The distribution and implications of BCG scars in northern Malawi.

Research!Recherche The distribution and implications of BCG scars in northern Malawi P.E.M. Fine,' J.M. Ponnighaus,2 & N. Maine3 Reported are data on the BCG scar status of more than 112000 individuals who were surveyed in Karonga District, northern Malawi, between 1979 and 1984. The age and sex patterns of apparent BCG scars reflect the history of BCG vaccination activities in the district. Repeated independent examinations of large numbers of people revealed that the proportions remaining with the same observed scar status among those initially classified as being scar "positive" or scar "negative" were each approximately 90%. The repeatability of positive scar reading was lower among children and older adults than among young adults aged 15-24 years, and blind follow-up of children known to have been vaccinated as infants in child health clinics indicated that less than 60% had a detectable scar 3 years after receiving the vaccine. "Negative" repeatability increased consistently with age. The implications of these findings for estimating BCG vaccine uptake and for assessing its efficacy in case-control and cohort studies are discussed. The finding that BCG scars may be difficult to read suggests there is a danger of observer bias that could lead to distortion-in particular, to overestimates of vaccine efficacy. Although BCG vaccines are among the most widely used they are also among the most controversial, since the protection they impart against tuberculosis and leprosy varies widely between different popu- lations for reasons that are not understood (1). There is considerable current interest both in increasing the uptake of BCG vaccines in many countries, as part of WHO's global Expanded Programme on Immuni- zation (EPI), and in understanding the impact of past and current immunization campaigns. Report- ing post-BCG vaccination scars plays an important role in each of these efforts. Intradermal injection of live BCG vaccine nor- mally leads to development of a local ulcer, and this often leaves a scar after healing. The scar has been described as ". . . a round, slightly depressed area with irregular edges, 4 to 7 mm in diameter. Occasionally it is raised a few millimeters above the 1 Reader, Department of Tropical Hygiene, London School of Hygiene and Tropical Medicine, Keppel Street, London WC1E 7HT, England. Requests for reprints should be sent to this author. 2 Lepra Evaluation Project, Chilumba, Karonga District, Malawi. 3 Research Fellow, Department of Tropical Hygiene, London School of Hygiene and Tropical Medicine, London, England. skin as a result of fibrous tissue formation and is hard to the touch (hypertrophic scar) or (especially after a local abscess) healing may be accompanied by fibrous cicatrization (retracted scar). Rarely it grows upwards and outwards to become a mushroom-like growth with overhanging margins (keloid scar)" (2). The presence of such a scar on an appropriate area of the body has been considered characteristic enough for it to be used as evidence of prior BCG vaccination. Tuberculin surveys carried out to assess incidence rates of infection with the tubercle bacillus routinely exclude individuals with BCG scars in order to minimize confounding by BCG-derived nonspecific tuberculin sensitivity. Some vaccination schedules recommend that BCG be given with the first or second dose of diphtheria-pertussis-tetanus (DPT) vaccine so that the scar can be used as a marker at subsequent visits to the vaccination clinic; and some countries, e.g., Poland, recommend that any child without a BCG scar should be given a dose of the vaccine at the next opportunity (3). Simi- larly, information on BCG scars is often collected by EPI surveys in order to assess vaccine uptake (4). The presence of such scars has also been used to validate vaccine codes in randomized controlled trials (5) and as evidence of prior vaccination in Bulletin of the World Health Organization, 67 (1): 35-42 (1989) © World Health Organization 1989 35 P.E.M. Fine et al. order to evaluate the protective efficacy of ongoing BCG vaccination programmes against tuberculosis and leprosy in case-control (6-9), cohort (9), or household contact (10, 11) studies. In all of the above-mentioned uses it is assumed that the BCG scars are sensitive and specific indica- tors of prior vaccination. The reported sensitivity of scar reading (the proportion of vaccinated individ- uals who develop a recognizable scar) has been reported to vary from 98.9% among recipients of 0.1 mg BCG in a vaccine trial in south India (Chingleput), as assessed 4 years afterwards (5), to 60% among Swedish children 14 years after having been vaccinated at birth (12). The propensity to develop a recognizable scar is probably, in part, a function of the type, dose, and method of injecting the vaccine (13-15). The specificity of scar reading (the proportion of individuals who have no evidence of a BCG-like scar among those who have never been vaccinated) is, however, more problematic to measure, since it is difficult to confirm a negative vaccination history. None the less, it has been noted that traumatic and decorative scars in some popu- lations may be mistaken for BCG scars and hence lead to erroneously high estimates of BCG uptake (16). We report here the results of an investigation of the epidemiological pattern and implications of the distribution of BCG scars, based on data collected in the Lepra Evaluation Project (LEP), a longitudinal epidemiological study of leprosy and tuberculosis in Karonga District, northern Malawi. Methods The methods used in the LEP have been described in detail elsewhere (17). In this context we note that more than 112000 individuals, virtually the entire population of Karonga District, were interviewed and examined between 1979 and 1984. The data col- lected for each interviewee included personal details such as age, sex, schooling, and history of persistent coughs. Physical examinations were carried out by trained paramedical workers (called leprosy control assistants) according to a strict protocol. The first step in each examination was to scrutinize the right deltoid area for evidence of a BCG scar. The results were coded as "positive" (P), "negative" (N), or "doubtful" (D)-use of the doubtful category being encouraged if there was uncertainty as to whether a mark on an individual's arm was indeed a BCG scar. In addition to the P, N, and D categories, a fourth category ("unknown" (U)) was used for individuals whose scar status was not recorded (because they refused to be examined or the field staff omitted to inspect or record the information, or because of a key-punch error). In order to minimize possible bias in obtaining scar data, an effort was made to ensure that the control assistants did not have access to information that might prejudice their judgement. Thus, data on tuberculosis, e.g., cough history, were collected by independent interviewers and recorded on separate forms from those used by the leprosy control assist- ants. Furthermore, as a matter of policy, the poten- tial implications of BCG vaccination for leprosy were never discussed with the staff, with the result that the assistants, who were trained to diagnose leprosy, were not aware of the effect that BCG vacci- nation might have on this disease. Many individuals were examined more than once during the course of the 1979-84 survey, because they either were leprosy suspects or had been included in one or another special investigation or because they had moved from an area covered early to one covered later in the survey. Further- more, a second survey was begun in Karonga Dis- trict in January 1986. This included a complete re-examination, including an assessment of BCG scar status and measurement of the scar diameter, of all individuals still resident in the area. Records from previous examinations have never been available to the control assistants in the field, and so each exami- nation has been independent of all others. As a special exercise, project staff were assigned to attend each vaccination session at the local child health clinic in Chilumba, a large village in the southern part of the district, from 1982 to 1984, where they recorded the name, sex, date of birth, parents' names, and address of each child who actually received BCG vaccine. This information was then compared with the results of independent examinations of these children carried out during the course of the survey. History ofBCG use In the study population BCG vaccine was first used in Karonga District in 1974 by mobile teams during an attempted mass- vaccination campaign, without prior tuberculin testing, of all inhabitants under approximately 15 years of age. The teams visited all the schools in the district, and hence the vaccine was allocated prefer- entially to the 40-50% of children who were enrolled in schools at that time. After this initial phase, which lasted for 3 years, the responsibility for BCG vacci- nation was handed over to the child health services, which have subsequently made the vaccine available to infants during their first year of life. In so far as this is concerned, the district is typical of many areas in the developing world. BCG vaccination in the district has consistently been by intradermal injection into the deltoid region 36 Epidemiology of BCG scars Fig. 1. Proportions of Individuals recorded as BCG scar "positive", by age, sex, and schooling status, Karonga District, northern Malawi, 1979-84; (a) males, (b) females. No schooling --- 1 -5 years' schooling 5 years' schooling 20 30 40 50 60 70 Age at examination (years) FEMALES - No schooling --- 1 - 5 years' schooling .... > 5 years' schooling 1% and 3% of individuals were classified as having "doubtful" scar status. The scar status of 2.4% of the total population surveyed was "unknown", mainly because of individuals who were not examined. Fig. 1 shows the scar data analysed according to the schooling status, age, and sex of individuals. A clear association is apparent between the prevalence of BCG scars and the number of years of schooling, with peak scar rates (86.2% for males and 82% for females) occurring among 15-19-year-olds who had attended school for more than 5 years. The frequency distributions of scar sizes for dif- ferent age groups, shown separately for lesions that were classified as "positive" or "doubtful", are illus- trated in Fig. 2. The size of the BCG scars increased with age up to 25 years and then decreased; "doubt- ful" scars were in general smaller than "positive" scars. Fig. 3 shows the repeatability of scar reading according to the interval between examinations and by age at first examination. Both the positive and negative repeatability remained quite constant at about 90% for examinations that were separated by up to 6 years. Some trends were apparent, however, depending on the age at the initial examination. For Fig. 2. Frequency distribution of scar size for individuals recorded as (a) BCG scar "positive" or (b) BCG scar "doubtful", by age and sex, Karonga District, northern Malawi, 1986. 0 0 10 20 30 40 50 WHO U1046 Age at examination (years) 60 70 of the right arm. From available records it appears that most if not all of the vaccine used has been Glaxo freeze-dried. The dose used for infants up to 3 months of age has been 0.05 ml, while the standard 0.1-ml dose has been used for anyone older than this. Results Table 1 shows the scar status by age and sex as recorded at the first LEP examination of 112821 individuals interviewed between 1979 and 1984. Among young children the proportions of males and females with apparent BCG scars were quite similar-approximately 60% among 0-4-year-olds, falling slightly to 55% among 10-14-year-olds. Appreciably higher proportions of 15-24-year-olds exhibited scars (up to 78% among males aged 15-19 years) and a clear difference occurred between the sexes for 20-24-year-olds, among whom approx- imately 74% of males, but only 60% of females, had evidence of a BCG scar. In each age group, between (a) BCG scar "positive" Age group (years) = 2937 04 (I L 10 20 30 40 50 2 n = 3172 5- le 10 20 30 40 50 1014 c%i 2 = 2711 10 20 30 40 50 15-19 I%lf n 1 0 20 30 40 50 2024 % 1977 10 20 30 40 50 20, 2=1825-34 le 0I lil =7 10 20 30 4'0 50 35-44 l%) 20 1 = 234 10 20 30 40 50 45-59 % 10A_ 10 20 30 40 50 260N 20 = 202 >Ole 0 20 30 '0 50 WH o z10a8 Scar size (mmi (b1 BCG scar "doubtful" Age group(years) = 136 0-4 lo 0. 10 20 30 40 50 36t 20 255 = 241 5-9 lIOI 10 20 30 40 50 20 .. n = 19410-14 .I 10 20 30 40 50 20 = 74 15-19 eo 10 0 30 4 50 20 241%I 10 20 30 40 50 20,oJ 5= 127 25-34 1o 2 10 20 30 40 50 35-44 N 20 _=8 10 20 30 40 50 45591N201f 0 = 119 1 0 20 30 40 50 101 0 1S0 20 30 40 5 Scar size (mm) 100 > 'G 50- ae 100 - L'0 o 50- h. 0 10 ( b ) 37 ( a I MALES P.E.M. Fine et al. Fig. 3. Repeatability of a "positive", "negative", or "doubtful" scar reading (a) by Interval between exami- nation and (b) by age at first examination. Repeatability Is taken here to be the proportion of Individuals recorded as scar "positive" (or "negative" or "doubtful") on the first examination, whose scar status remained the same in the subsequent examination. (a) Scar "positive" - - - Scar "negative" . . . - - Scar "doubtful" . 0-24 25-48 49-72 ' 72 Interval between examinations (months) (b) 100 at 15M 50 - O L 0.. WHO 851047 .0 I.- - Scar "positive" --- Scar "negative" * - - g - Scar "doubtful" .-1----.----,.---------e-----------1-------------e----------1- U 0-4 5-9 10-14 15-19 20-24 25-34 35-44 45-59 260 Age group (years) example, the positive repeatability was 80% among 0-4-year olds, increasing to 96% for those aged 15-19 years at the first examination, and falling gradually to only 60% among those aged > 60 years. Negative repeatability was approximately 75% among 0-4-year olds, rising to over 90% for those aged 10-14 years, and increasing slowly up to 96% among the oldest individuals. In contrast, the pro- portion of "doubtful" scars that were considered still to be "doubtful" at subsequent examinations was approximately 15% for all age groups and all inter- vals between examinations. The 0% level for the repeatability of the doubtful category at 0-24 months after the initial examination is based on only three observations. Table 2 shows the results of an independent subsequent assessment of the BCG scars of 494 children who had been observed receiving the vaccine in child health clinics. The data are broken down by age at vaccination and by the time from vaccination to the follow-up examination. Stan- dardized positivity ratios, calculated by the indirect method (19) and expressed as the quotient of observed number of "positive" scars . expected number of "positive" scars x 100, are also shown. Positivity ratios greater or lesser than 100 reflect, respectively, higher or lower proportions of individ- uals with scars than expected on the basis of all the data taken together. The data indicate that the sensi- tivity of scar assessment rose to a peak of 95% at 7-12 months after vaccination and fell, respectively, to 91%, 70%, and 54% at 13-18 months, 19-24 months, and > 25 months. There is no evident relationship between the sensitivity of scar reading and the age at vaccination, while the standardized statistics reveal no important confounding between age at vaccination and the follow-up time. Discussion This analysis has revealed a complex pattern under- lying the distribution of BCG scars in Karonga Dis- trict, northern Malawi. In this respect, the striking age and sex trends observed during the 1979-84 survey (Table 1) largely arose because of the history of BCG vaccination in the district, in particular its introduction in about 1974 with initial emphasis on schoolchildren. This is consistent with the low preva- lence of scars among adults and the higher preva- lence among males than females aged 20-24 years-reflecting the predominance of boys in the upper levels in primary schools in Malawi during the mid-1970s. The strong association between the prevalence of scars and schooling status is shown clearly in Fig. 1. This probably largely reflects the concentration of mobile vaccination teams on schools, but it may also have a broader socioeco- nomic implication, in that the better-off may be both better educated and also more likely to seek or receive the vaccine. In so far as the history of BCG use in Karonga District is similar to that in many areas of the developing world, these associations signal important factors to be considered in any analysis of the protective efficacy of BCG (9). Although the observed distribution of BCG scars by age, sex, and schooling status is perhaps as expected, there is evidence of appreciable misclassifi- cation in these data. For example, the repeatability and sensitivity data in Fig. 3 reveal different patterns of misclassification among the very young, teenagers, and older adults; the reasons for this are discussed below. The low negative repeatability among the very young may arise because some of these children were vaccinated with BCG between two successive exami- 100 I 9 D 50 o CL 38 Epidemiology of BCG scars Table 1: Distribution of BCG scar status by age and sex, Karonga District, northern Malawi, 1979-84 Males Females Age Scar status" Scar status" group (years) P N D U Total P N D U Total 0-4 5871 3660 179 169 9879 5851 3949 202 153 10155 (59.4)b (37.1) (1.8) (1.7) (57.6) (38.9) (2.0) (1.5) 5-9 5139 3502 275 67 8983 5045 3641 278 60 9024 (57.2) (39.0) (3.1) (0.8) (55.9) (40.4) (3.1) (0.7) 10-14 3887 2861 165 83 6996 3674 2750 154 52 6630 (55.6) (40.9) (2.4) (1.2) (55.4) (41.5) (2.3) (0.8) 15-19 4267 1033 56 103 5459 3850 1175 49 114 5188 (78.2) (18.9) (1.0) (1.9) (74.2) (22.7) (0.9) (2.2) 20-24 2873 805 34 167 3879 2715 1633 64 146 4558 (74.1) (20.8) (0.9) (4.3) (59.6) (35.8) (1.4) (3.2) 25-34 1324 3766 117 283 5490 1535 5835 217 272 7859 (24.1) (68.6) (2.1) (5.2) (19.5) (74.3) (2.8) (3.5) 35-44 485 3910 114 188 4697 888 6265 174 239 7566 (10.3) (83.2) (2.4) (4.0) (11.7) (82.8) (2.3) (3.2) 45-59 505 4426 116 160 5207 534 4977 119 213 5843 (9.7) (85.0) (2.2) (3.1) (9.1) (85.2) (2.0) (3.7) )60 192 2517 50 71 2830 112 2308 31 127 2578 (6.8) (88.9) (1.8) (2.5) (4.3) (89.5) (1.2) (4.9) Total 24543 26480 1106 1291 53420 24204 32533 1288 1376 59401 (45.9) (49.6) (2.1) (2.4) (40.7) (54.8) (2.2) (2.3) P = positive; N = negative; D = doubtful; U = unknown. b Figures in parentheses are percentages. Table 2: Proportions of children recorded as BCG scar "positive", by age at vaccination and by the Interval between the date of vaccination and the date when the site of vaccination was examined Age at Interval since vaccination (months) vaccination (months) 0-2 3-6 7-12 13-18 19-24 A25 Total SPR-time' 0-2 23/28 28/32 48/48 27/28 7/9 3/7 136/152 (82)b (88) (100) (96) (78) (43) (89) 102.3 3-5 30/36 43/48 39/41 14/16 3/4 4/6 133/151 99.8 (83) (90) (95) (88) (75) (67) (88) 6-8 4/4 2/2 1/2 1/1 - 0/2 8/11 91.3 (100) (100) (50) (100) (0) (73) 9-11 2/2 - 2/2 2/2 - 6/6 121.0(100) (100) (100) (100) 12-14 0/1 - _ _ _ 0/1 0 (0) (0) Unknown - - 11/13 32/36 35/52 40/72 118/173 98.0 (85) (89) (67) (56) (68) Total 59/71 73/82 101/106 74/81 47/67 47/87 401/494 (83) (89) (95) (91) (70) (54) (81) SPR-age' 96.3 100.3 110.5 114.4 95.8 76.0 8 SPR-time = Standardized positivity ratio based on indirect standardization for sex and time since vaccination. b Figures in parentheses are percentages. c SPR-age = Standardized positivity ratio based on indirect standardization for sex and age at vaccination. P.E.M. Fine et al. nations, while the low positive repeatability is due in part to the disappearance of BCG scars among children who had been vaccinated at a very young age. The follow-up data on children observed to have been vaccinated in infancy suggest that the pro- portion with a recognizable scar falls to less than 60% within 3 years (Table 2). An additional reason for the low sensitivity among young children might have been the low dose of vaccine used for those aged less than 3 months (0.05 ml instead of 0.1 ml). The data in Table 2 indicate that it may be several weeks after vaccination before a recognizable scar forms. Our findings therefore suggest that scars may be poor indicators of BCG vaccination in infancy, and this is corroborated by the results of several other studies. For example, Beskow et al. reported that of 383 Swedish children vaccinated at birth with an adult dose of Gothenburg-strain vaccine only 225 (59%) had recognizable scars at 7 years of age (12). Also, Grindulis et al. reported that of 112 Asian children in the United Kingdom given Glaxo BCG vaccine at birth, only 84 (75%) had recognizable scars at 22 months of age (20). Furthermore, the report of the BCG trial in south India contains the following statement: "The frequency of persons without a scar, though vaccinated ... was highest in the youngest age group" (5). In contrast, reports have appeared of a higher prevalence of scars follow- ing vaccination of infants, e.g., 97% among 740 Sri Lankan children, but it is not clear whether the vac- cination history of the children was known (21). Although the rates will no doubt vary with different vaccines and methods of administering them, an appreciable proportion of infant BCG vaccinations probably do not leave a permanent scar, a fact that must be recognized in attempts to evaluate the effi- cacy of infant BCG vaccination by case-control methods. Both the positive and negative repeatability were high among teenagers (Fig. 3(b)). Here, the high positive repeatability may reflect the following. First, as indicated by the results of the south Indian BCG trial (5), BCG vaccination of older children may be more likely to lead to a permanent scar than that of infants and young children. Second, most of the children in the present study were vaccinated during mass campaigns, and it is likely that both the quality of the vaccine and the administration technique were better during the campaigns than thereafter. Third, because most of the children were vaccinated almost 10 years prior to being examined in the LEP, those vaccine lesions that remained belonged to a subset that were more durable. The low positive repeatability among adults aged over 24 years, and in particular among those aged over 44 years, probably indicates that an appre- ciable proportion of scars observed in such individ- uals were not due to BCG vaccination. This is supported by the history of BCG use in the district, in so far as the vaccine was primarily used for those under 15 years of age when it was introduced in 1974. Although older individuals were not refused the vaccine if they requested it, they were not actively encouraged to come forward for vaccination. Further evidence of an increasing proportion of false-positive scars among older individuals is the decreasing size of lesions that were attributed to BCG vaccination among those aged > 25 years (Fig. 2(a)). The true proportion of vaccinated individuals among those aged > 35 years in Karonga District at the time of the survey was therefore probably appre- ciably less than the 10% shown in Table 1. Accord- ingly, data on individuals aged > 35 years of age were not used to calculate BCG efficacy in the study population (9). The difficulty in assessing BCG scars, e.g., in evaluating the proportion that were "doubtful" and the poor repeatability statistics, has another impor- tant implication. The less clear the mark or scar, the greater the potential for subjective bias in observing or interpreting it. For example, prejudiced exami- nation of an arm may interpret a small blemish as a BCG scar or provide an alternative explanation for a suspicious mark, even if it is in the correct place for an vaccination scar. The data on lesion sizes (Fig. 2) and repeatability (Fig. 3) reveal considerable scope for fatigue, imagination, or prejudice on the part of the observers. The extent of the bias that may be introduced depends on the context and the exam- iner, but is potentially considerable and should be taken into consideration in the design and interpre- tation of studies that are dependent on BCG scar data. In this respect, vaccine-efficacy studies based upon scar evidence of vaccination should be designed such that scars are assessed without any knowledge of the clinical status of the subject. In some circumstances it may be possible to validate scar information against documentary evidence of vaccination (10, 11). The implications of misclassification of vacci- nation status are important. Table 3 illustrates the extent to which non-differential misclassification of this status, i.e., misclassification that is independent of outcome or disease status, will tend to reduce esti- mates of vaccine efficacy. The reduction can be con- siderable; for example, a sensitivity of 60% and a specificity of 75% (values that may not be impossible after neonatal vaccination, according to the data in Table 2, Fig. 3(b), and other published work (11, 12, 20)) would lead to a 50% underestimation of true vaccine efficacy. On the other hand, differential (biased) scar assessment may well have the opposite 40 Epidemiology of BCG scars Table 3: Vaccine efficacles that would be observed, given different levels of non-differential misclaslfication of vaccination staus, assuming true vaccine efficacies of 90%, 75%, 50%, and 25%" Sensitivity of scar readingb Specificity of scar readingc 100% 90% 80% 70% 50% True efficacy: 90% 100% 90% 89% 88% 87% 86% 90% 82% 79% 76% 73% 63% 80% 75% 71% 66% 60% 45% 70% 69% 63% 57% 49% 30% 50% 60% 50% 40% 28% 0 True efficacy: 75% 100% 75% 73% 71% 70% 67% 90% 68% 65% 61% 58% 49% 80% 63% 58% 53% 48% 35% 70% 58% 52% 45% 39% 23% 50% 50% 41% 31% 22% 0 True efficacy: 50% 100% 50% 48% 45% 43% 40% 90% 45% 42% 39% 36% 29% 80% 42% 37% 33% 29% 20% 70% 38% 33% 28% 24% 13% 50% 33% 26% 19% 13% 0 True efficacy: 25% 100% 25% 23% 22% 20% 18% 90% 23% 21% 19% 17% 13% 80% 21% 18% 16% 14% 9% 70% 19% 16% 13% 11% 6% 50% 17% 12% 9% 6% 0 ' The calculations assume follow-up of equal-sized cohorts of vaccinated and non-vaccinated individuals (22). A similar effect is predicted if case-control methods are used. b Sensitivity is defined as the proportion of truly vaccinated indi- viduals who have a recognizable scar. c Specificity is defined as the proportion of non-vaccinated indi- viduals who lack an apparent scar. effect on estimates of vaccine efficacy. For example, if the observers' bias favours BCG, i.e., if they believe that BCG vaccine imparts some protection against either tuberculosis or leprosy, they may be inclined to see fewer BCG scars on diseased individuals than they would otherwise. This would lead to an increase in the calculated estimate of BCG efficacy. It may be noted that the tendencies for these opposite effects are correlated, in so far as situations in which a high proportion of scars are unclear will lead to greater non-differential as well as differential misclassifica- tion; however, the overall effect on apparent vaccine efficacy is difficult if not impossible to assess. A further implication of the use of scars in vaccine efficacy studies is that there may be an association between the tendency for vaccination to leave a scar and that for vaccination to protect against disease. This might occur because improperly administered vaccines are unlikely either to leave a scar or to protect; or else, there may be a more subtle immunological reason. Whatever the explana- tion, such an association would indicate that uncor- rected vaccine efficacy estimates based upon scars as evidence of vaccination are, in effect, measures of the protective efficacy not of receiving BCG vaccine but of being scarred by it. It should be emphasized that the data we have reported here for Malawi were collected by para- medical workers with considerable experience in examining skin and scars. We suspect that the prob- lems described here will be encountered in many populations, and that BCG scars are not as simple and reliable as indicators of past vaccination as some workers have claimed. Acknowledgements We thank Dr A. Galazka and Mr. H.G. ten Dam for helpful comments on a draft of this paper and Miss Mona Maji- thia for secretarial assistance. The Lepra Evaluation Project is funded largely by the British Leprosy Relief Association (LEPRA), with additional support from the IMMLEP component of the WHO/UNDP/World Bank Special Programme for Research and Training in Tropi- cal Diseases. Resume Repartition et signification des cicatrices de BCG dans le nord du Malawi La presence ou I'absence d'une cicatrice de BCG est utilisee comme indicateur de la vaccination par le BCG. Ces donnees sur les cicatrices sont utiles lors des enquetes sur la sensibilite a la tuberculine afin d'estimer l'incidence de l'infection et de la presence de Mycobacterium tuberculosis dans la communaute, comme marqueur dans les programmes de vaccination, et comme moyen de validation des codes dans les essais vaccinaux, de meme que pour servir de base a l'estimation de 1'efficacite du vaccin BCG lors des etudes cas- temoins ou des etudes sur les contacts domes- tiques. L'article presente des donnees sur les cica- trices de BCG chez plus de 112000 personnes examinees au cours d'une enquete sur la lepre portant sur 1'ensemble de la population du district de Karonga, dans la partie nord du Malawi, entre 1979 et 1984. Le vaccin BCG avait tout d'abord ete administre a la population au cours d'une cam- pagne de vaccination de masse entre 1974 et 1977, campagne qui etait axee sur les sujets de moins de 15 ans. Apres cette campagne, ce sont les dis- pensaires pediatriques qui ont et charges de la vaccination BCG. La repartition par Age et par 41 P.E.M. Fine et al. sexe des cicatrices de BCG visibles, selon les re- sultats de l'etude, traduit bien cette situation, avec notamment l'intervention initiale des equipes mobiles sur les enfants d'Age scolaire. Des examens independants et repetes portant sur un grand nombre de sujets ont montre que la proportion de sujets classes dans le meme groupe en ce qui concerne la presence ou I'absence de cicatrice variait sensiblement selon l'age des sujets examines. Par exemple, chez les sujets jeunes, la reproductibilite d'une lecture positive etait faible, d'environ 75%. Cette constatation a ete corroboree par le suivi d'enfants dont on savait qu'ils avaient requ le vaccin BCG dans un dis- pensaire pediatrique: au bout de trois ans, moins de 60% de ces enfants avaient une cicatrice reconnaissable. Le reproductibilite de la lecture etait en revanche elevee (95%) pour les sujets ages de 15 a 24 ans puis baissait de nouveau pour n'etre plus que de 60% au dela de 60 ans. Ces observations laissent a penser que la cicatrice de BCG n'est pas un indicateur tres sensible de la vaccination des nourrissons et est un indicateur peu specifique chez les sujets ages. Les donnees presentees montrent dans quelle mesure differents niveaux de classification erro- nee des cicatrices de BCG conduiront a une sous- estimation de l'efficacite protectrice du vaccin. Les resultats montrent que la tendance a commettre une erreur de classification en raison d'un biais inherent A l'observateur est correlee a cette ten- dance en I'absence de tout biais, et que ces obser- vations erronees peuvent produire des effets opposes sur 1'estimation de l'efficacite des vaccins. Les auteurs notent que 1'estimation de la protection d'apres l'observation des cicatrices de BCG donne en fait des renseignements sur 1'efficacite protectrice que confere le fait d'avoir une cicatrice et non sur l'efficacite protectrice de la vaccination. References 1. Fine, P.E.M. BCG vaccination against tuberculosis and leprosy. British medical bulletin, 44: 691-703 (1988). 2. Tuberculosis control, a manual on methods and pro- cedures for integrated programs. Washington, Pan American Health Organization, 1986 (PAHO Scientific Publication No. 498). 3. Velimirovic, B. et al. Infectious diseases in Europe: a fresh look. Copenhagen, WHO Regional Office for Europe, 1984. 4. Expanded Programme on Immunization. Programme review-Zambia. Weekly epidemiological record, 60: Reprint No. 4946 56-59 (1985). 5. Tuberculosis Preventatlon Trial, Madras. Trial of BCG vaccines in south India for tuberculosis preven- tion. Indian journal of medical research, 72 (Suppl. 1): 1-74 (1980). 6. Smith, P.G. Retrospective assessment of the effect- iveness of BCG vaccination against tuberculosis using the case-control method. Tubercle, 62: 23-35 (1983). 7. Smith, P.G. Evaluating interventions against tropical diseases. International journal of epidemiology, 16: 159-166 (1987). 8. Putrall, J. et al. A case-control study of effectiveness of BCG vaccination in Jakarta, Indonesia. Pro- ceedings of the Eastern Regional Tuberculosis Con- ference of the International Union against Tuberculosis. 1983, pp. 194-200. 9. Fine, P.E.M. et al. Protective efficacy of BCG against leprosy in northern Malawi. Lancet, 2: 499-502 (1986). 10. Tldjanl, 0. et al. The protective effect of BCG vacci- nation of the newborn against childhood tuberculosis in an African community. Tubercle, 67: 269-281 (1986). 11. Padungchau, et al. The effectiveness of BCG vacci- nation of the newborn against childhood tuberculosis in Bangkok. Bulletin of the World Health Organiz- ation, 64: 247-258 (1986). 12. Beskow, R. et al. Tuberculin sensitivity in Swedish schoolchildren vaccinated with BCG at birth. Bulletin of the International Union against Tuberculosis, 55: 100-104 (1980). 13. Edwards, L.B. et al. BCG vaccination: studies by the WHO Tuberculosis Research Office, Copenhagen. Geneva, World Health Organization, 1953 (WHO Monograph Series No. 12). 14. Nybo., J. & Bunch-Christensen, K. Assay in man of different BCG products. Bulletin of the World Health Organization, 35: 645-650 (1966). 15. ten Dam, H.G. et al. The use of jet-injectors in BCG vaccination. Bulletin of the World Health Organiz- ation, 43: 707-720 (1970). 16. Pyakalla, T. & Pust, R. BCG scar surveys in Enga province: a medical anthropology note. Papua New Guinea medical journal, 20: 141-142 (1977). 17. Ponnighaus, J. M. et al. The Lepra Evaluation Project (LEP) an epidemiological study of leprosy in northern Malawi. I. Methods. Leprosy review, 58: 359-375 (1987). 18. Bernler, H. Manual for under-five clinics in Malawi. Ministry of Health, Lilongwe, 1978. 19. Rothman, K. J. Modern epidemiology. Boston, Little, Brown & Co, 1986. 20. Grlndulls, H. et al. Tuberculin response two years after BCG vaccination at birth. Archives of disease in childhood, 59: 614-619 (1984). 21. Karallledde, S. et al. Tuberculin response of Sri Lankan children after BCG vaccination at birth. Tubercle, 68: 33-38 (1987). 22. Copeland, K.T. et al. Bias due to misclassification in the estimate of relative risk. American journal of epi- demiology, 105: 488-495 (1977). 42