BCG vaccination programme (protection = 50%). Although constant fertility rates would change future incidence considerably (in both absolute and H. T. WAALER ET AL. relative terms), the A P-values expressed in per- centages remain fairly constant. DISCUSSION The present simulations have been based on pre- liminary estimates from the findings of the longitu- dinal study in the Bangalore district of South India, in which a rural population was examined four times (at intervals of 11/2, 11/2, and 2 years.) Since neither vaccination nor a CF/T programme was carried out in the area, the estimates of the parameters are very suitable for predicting the effect of any type of programme on a non-interference situation. There are, however, important limitations to the inter- pretation of the results. Most of the flow-rates considered are presumed to be time-constant in the model simulations supposed to reflect basic rela- tionships between the bacilli and the host. This is somewhat unrealistic if marked changes take place in them. Thus, changes in the standard of living may affect the transfer rates. However, the quantitative relation between the standard of living, infection and morbidity, and demography are not very well known. Consequently the results presented should not be regarded as the best over-all estimates but rather as estimates of potential trends and potential effects under unchanged conditions. Problem reduction will be higher than is reported here if improvements in the standard of living are expected during the com- ing years. In spite of the foregoing arguments, can one, on the basis of the predictions from the model, say anything today about the most likely trend of tuberculosis in India? Whereas the non-interference trend shows an increase in the absolute number of new cases, incidence rates appear to remain con- stant. However, as millions of vaccinations have been carried out and CF/T programmes have been gradually implemented since 1962 in different parts of India, there is reason to believe that the incidence of tuberculosis in that country is on its way down. However, the actual trend is of limited value in itself. What is important is the demonstration of the considerable potential impact of various realistic antituberculosis programmes. The problem-reducing capacity of a CF/T pro- gramme and of a BCG programme could be con- siderable in India if coverages of the order of 66% could be achieved for either. Even a BCG pro- gramme assuming a low-30 %/-protective effect for BCG is able to produce a marked effect. A combined programme (CF/T + BCG) with 66% coverage for both and a 80% protective effect for BCG will bring down the incidence over 25 years to 50% of what it would have been without a programme. It might be mentioned that the 25-year period 1945-1970 showed a reduction of the incidence of tuberculosis in Norway to 25% of what it would have been had the prewar downward trend merely continued. The immediate postwar years in Norway were character- ized by intensive activities in tuberculosis control. Selective X-ray screening of the total adult popula- tion was started and is still working. Mass BCG vaccination of the population above 15 years of age was begun and was followed by a policy of mass vaccination of school-leavers (coverage: 95 %). The benefits of new drugs have been fully exploited. Moreover, a marked increase in the standard of living has taken place. That the programme in Norway had a greater impact than the above- mentioned programmes in India can be attributed to the intensive antituberculosis activities and im- provements in the standard of living. In developing countries it is not always possible to maintain a 66% coverage for BCG programmes, let alone CF/T programmes. Even a BCG programme with a coverage of 30% of vaccinations and a protective effect of 50% for BCG may have a fairly strong impact. However, if a CF/T programme, even with a coverage as low as 20% is added, the effect is considerable. On the other hand, a CF/T programme alone, with a coverage of only 20%, has very little impact-only a 12% reduction in incidence after 25 years. The importance of economic and other constraints in the actual formulation of programmes has not been dealt with in this report. The apparent balance of efficacy between the two programmes with 66% coverages is, of course, only epidemiological. Even without any profound analysis, there can be no doubt that a BCG programme will require far fewer resources than a CF/T programme, and the problem reduction (A P) per resource unit will give the BCG programme a marked advantage. However, the optimum balance between these can be achieved only after a careful study of costs and judicious selection of the social time-preference value to be applied. 270 TUBERCULOSIS 271 RtSUMI, LA TUBERCULOSE DANS DES RIEGIONS RURALES DE L'INDE MERIDIONALE: tTUDE DES TENDANCES POSSIBLES ET DE L'EFFICACIt POTENTIELLE DES PROGRAMMES ANTITUBERCULEUX On a utilise un modele mathematique afin d'analyser les donn6es 6pidemiologiques recueillies au cours d'une serie d'enquetes sur la tuberculose en milieu rural dans un district de l'Inde m6ridionale. La methode foumit des estimations sur les tendances naturelles de la maladie ainsi que sur les resultats a attendre des divers programmes antituberculeux (vacci- nation par le BCG, depistage et traitement) exprim6s en termes de couverture et d'efficacit6 technique. Dans les regions de l'Inde ott aucun programme antituber- culeux specifique n'est actuellement mis en ceuvre, il est probable que les taux d'incidence ne se modifieront pas beaucoup a I'avenir, meme si les chiffres absolus sont en augmentation. Dans la plus grande partie de l'Inde, des mesures sp6 cifiques de lutte antituberculeuse sont pour le moment appliqu&es. Les simulations r6alis6es grace au modele permettent d'esp6rer dans ces r6gions une baisse des taux d'incidence. L'ampleur des resultats dependra pour une grande part de la couverture des programmes et de leur efficacit6 technique. Les simulations semblent indiquer que meme des programmes de vaccination par le BCG ne couvrant que 30% de la population, avec un taux de protection limit6 a 50%, auront une influence considerable sur l'incidence de la tuberculose. REFERENCES 1. FELDSTEIN, M. S. ET AL. Resource allocation model for public health planning: A case study of tuber- culosis control. Geneva, World Health Organization, 1973 (Supplement to Bulletin of the World Health Organization, vol. 48), p. 68. 2. WAALER, H. T. American review of respiratory dis- eases, 98(4): 591 (1968). 3. WHO EXPERT COMMIrEE ON TUBERCULOSIS. Eighth report. Geneva, 1964 (WHO Technical report series, No. 290). 4. WAALuR, H. T. & PIOT, M. A. Bulletin of the World Health Organization, 43: 1 (1970). 5. RAJ NARAIN ET AL. Bulletin of the World Health Organization, 39: 701 (1968). 6. WAALER, H. T. ET AL. (1975) (in press). 7. Great Britain, Medical Research Council. British medicaljournal, 1: 973 (1963).