Distribution of the snail Biomphalaria glabrata, intermediate host of Schistosoma mansoni, within a St Lucian field habitat

SCHISTOSOMIASIS Distribution of the snail Biomphalaria glabrata, intermediate host of Schistosoma mansoni, within a St Lucian field habitat R. F. STURROCK 1 A total of 6360 mud samples were obtained, in 62 collections made with an exhaustive sampling device, from banana drains on the West Indian island of St Lucia during fortnightly samplings over a 2'12-year period. Analysis of counts of the snail Biomphalaria glabrata from these samples showed that this species had a contagious distribution. This finding is consistent with other evidence that banana drains form a rigorous habitat for B. glabrata. Its distribution was more contagious than that of Oncomelania quadrasi in certain Philippine habitats and several species of aquatic snail in various African irrigation canals. The exact transformation for normalizing the snail counts for standard statistical techniques was z = X0-287 but the more convenient cube root transformation is probably adequate. However, if too few snails are collected (15 or fewer per 100 samples) or if the frequency distribution ofsnail counts is discontinuous, with too many widely separated high frequency counts, neither transformation will be entirely satisfactory. Yeo (23), using the method of Taylor (20), demon- strated the aggregated distribution of several species of aquatic snail within an experimental irrigation canal in Tanganyika (now Tanzania). Mud samples were collected by a modification (1) of an exhaustive tube-sampling technique (9). The individual snail counts could be transformed to y=xlI' to permit normal statistical analysis techniques. However, Yeo warned that this transformation might not be appli- cable to snail counts from other habitats. The same sampling technique was used by Stur- rock (17) to study Biomphalaria glabrata in banana drains-one of the snail habitats on the West Indian island of St Lucia. The distribution of the species in that habitat is examined in the present paper. MATERIALS AND METHODS A detailed description of the study area and habitats is given by Sturrock (17). Banana drains remove excess rain water to prevent flooding of the fields and drowning of the banana plants. Parallel, earth-lined lateral drains, 0.5 to 0.7 m wide, 0.3 to 1.0 m deep, and 5 to 6 m apart, empty via similar but 1 Research Fellow, Wellcome Trust Research Labora- tories, P.O. Box 43640, Nairobi, Kenya (seconded from the London School of Hygiene and Tropical Medicine, Keppel Street (Gower Street), London WC1E 7HT, England). larger main drains into natural water courses. Depending on the terrain and rainfall conditions, these drains may contain standing or flowing water permanently or at irregular intervals. Eight mud samples (1) were collected at each of 13 sampling stations at fortnightly intervals from March 1967 to June 1969. Exceptions were the first and third collections, when 32 and 88 samples, respectively, were collected. In total, 6360 samples from 62 collections were examined. Taylor's empiri- cal method, as outlined by Southwood (14), was used to examine the data. RESULTS The frequency distribution of snail counts for the 6360 samples is shown in Table 1. Zero counts predominated, but there were some high counts. The counts do not follow a Poisson distribution, because the mean count (x) was 0.5607 and the variance (s2) was 6.5601. The cumulative frequency of the counts plotted on log probability paper (Fig. 1) gave a smooth curve up to 99.8 %. Thus, although the counts do not follow a normal distribution, there is no immediate evidence of polymodality. The nega- tive binomial parameter k was calculated from Nln[1+(x/k)] = E[AxI(k+x)] 3348 - 267 - BULL. WORLD HEALTH ORGAN., Vol. 52, 1975 R. F. STURROCK Table 1. Frequency distribution of snail counts among 6360 banana drain samples No. of snails per No. of samples No. of snails sample collected x f fx Fig. 1. Plot of logio cumulative frequency of snail counts (fx) expressed as a percentage of the total number of snails collected (E fx) against the individual snail counts (x). where N = the total number of samples and Ax = the sum of all the frequencies of sampling units containing more than x individuals (14). In this case k = 0.0711. These was a linear relationship between log 10 x and log1os2 where x and s2 are the means and variances of the 59 collections that yielded snails (Fig. 2). The linear regression line was fitted by the method of least squares (13). The slope (b) was 1.426 ± 0.0491 at the P = 0.95 level of significance. According to Southwood (14), the exact normalizing transformation is x0-287, calculated from z = xP where p = 1 - 1/2b. The value calculated is inter- mediate between the two approximate transforma- tions suggested by Southwood (i.e., the square root and the logarithmic transformations for values approaching 0.5 and 0, respectively), but is close to the cube root transformation used by Yeo (23). The effects of both transformations on the rela- tionship between x and s2 are shown in Fig. 3 for the 59 collections illustrated in Fig. 2. The linear relation- ship is still discernible, but, for many of the data, the " variance is relatively independent of the mean " (14). However, two factors appear to limit the truth 60 x 40 x X x30 20 E 15 10 x1 x .-8 X 6 X ° 5 X 4- X E :1 3 - X z 2 x 1 x 90 95 98 99 99.5 99.8 99.95 99.99 WHO 75901 Percentage cumulative frequency 5441 387 169 93 62 38 31 18 25 13 10 8 11 4 8 5 3 2 5 4 4 2 2 2 3 1 1 0 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 28 29 35 39 41 53 56 63 0 387 338 279 248 190 186 126 200 117 100 88 132 52 112 75 48 34 90 76 80 21 44 23 48 50 78 28 29 35 39 41 53 56 63 Total 6360 (£f) 3566 (Efx) 268 SCHISTOSOMIASIS 0.1 0.01 K X X xK xK 0.01 0.1 1.0 10 WHO 7S902x- log scale Fig. 2. Plot of logiox (mean snail count) against logios2 (variance for 59 collections yielding snails). of this statement. It does not hold where the mean is very low-i.e., where x is less than 0.1 in both Fig. 3a and Fig. 3b, which represents total collec- tions of fewer than about 15 snails in 100 samples. Examination of the original data from such collec- tions shows that the majority of snail counts were in the range 0-2, which inevitably imposes an upper limit on the size of the variance. A second factor became progressively more ap- parent as the number of snails collected increased. There tended to be a discontinuity in the frequency of snails per sample: clusters of high counts were widely separated from a smoother distribution of counts at the lower end of the scale. A relatively simple transformation, such as those used in Fig. 3, Fig. 3. Relationship of mean snail count (x) to the variance (s2) after normalizing the data to (a) z = xO-287 and (b) z = xl1s. is unlikely to normalize this type of frequency distribution satisfactorily. DISCUSSION Accurate quantitative field data are needed to utilize fully the mathematical models currently being developed to study the dynamics of schistosome transmission and to design and assess control pro- grammes. Such field data are difficult to obtain for a single species: the problems are compounded in schistosomiasis, in which three species (man, the parasite, and the snail), each with its own peculiar sampling problems, should be studied simultane- ously. Snails inhabit a variety of complicated habitats. Some authors have been fortunate enough to be able to use exhaustive sampling techniques without unduly disturbing specific habitats (1, 4, 6, 9, 10). More often, though, less reliable relative sampling techniques have been unavoidable (5, 17, 21, 22). Whereas the latter methods allow general trends to be studied and some attempt to be made at deriving population dynamics parameters (7, 16), little infor- mation can be obtained on absolute snail numbers. 269 W 3 Vl R. F. STURROCK It is generally accepted that snails are clumped in aggregated or contagious distributions within habi- tats. There is some evidence that this clumping is related to specific ecological microhabitats, such as certain plants (2, 9, 12). Alternatively, such clumping may be caused by social behaviour by the snails (11). However, little effort has been made to describe these clumped field distributions mathematically, even when exhaustive sampling procedures have been used. A notable exception is the work of Pesigan et al. (9), who showed in the Philippines that Oncome- lania quadrasi followed a negative binomial distribu- tion in several habitats with k = 1, approximately. (The lower St Lucian value for k (0.07) suggests that B. glabrata had a more contagious distribution.) Later, Yeo (23) showed empirically that the snail counts from a Tanganyikan irrigation canal could be normalized by the cube root transformation. Hair- ston (4) used the same transformation, modified by adding 0.5 to each count to allow for zero values, for snail counts from irrigation canals in the Sudan and Egypt because it was simpler than more complicated transformations based on the negative binomial dis- tribution. In fact, the addition of a constant to the raw counts is unnecessary for the inclusion of zero values when a power transformation is used: it is required only for logarithmic transformations. The similarity of snail distributions in these irrigation canals may reflect the homogeneity and similarity of this type of habitat. However, Yeo pointed out that other habitats might have different snail distribu- tions, and variations in the slope (b) of logl0 s2 on log10 x2 might have ecological significance in defining the suitability of different habitats for snails. In Tanganyika, b = 1.29 + 0.08; in St Lucia, b 1.426 ± 0.049. The latter figure is higher, probably significantly so, suggesting that B. glabrata is more contagiously distributed in the heterogeneous condi- tions of the banana drains than are other aquatic snails in the relatively homogeneous African irriga- tion canals. In Yeo's data, the exact normalizing transforma- tion (x0O355) is not significantly different from the more convenient cube root transformation (xl/a) that may be used prior to the application of robust statistical techniques such as the analysis of variance. The same is probably true for the St Lucian data but, as shown in Fig. 3, the cube root transformation is less effective than the exact transformation in reducing the dependence of the variance on the mean. The contagious distribution of B. glabrata un- doubtedly reflects primarily the heterogeneous nature of the banana drains. At any time, except during floods, conditions vary widely, ranging from dry, sun-baked drains, devoid of vegetation and extremely unsuitable for snails, to shaded drains with lush vegetation and gently flowing water in which snails thrive. These conditions, moreover, are not constant and one drain may revert from one extreme to another in a quite short time, owing mainly to seasonal variations in rainfall but also to periodic cleaning of the drains in routine banana husbandry. In adverse conditions many snails die, but those that survive-often by aestivation (15)-can repopulate the habitat explosively when conditions become favourable again, because of their high intrinsic rate of natural increase (18). Nevertheless, conditions in the drains are on the whole very harsh for B. glabrata, as was borne out by field observations (16, 17): hatchling and juvenile snails were relatively less common than in other habitats, birth rates were generally low, and death rates were frequently high. How S. mansoni transmission is affected by the aggregated distribution of the snails depends on the pattern of human contact with the habitat. On St Lucia, human contact with banana drains is low and sporadic and, judged by snail infection rates, this habitat is one of the least important in S. man- soni transmission. However, it is possible, in other habitats with close human contact, that aggregation of the snails may potentiate transmission. Thus, the local situation must be taken into account in mathe- matical models, such as that of Goffman & War- ren (3), in which prominence is given to the role of the snails. It might be thought that the harshness of the environment would favour snail control by mollusci- cides. Paradoxically, the reverse was the case in the St Lucian banana drains. High snail kills were possible where conditions were favourable to snails, because adequate water was present to allow the molluscicide to work. In damp or dry drains, how- ever, aestivating snails were often immune to the molluscicide and rapid repopulation by the survivors was possible when conditions became more favour- able. In such circumstances, a molluscicide control programme must incorporate a stringent surveil- lance-treatment element to prevent a post-treatment resurgence of the snail populations. After several years of such a programme on St Lucia, large colonies though scarcer than before were still 270 SCHISTOSOMIASIS being detected in the banana drains (19). These colonies probably played little part in transmission, but they were still important because they formed a reservoir of snails to reinvade other adjacent habi- tats. One method of assessing the effectiveness of mol- lusciciding is to measure the percentage mortality caused among the snails (4). Such data, in conjunc- tion with data on changes in incidence among the human population, might allow the quantitative estimation of the break point in transmission postu- lated by Macdonald (8). The cost-effectiveness of mollusciciding would then also be calculable. Be- cause of the clumped distribution of the snails, the use of raw snail-counts is inappropriate and, pos- sibly, dangerously misleading. Suitable transforma- tions must be used (4, 23). The transformations illustrated in Fig. 3 would probably suffice for many of the data, but the limitations already noted should not be overlooked. The precontrol assessment will probably sample large snail populations and, if the frequency distribu- tion of the snail counts is discontinuous, these transformations will be inappropriate. If the mol- lusciciding has been successful, then post-application snail counts will be very low. Again, the counts may not be adequately normalized by the transforma- tions shown above. A further complication, to some extent suggested by qualitative rather than quantita- tive data obtained in St Lucia (19), is that the distribution of the snails may be altered by the application of molluscicides. In this case, the same transformation is unlikely to be appropriate for pre- control and post-control snail counts. In fact, no attempt was made to estimate the total pre-control snail population in the banana drains on St Lucia, because the sampling stations had not been selected to represent proportionately the different types of drain present. To design such a sampling programme for 1000 km or more of drains in which conditions are variable and unstable would have been an enormous and, probably, impossible under- taking. Similarly, no assessment of the size of the post-treatment snail population was attempted and the efficiency of the programme was judged primari- ly by the effect on the incidence of human infections ACKNOWLEDGEMENTS The data analysed in this paper were collected while the author was a regular staff member of the Rockefeller Foundation, New York, USA, on assignment to the Research and Control Department, St Lucia, West Indies. R1ESUMIt DISTRIBUTION DU MOLLUSQUE BIOMPHALARIA GLABRATA, HOTE INTERMEDIAIRE DE SCHISTOSOMA MANSONI, DANS UN HABITAT AGRICOLE A SAINTE-LUCIE Dans l'ile antillaise de Sainte-Lucie, Schistosoma man- soni est transmis par le mollusque Biomphalaria glabrata qui a notamment pour habitat le systeme de drainage des bananeraies. Utilisant une technique complete d'echantil- lonnage, A intervalles de 15 jours, on a recueilli, entre mars 1967 et juin 1969, 6360 echantillons dans 13 stations de drainage. Le resultat predominant des numerations 6tait nul mais le nombre moyen de mollusques par echantillon (i) etait de 0,5607, avec une variance (s2) de 6,5601. La distribution etait fortement groupee; pour une distribution binomiale negative, la valeur de k etait de 0,0711. Parmi les 59 collQctes bimensuelles qui ont fourni des mollusques, il y avait relation lineaire entre le log1o x et le log1o S2 avec une pente de 1,426 i 0,0491 au niveau de signification P = 0,95. La transformation exacte en vue de normaliser les nombres trouves aux fins des techniques statistiques traditionnelles etait: z = x0 287 (ou Z = (X + 1)0,287 pour tenir compte des resultats nuls) mais on peut se contenter de la formule plus commode z = (x + 1)'I/. Toutefois, il n'est pas possible d'obtenir une normalisation satisfaisante si le nombre de mol- lusques recueillis est inferieur a 15 pour cent echantillons ou si la distribution de frequence des nombres trouves est discontinue, avec un grand nombre de hautes frequences tres espac&es. Il semble donc que la distribution de B. glabrata dans le systeme de drainage des bananeraies de Sainte-Lucie soit plus groupee que celle de mollusques semblables dans un certain nombre de canaux d'irrigation 271 272 R. F. STURROCK africains et dans divers habitats des Philippines. Cette observation correspond au caractere heterogene des sys- temes de drainage des bananeraies et au fait qu'ils constituent un habitat tres penible pour B. glabrata. II est important d'obtenir ce type de donnees quantitatives afin de pouvoir utiliser efficacement les modeles mathema- tiques, qui sont en preparation pour 1'etude de la dynamique de la transmission, ainsi que pour la concep- tion et l'evaluation de programmes de lutte. REFERENCES 1. CROSSLAND, N. 0. Bulletin of the World Health Organization, 27: 125 (1962). 2. DAWOOD, 1. K. ET AL. Bulletin of the World Health Organization, 32: 269 (1965). 3. GOFFMAN, W. & WARREN, K. S. American journal of tropical medicine and hygiene, 19: 278 (1970). 4. HAIRSTON, N. G. Bulletin of the World Health Organization, 32: 289 (1965). 5. HIRA, P. R. & MULLER, R. Annals of tropical medicine and parasitology, 60: 198 (1966). 6. JOBIN, W. R. American journal of tropical medicine and hygiene, 19: 1038 (1970). 7. JORDAN, P. & WEBBE, G. Human schistosomiasis. Springfield, IL, Thomas, 1969, p. 143. 8. MACDONALD, G. Transactions of the Royal Society of Tropical Medicine and Hygiene, 59: 489 (1965). 9. PESIGAN, T. P. ET AL. Bulletin of the World Health Organization, 18: 481 (1958). 10. SHIFF, C. J. Annals of tropical medicine and parasitol- ogy, 58: 240 (1964). 11. SIMPSON, A. W. ET AL. Behavioral biology, 9: 731 (1973). 12. SMITHERS, S. R. Transactions of the Royal Society of Tropical Medicine and Hygiene, 50: 354 (1956). 13. SNEDECOR, G. W. & CHOCHRAN, W. G. Statistical methods, 6th ed. Ames, Iowa State University Press, 1967, p. 135. 14. SOUTHWOOD, T. R. E. Ecological methods with particular reference to insect populations, London, Methuen, 1966, p. 6. 15. STURROCK, R. F. Annals of tropical medicine and parasitology, 64: 365 (1970). 16. STURROCK, R. F. International journal of parasitol- ogy, 3: 165 (1973). 17. STURROCK, R. F. International journal of parasitol- ogy, 3:175 (1973). 18. STURROCK, R. F. & STURROCK, B. M. Annals of tropical medicine and parasitology, 66: 385 (1972). 19. STURROCK, R. F. ET AL. International journal of parasitology, 4: 231 (1974). 20. TAYLOR, L. R. Nature, 189: 732 (1961). 21. WEBBE, G. Bulletin ofthe World Health Organization, 27: 59 (1962). 22. WEBBE, G. Population studies on intermediate hosts in relation to transmission of bilharziasis in East Africa. In: Wolstenholme, G. E. & O'Connor, M., ed. Bilharziasis: CIBA Foundation Symposium. London, Churchill, 1962, p. 7. 23. YEO, D. Bulletin of the World Health Organization, 27: 183 (1962).