Ori g inal Article PROBLEMS IN ESTIMATING THE BURDEN OF PULMONARY TUBERCULOSIS IN INDIA: A REVIEW M.S. Krishnamurthy* (Received on 5.2.2001, Revised version received on 2.7.2001, Accepted on 9.7.2001) Summary :It is generally agreed that there is need for developing an acceptable estimate of the tuberculosis problem in India. Estimates obtained by the National Sample Survey have been found to be deficient for the purpose of enabling rational allocation of resources and evaluating efficiency of anti-tuberculosis programmes. The WHO commissioned a study on the subject and subsequently published a review of the global burden of tuberculosis, in which the Indian situation was also considered. The Government of India also formulated a Committee in 2000 to address this problem. The author reviews the estimates so far obtained. Burden of pulmonary tuberculosis - the current corresponding to the population escalation taking estimates place in the country. For developing an appropriate strategy for combating tuberculosis in the country, it is necessary to obtain a precise estimate of the disease in the community. The process of estimating it for as large and diverse a country as India is not only difficult, but is also expensive and time consuming. It is common knowledge that the nation wide sample survey carried out by the Indian Council of Medical Research (ICMR) in 1955-58 (NSS), followed by several similar surveys conducted at the National Tuberculosis Institute, Bangalore (NTI), New Delhi Tuberculosis Centre and the Tuberculosis Research Centre, Chennai (TRC) had provided the basis for estimating the prevalence of tuberculosis in India. On an average, the same was taken to be 4.0 and 16.0 per thousand for bacteriological (C+) and radiologically active tuberculosis cases (X+) respectively. Although the tuberculosis situation and the prevalence rates were considered to be static in time, as per the findings of several of the studies carried out in India, the absolute number of cases in the country had to be constantly revised upwards, The case for revision of estimates The global resurgence of tuberculosis, especially in the context of Human Immuno- deficiency Virus, has created an unprecedented awareness for tuberculosis in recent times. Several international bodies have, as a result, involved themselves in tuberculosis control with an intensity never evidenced before. The question, however, which has been disturbing the intervention-strategists is how to evaluate changes in tuberculosis situation, following the implementation of the control measures requiring investment in money and manpower. Evidently, some kind of change in the tuberculosis situation could result and this would require measurement. It is, however, open to question how precise the estimate for the disease should actually be to make it amenable to measurement. On the one hand, we are made aware by Chakraborty, in a recent communication 1 , and even in an earlier paper 2 , that he had found the available data in India unsuitable for efficacy and efficiency evaluation exercises. On the other hand, an editorial in the Indian Journal of Tuberculosis (IJT) suggests that ideals (in estimating) * Health Scientist, Bangalore Correspondence: Mr M S Krishnamurthy, 2315, 21st Cross, Banasanhai II Stage, Banglore 560 0070 The Indian Journal of Tuberculosis 194 M.S.KRISIINAMURTIIY need not be pursued, as long as practicable estimates were taken as ideals 3 . It is also understood from the editorial published elsewhere in the same issue of the IJT, that amidst this controversy, the Govt. of India had arranged a 'consultation' at the NTI, Bangalore for resolving some of the issues and providing an estimate of the tuberculosis burden in India 4 . It is presently intended to discuss some of the problems here, in arriving at the estimate of the tuberculosis burden in the country. As has been brought out in an earlier para, the rates of both bacteriological and radiologically positive cases were more or less considered to be unchanged in time so far. In the seventies, the NTI had attempted to refine both the prevalence of bacteriological and radiological cases 5, 6 . This, in the first place, was in recognition of the limitation of the yield of bacteriological cases by the investigation of only two samples of sputum from the X-ray abnormals in a survey. The limitation of X-ray interpretation by two independent X-ray readers, followed by an umpire reader, was also a factor to consider in estimating the radiologically active disease. Instead of investigating only two sputum samples from X-ray abnormals, eight samples were examined in a study by the NTI, in order to arrive at an estimate of the total bacteriological case load in a given community 5 . Further, in another study at the NTI, instead of the conventional X-ray reading technique, an innovative system of interpretation of the radiographic abnormalities along with a series of follow-up X-rays and other examination findings, was relied upon in order to get a correct estimate of the initially radiologically active cases in the community 6 . This technique of X-ray reading was termed as Joint Parallel Reading (JPR). From the former study, it was found that the yield of bacteriologically positive cases was more by 37% in multiple sputum examination method ; and in the latter study the JPR method of reading showed that only 22% of the radiological cases, classified as active cases by conventional method, could actually be confirmed as truly active TB cases. Apart from the JPR method, there is other amply corroborated evidence to support the finding of an over-estimate of radiologically active cases as obtained through the method of conventional X-ray reading. In longitudinal surveys carried out by the NTI, the prevalence of these cases was 10.6 per 1000 in I survey & 6.8, 4.2 & 4.3 in II, III & IV surveys 7 . Further, in the Chengalpattu study, while the observed high prevalence of bacillary cases was at 10.68 per 1000, the prevalence of radiological cases was 4 29 per 1000 8 . The New Delhi Tuberculosis Centre survey also had shown substantial reduction of the X-ray positive case prevalence in later years 9 , ostensibly due to an improved interpretation of X-ray shadows, probably influenced by the results from the NTI study on the fate of X-ray cases diagnosed in a survey (JPR study), available by then. It is strange that these significant findings on the JPR study were not incorporated by the NTI in estimating the burden of disease, even to this day, till Chakraborty in his report to the WHO used them to present a set of new prevalence rates for the country 10,11 . In the seventies, NTI had realised the need to modify the expensive conventional survey technique by adopting alternative cheap investigation procedures. In this pursuit, NTI used 'symptom elicitation' as the initial screening tool, instead of X- rays 12 . In more recent times, Chakraborty et al found that the estimates of prevalence of disease made by the use of either of the screening tools should be the same, if the symptom elicitation was made either by a social investigator or by a senior experienced worker in the field of tuberculosis 13 . These significant findings influenced many of the tuberculosis workers to carry out similar surveys in different areas of the country, beyond the seventies. They carried out these surveys in relatively larger population groups, from which valid statistical estimates of prevalence of disease could be computed. The findings of such surveys, with and without applying the coirection factor as suggested by Gothi et al 12 , and the findings of two major X-ray surveys are given in Table 1. The Indian Journal of Tuberculosis ESTIMATING THE BURDEN OF PULMONARY TUBERCULOSIS IN INDIA 195 Table 1 : Prevalence of tuberculosis in India according to screening tools used 10 Survey Period of survey Population covered Initial screening tool Prevalence of C+ cases (11 respective of smear result) in 15+ age group Prevalence after applying correction factor Significance of the result in relation to NSS at 95% confiden- ce limits (for corrected rate) National Sample Survey(NSS) 1955-58 2,00,429 X-ray 5.4 (5.05-5.69) N.A. Chengalpattu (Tamil Nadu) 1968-71 2,04,624 X-ray 10.8 (10.36-11.26) N.A Wardha (Maharashtra) 1982-88 4,57,136 Symp 2.4 3.6 P<0.05 Raichur (Karnataka) 1988-89 40,496 Symp 8.8 (7.89-9.71) 13.2 (12.87-13.53) P<0.05 Karhal Block Morena District (Madhya Pradesh) 1991-92 11.097 Symp 9.7 (7.87-11.52) 14.6 (13.89-15.21) P<0.05 (Range for 95% confidence limits given in brackets) C+: culture/ smear positive cases Symp: Symptoms It can be seen from Table 1 that the NSS and Chengalpattu surveys were carried out using X- rays as initial screening tool and for the rest of the surveys, the initial tool of investigations was symptom elicitation. For the surveys carried out using symptom elicitation as initial screening, the respective rates of prevalence of disease as observed and as computed after applying the correction factor (as suggested by Gothi et al 12 ) have been furnished in col 5 & 6. The observed prevalence rates, as well as those obtained after applying the correction factors, were found to be statistically different (P<0.5) from those observed in NSS. In the light of significantly different prevalence rates found in different surveys, beyond the seventies, the assumption that the prevalence of disease is uniform a l l through the country, (“distributed ubiquitously” as suggested in Forum under Editor replies 3 ), may not be correct. Unfortunately, and strangely at that, the same hypothesis developed during the NSS, receives support of the scientific community, even to this day 4 in the face of scientifically analysed data pointing otherwise. Stranger still is the fact that the same hypothesis is still being followed even by the NTI for the purpose of monitoring the programme, nationally. ARI studies currently being undertaken by the NTI all over the country, are likely to throw more light on this aspect, and would hopefully resolve the issue. Attempts at redefinition of the problem in recent times As has been brought out in an earlier paragraph, that warning signals in respect of the tuberculosis situation globally have been taken seriously and there are now attempts to work out the global burden of tuberculosis, including India. Inspite of efforts to underplay the need and the method used for revising the estimate for the country 3,4 , there have been two attempts from the WHO Geneva, to analyse the problem and arrive at as precise an estimate as possible, in the Indian context. Under the auspices of the WHO, Chakraborty re-estimated the burden of disease in The Indian Journal of Tuberculosis M.S.KRISHNAMURTHY 196 the country, updating the technique followed heretofore, by incorporating the findings of JPR and multiple sputum investigation studies as explained earlier 10 . This was followed by the work by Christopher Dye et al in 1999, as a part of the global exercise and expressed as a consensus statement by the WHO Geneva 14 . It was against this background, that the Government of India convened an 'Expert Committee' in 2000 and assigned them the task of estimating the burden of TB in the country 4 . The Committee reviewed all the available data, including the estimates made by Chakraborty and Christopher Dye. It identified the various surveys carried out in the country which had followed similar investigation procedures. The average both sexes all ages rates from these surveys were standardised on the basis of the observed prevalence rates in different age and sex groups of each survey, with the projected population structure of 2000. A weightage was given for the size of the population covered in each survey to estimate the burden of disease in the country. Table 2: Prevalence of tuberculosis cases in India (average rate/number) for standard population Individual/ Group PievalenceforC+/X+ cases (per 1000) No. of C+/X+ (in , 000) No. of Sm+ Potential (annual) of case finding under NTP (No.) Ratio of Incidence: Prevalence Conventional Prev of C+ cases @ 4.0 in 5+ age 3265 1632 1,305 1:3 estimate Grou p Prev. of X+ cases @ 16.0 for 5+ 13 , 060 age group (Sm+ cases 50% of C+) Christopher Prev of C+ cases® 5.05 4,851 2,182 1,745 1:27 D y eetal 14 Prev of X+ cases not estimated ) ( 44.9% of C+ ) Chakraborty Prev. of C+ cases @ 6.0 4,897 2,448 1,958 1:3 A.K. 10 Prev of X+ cases @ 3.0 2,448 - Expert Group Prev. C+ cases® 4.8 3,926 2,473 Govt. of India ( 63% of (2000) C+) Chengalpattu 17 Prev. of C+ cases @ 8.75 5,881 3,071 2,352 Tamil Nadu in 1 0+ a g e g rou p (@ 52.23% of ( estimates made ( 1968-70 ) I Surve y C+ cases of based on ( considered p o p ulation that round ) observed rates below as 30% ) in 1/3 of the original sample Prev of C+ cases @ 6.94 4,664 2,876 1,865 selected for in 1 0+ a g e g rou p (@ 6 1.67% the BCG trial) (1984-86) V I I Survey of C+ cases) Population considered: 960,178 ( in thousand) as per Christopher Dye et al 14 C + : Bactenologicallv positive cases (culture / culture and smear positive) X+ : Rad iol ogically active patients Sm+ : Smear positive cases The Indian Jouinal of Tuheiculosis ESTIMATING THE BURDEN OF PULMONARY TUBERCULOSIS IN INDIA 197 Based on the above exercise, one has now access to a fresh set of estimates of the burden of tuberculosis in the country, as of today. It serves to provide one standard average prevalence rate, corrected to represent the current Indian population structure, age and sex-wise. This would eliminate the confusion created by different rates put forward by painstaking work by a number of individuals and agencies. Table 2 provides information on the estimates of prevalence obtained from various sources, compared to that arrived at by the expert group convened by the Government of India (2000). Information is available in the Table on the estimate of prevalence rates followed by the individual / groups and the computed number of culture and smear positive cases for a standard population. Suitability of the newly computed estimate : opening Pandora's box? The estimates given in Table 2 and their suitability for use require to be discussed here in some detail. Since the conventional estimates, which were so far followed by the NTI, had not taken cognizance of the findings of its own studies in respect of additional yield of 37% of the culture positive cases by multiple sputum examination, and the likely fact of a confirmation of only 22% of the radiological cases arrived at by following the conventional method of X-ray reading, the estimate of 32,65,000 bacteriological cases, could be an underestimate. On the other hand, the number of 1,30,60,000 taken as the burden of radiologically active cases is a gross overestimate. To consider 4.0 and 16.0 per 1000 as the prevalence of bacillary and X-ray active cases, respectively, does not seem to be correct, on the available evidence. From Table 1, it can be seen that the prevalence rate of bacillary cases is not uniform all through the country. In support of this, it may be stated that, even under NSS, pockets of high prevalence like Calcutta slums were found. Therefore, how far it is rational to consider an uniform prevalence for the entire country, given its diverse socio economic scenario and the sheer size, is anybody's guess. Besides the differences by space, there are changes in time as well, as seen from the NTI and TRC studies in rural Karnataka and Tamilnadu. NTI, from its own studies had found that the proportion of smear positive to total number of culture positive cases as observed in I, II, III & IV surveys had changed from about 50% to just about 15% in the span of some 22 years 15 . Moreover, there was an observed change in the ARI as well for the given area 16 . The TRC study in rural areas of Tamilnadu has also shown a long-term and sustained decline in the disease rates with time 17 . However, no such change has been observed in the urban population studied in New Delhi 9 . Thus, to consider a uniform prevalence rate for average is just an untenable hypothesis, both for space as well as time for this country. A look at the estimate arrived at by Christopher Dye et al 14 suggests that they had taken into cognisance the findings of most of the surveys wherein X-ray was the initial screening tool. They had chosen to ignore the findings of surveys, wherein the initial screening was 'symptom elicitation'. Disregarding the findings of these surveys had led Dye et al to estimate of prevalence of bacteriological cases @ 5.05/1000, which was possibly less representative in the context of the whole country. Further, the observed proportion between incidence and prevalence cases in longitudinal study of the NTI was 1:3 and this proportion was modified by Dye et al as 1:2.7. The reason for this change requires an explanation. Dye et al further chose to refrain from estimating the prevalence of radiologically active case load, the most likely and an understandable reason for this being the lack of confirmation of such cases, as brought out by JPR study of the NTI and also by TRC studies, as explained above. A look at the estimates of Chakraborty 10,11 the other hand, shows that he has refined the estimates of bacillary as well as radiologically active cases, based on earlier NTI studies on multiple sputum examinations and JPR technique of X-ray reading. It is true that when the estimate of bacteriological cases is increased, with the assumption of investigation of eight sputum samples, there would be considerable decrease in radiological cases, as most of these cases, could be represented as bacteriologically positive cases, with any of the eight sputum sample becoming positive. Thus, the decrease may not appear illogical at all, and is no The Indian Journal of Tuberculosis M.S.KRISHNAMURTHY 198 doubt affected by the use of evaluated data from field studies. However, it could be seen that he had chosen not to consider the fact of changes occurring in age, sex structure of the community. In all this, he had followed the hypothesis of no change in prevalence rates over time, calculating the absolute numbers on the basis of current population size, without taking into account the age-sex changes in the population, affecting the prevalence rate. The higher and lower prevalence rates, found in other surveys are no doubt represented by him as ranges, with an arbitrary midpoint giving the average. The exercise made by the Expert Committee convened by the Government of India (2000) is unique indeed, attempting the standardisation of observed prevalence in different age groups with the projected population of 2000. Weightage was accorded for each of the surveys, based on the populations investigated in each survey for finally arriving at an estimate of both C+ and X+ cases. However, not considering the findings of surveys wherein the initial investigation tool was 'symptom elicitation' made the estimate less representative in the country's context. Further, and more importantly at that, failure to take cognizance of the findings from NTI studies on multiple sputum examination and JPR technique of X-ray reading, as done by Chakraborty, is bound to lower the estimate of bacillary cases and raise the estimate of radiological cases. These appear to be serious flaws in the estimate. Though the proportion of smear positive cases out of the total bacteriologically positive cases was found to vary from 44% to 58% in the first five longtudinal surveys 7 (it was as low as 15.8% in 1984 survey 15 ), by the NTI and about 48% in the TRC study 17 , the expert committee's decision to consider the proportion at 68% does not appear logical. This unreasoned stance makes the estimate of smear positive cases unreasonably high and unacceptable. CONCLUSION With the renewed concern for tuberculosis, globally and in Ind ia, there is an augmented intervention effort for tuberculosis control. In view of this, there is a perceived need for developing an acceptable estimate of the problem, to enable the programme planners to allocate resources, as wel l as to evaluate the efficiency of programme delivery. The available estimate obtained from the NSS has been found deficient for these purposes. The WHO had commissioned a study of the tuberculosis situation in India based on available data, published under its technical reports series 10 . They had followed this up with publication of a review of the global burden of tuberculosis, in which the Indian problem was also estimated 14 . Later still, the Government of India had formulated an expert committee in the year 2000, which had deliberated on the issue and suggested a revised set of prevalence rates for the country. In this paper the above estimates have been reviewed. The conventional rates of C+ and X+ disease at 4.0 and 16.0 per thousand on the average were not acceptable, since they had not tak en into consideration the under-diagnosis due to the number of sputum samples (two) customarily examined in surveys, and the over-diagnosis due to the X-ray reading technique. The revised rates as worked out by Chakraborty in the Report commissioned by the WHO, had made good this deficiency 10 . However, he had not taken into consideration the fact of demographic changes occurring due to age and sex variations in the population with time. Both the reports by Chakraborty as well as by Christopher Dye, as brought out by the WHO, had not taken into consideration the surveys carried out on the basis of symptom elicitation in the population, and thus had missed one of the essential features of the epidemiological situation in India, as shown in Table 1, that is, the prevalence rates were different from area to area and not similar, as was their hypothesis. In giving an average prevalence rate for the country, they could thus be on untenable grounds. However, this deficiency may seem to be offset to an extent by the ranges they had chosen to give for each of the estimates 10,14 , thus admitting the fact of disparate rates from area to area. Even then, it could be a debatable exercise to present an average for such rates, as were found to be lying outside the 95% confidence limits, as in this case. It appears to be as far fetched an exercise as presenting an average rate of pulmonary tuberculosis, which would be globally applicable. To the credit of Chakraborty, it can be said that though he had presented an average rate The Indian Journal of Tuberculosis ESTIMATING THE BURDEN OF PULMONARY TUBERCULOSIS IN INDIA 199 for the country 10 , he also found it indefensible 1 . It appears that he had compromised on the rates in vogue, even as he found them inappropriate. Further, the reason for the ratio of incidence : prevalence, as shown to be 1:2.7 by Dye et al, is not justified on the basis of available information 14 . The rates, as worked out by the Government of India expert group, suffer from the same lacuna of not considering the results of surveys on the basis of symptom elicitation and concluding on an equal average prevalence rate for the entire country. The latter had also erred in disregarding the overwhelming evidence of under-diagnosis of bacteriological cases in the surveys, related to the number of sputum specimens examined, as well as on the over-diagnosis inherent in the X-ray reading technique. Moreover, they had chosen to consider the proportion of smear positive cases to be over 60%, for which there is no support from the available data. Thus, the prevalence rates given above by different groups were different from each other, and each with obvious lacunae in estimating procedures. To be meaningful, these differences need to be resolved, as they are bound to influence the decision of health planners for allocation of funds and for monitoring of the programme. REFERENCES 1. Cliakrahorty AK, Epidemiology of TB ; Some ifs and buts, IndJ Tub, 2000,47, 184 2. Chakraborty AK, TB situation in India: measuring it through time, IndJ. Tub, 1993,40,215 3. Editorial, On estimation of burden of TB in India, Ind J. Tub, 2000, 47, 127. 4. Editor replies, Ind J. Tub, 2000, 47, 186 5. Nair SS et al, Precision of estimates of prevalence of bacteriologically confirmed pulmonary TB in general population,/m/./. Tub, 1976, 23, 152. 6. Gothi GD et al, Interpretation of photofluorograms of active pulmonary TB patients found in epidemiological survey and their five year fate, Ind J Tub, 1974, 21, 90. 7. NTI, Bangalore, TB in a rural population of south India : a five year epidemiological survey, Bull.Wld Hlth Org, 1974, 51, 473. 8. TB Prevention Trial, Madras, Trial of BCG vaccines in south India for TB prevention, Ind J Meet Res., 1980, 72 (suppl), 1. 9. New Delhi TB Centre, Study of epidemiology of TB in an urban population of Delhi ; Report on 30 year follow up, IndJ. Tub, 1999,46, 113. 10. Chakraborty AK, Prevalence and incidence of tuberculosis infection and disease in India, WHO Technical Reports Series, WHO/TB/97/231 11. Chakraborty AK, Global epidemiology of TB in TB Research into the 21 st century, 1998, Pp I -190, Tuberculosis Research Centre, Chennai. 12. Gothi et al, Estimation of prcv. of bacillary TB on the basis of Chest X-ray and / or symptom screening, Ind J Med Res, 1976,64, 1150. 13. Chakraborty AK et al, Prevalence of pulmonary TB in a peri-urban community of Bangalore under various methods of population screening, IndJ. Tub, 1994, 41, 17. 14. Dye, Christopher et al, Global burden of TB; estimated incidence, prevalence and mortality by country, JAMA, Aug 18, 1999,282,677. 15. Chakraborty et al, Prevalence of TB in a rural population of south India by an alternative survey method without prior X-ray screening of population, Tubercle and Lung Dis, 1995,76,20. 16. Chakraborty et al, Tuberculosis situation in a rural population of south India ; 23 year trend, Tubercle and Lung Dis, 1992, 73,213 17. Tuberculosis Research Centre, Chennai, Trends in prevalence and incidence of tuberculosis in south India, Int J Tuber Lung Dis, 2001, 5(2), 142 The Indian Journal of Tuberculosis . the Govt. of India had arranged a 'consultation' at the NTI, Bangalore for resolving some of the issues and providing an estimate of the tuberculosis. are given in Table 1. The Indian Journal of Tuberculosis ESTIMATING THE BURDEN OF PULMONARY TUBERCULOSIS IN INDIA 195 Table 1 : Prevalence of tuberculosis