Analysis of survey data investigating the malarial endemicity of a mixed tribal population of Bihar, India

Q¨ UESTII ´ O,vol. 22, 2, p. 365-378, 1998 ANALYSIS OF SURVEY DATA INVESTIGATING THE MALARIAL ENDEMICITY OF A MIXED TRIBAL POPULATION OF BIHAR, INDIA T.K. BASU S. GANGULY S.K. SARKAR Indian Statistical Institute  R. ARNAB University of Durban  A section of the mixed tribal population of the Singhbhum district, Bihar, India is declared malaria epidemic zone. The tribal population of several generation is known to be suffering from malaria. A survey based on crosssectional data analysis, was conducted on the mixed tribal population for one month. The purpose of this study was to investigate the health status using collected blood samples. The main focus is on the comparative roles of the «defense mechanism»and «vitality factor»of the human system in context to the malarial infection. By gradual elimination of the blood parameters by statistical analyses, the «vitality»parameters probing malarial endemicity are assessed with a view to predicting the epidemic. The main findings from this survey are (i) of the selected twenty two parameters of blood, albumin, total cholesterol, total protein, β-Globulin, γGlobulin, Immuno globulin G seem to have some predictive capacity with respect to the malarial endemicity of the tribal people and (ii) Categorical variables like blood groups and sex are comparatively less important for prediction of malaria. Keywords: Vitality factors; defense mechanism; two-way Anova; discriminant function.  Biometry Research Unit. Indian Statistical Institute. 203 Barrackpore Trunk Road. Calcutta 700 035, India.  Department of Statistics. University of Durban, South Africa. – Received January 1997. – Accepted November de 1997. 365 1. INTRODUCTION The Indian Council of Medical Research, a leading medical research concern of India, identified a few villages of tribal habitat of the Singhbhum district of Bihar (India) as a malarial epidemic zone. The people of these villages show evidence of malarial suffering for several generations. A study was undertaken to investigate the malarial endemicity by checking the blood samples obtained from a mixed tribal population of the Singhbhum district. The investigation was executed by the Indian Council of Medical Research in conjunction with the Biometric Unit, Indian Statistical Institute. Malaria depends upon two principal factors- (a) vitality and (b) defense mechanism, regulated by the circulation of blood. Vitality is controlled by total, free and ester cholesterol, albumin, red blood corpuscles (RBC) and haemoblobin, and the defense mechanism is controlled by immunoglobulins (IgG, IgA and IgM), white blood corpuscles (WBC), monocytes, nutrophils, eosinophils and lymphocytes. Lymphocytes are subdivided into T and B lymphocytes. In addition, the role of globulin-containing fractions is also prominent on this aspect. Under malarial conditions, RBC, haemoglobin and albumin decrease whereas WBC, neutrophils and globulins tend to increase. This is a very broad overview of this complex mechanism. In this context, a brief description of the biology of malaria is presented. 1.1. Biology of Malaria Malaria is caused by a parasite protozoan infection. Four different species of the genus Plasmodium are known to infect humans. In the tropics, Plasmodium falciparum is most prevalent. The life cycle of the parasite is an interaction between a female mosquito of the genus Anopheles (the vector) and the human host. Transmission of disease occurs by a bite of the infectious mosquito. The parasite migrates to the liver, remains in the latent stage for several days while replicating. Ultimately, there occurs a penetration of host RBC with an asexual replication within the parasites which results in the lysis of the cells. Malarial symptoms are occurred by this asexual parasites in the blood. When there is a fresh case of mosquito biting, a sexual stage, called gametocytes develops (from the asexual parasites) which is responsible for transmission of parasites from the host. These transmitted parasites fertilize in mosquito gut, replicate and a new cycle of transmission starts. These types of epidemiological surveys are generally based on blood smears in which one can observe both the asexual parasites and gametocytes. Classifications of blood smears are generally done in terms of the appearance/absence of parasite or in terms of its density. According to above, there are actually three classes of population related to the malarial statussusceptibles, the proportion that are uninfected (Class 1); infecteds, the proportion with infection (Class 2); and immunes, the proportion with asymptotic infection which is not prominent 366 (Class 4). The Febrile Class i. e. Class 3 represented subjects with unidentified fever, not malaria (as confirmed by slide test). 1.2. Main Findings Subsequent analysis of data gives some indication that the levels of albumin, total cholesterol, total protein, βGlobulin, γ-Globulin and IgG in the circulating system may be used for the prediction of malarial endemicity of the tribal population. The categorical variables like blood groups and sex are less important for the prediction of the malarial incidence. 2. SAMPLING DESIGN AND COLLECTION OF DATA A simple random sample of 88 persons were selected from the list of volunteers from the tribal population. Among them, 45 were male and 40 were female. The selected individuals were invited at the Clinics of Kokda and Khanderber, organized by the Gram Vikas Kendra, a rural development centre of Tata Engineering and Locomotive Works, Jamshedpur, Bihar. The list of volunteers was prepared earlier by the Centre. 10 persons were invited daily in the Clinics serially from the chart. The average daily attendance was six. This is because most of the females did not respond owing to low literacy level, social-taboos or customs. The nature of nonresponse is quite random in nature and so this will not cause any bias in the experiment. However, the efficiency of the estimators will reduce because of the reduction of sample size. Relevant history regarding family members, total income of the family and other socioeconomic data of the volunteers were collected by the Centre and were sent to the Clinics for record. They were mostly nonvegetarians residing in huts as usually observed in tribal villages. According to their family income, they belonged to the «below poverty-level»group. 2.1. Experimental protocol Subjects, as instructed earlier, came in post-absorptive stage (overnight fasting) in the Clinics between 9 to 10 am. The attending physician checked the weight, pulse and blood pressure of each individual and drew blood from their bracial vein for subsequent testing. 10ml of blood was taken in an aseptic condition for the biochemical, haematological and immunological tests. Out of this 10ml blood, 4ml was kept in a sterilized glass vial with the sequestering agent. Blood smear was prepared on two glass slides, one thick (for the detection of Malarial parasite) and the other thin 367 (for differential count of the blood cells). The remaining portion of blood was taken in a sterilized test tube and was allowed to clot for release of serum. The electrophoretic separation of different protein fractions was carried out in the line of Smithies(1955) with 1% agar and 0.1M Veronal Buffer (pH 8.6). The fractions of ProteinAlbumin and Globulins were scanned by using a Densitometer. The relative percentage of different Protein fractions was measured by using a Planimetre. The Immunodiffusionstudy was performed according to the standard method (Mancini et al, 1965). The diametre of the precipitating ring was measured in mm by using the ‘Immunomeasure’ scale. The standard curve was plotted on a double log graph paper using standard antigens supplied by M/s Immunodiagnostics Pvt. Ltd., Delhi. The Immunoglobulin fractions were calculated from the standard curve. The conventional methods for the estimation of Total protein (Lowry et al, 1951), Total, Free and Ester cholesterol(Wootton,1974), Haemoglobin, RBC,WBC,Blood groups and Differential counts i.e.estimation of Leucocytes, Monocytes, Eosinophiles, Basophils (Davie and Lewis,1984), Tand BLymphocytes (Blood, 1977) were followed (Summary tables 1A-1C). On the basis of the above pathological tests the population was post stratified into four Classes according to their malarial status. Class 1. Normal: Those who had no fever currently and no past history of malaria within the last couple of years. Total number 20 (17 male and 3 female) Class 2. Malarious: Those who turned up with symptoms of malaria and showed the malarial parasite positive in the ‘slide test’. Total number 15 (11 male and 4 female) Class 3. Febrile: Those who emerged with fever but no malarial parasite was identified in the ‘slide test’. Total number 25 (15 male and 4 female). Class 4. Dormant: Those who had recurring malaria during the last two years but showed no current malarial symptoms. Total number 28 (21 male and 7 female). 3. STATISTICAL ANALYSIS A: Analyses of cardiovascular and blood parameters The means of the following twenty one blood parameters (viz. Total Protein (TP), Total Chlesterol (Tch), Free Chlesterol (Fc), Ester Cholesterol (Ec), Red Blood Corpuscles (RBC), White Blood Corpuscles (WBC), Neutrophils (N), Lymphocytes (L), Eosinophils (E), Monocytes (M), Haemoglobin (Hb), Immunoglobulin G (IgG), Immunoglobulin A (IgA), Immunoglobulin M (IgM), Albumin (Alb), α1-Globulin (α1), α2-Globulin (α2), β-Globulin, γ-Globulin, T-cell (Tc), B-cell (Bc) and two cardio368 vascular parameters, namely, Pulse (P) and Blood Pressure, Systolic/Diastolic [Bp (Sys/Dia)] for the four classes of people (Normal, Malarious, Febrin and Dormant) were estimated through the sample means which are given in the tables 1A, 1B and 1C. The estimated standard errors (SR) of the estimates are also presented in these tables. The following table gives an idea of the general health condition of the tribal community under consideration. Table 1 Parameters Normal Range mean standard standard Proportion Proportion Proportion sample Indian deviation error(se) below above outside range Standard normal normal normal min-max range range range P(c/min) 65-85 94.898 15.834 1.688 0 70.11 70.11 68-124 BP/sys 100-140 110.932 13.118 1.398 17.04 0 17.04 78-150 BP/dia 70-90 76.33 10.783 1.149 20.45 4.54 24.94 54-100 WBC(m/cmm) 4000-10000 7185.8 931.853 99.336 0 0 0 5000-9500 α1(% Tp) 52-68 3.555 1.351 0.144 26.13 23.86 50 0.9-8.7 α2(%Tp) 6.1-10.1 3.838 1.697 0.181 89.7 0 89.7 1-9.9 β(% Tp) 8.5-14.5 6.619 2.495 0.266 84.1 0 84.1 2.4-14.5 γ(% Tp) 10-21 18.991 5.38 0.537 4.54 27.27 31.88 6-42.9 IgG(mg%) 700-1500 2453.69 598.515 63.802 0 85.22 85.22 1205-3477 IgA(mg%) 90-450 193.818 18.698 1.993 0 0 0 152-237 IgM(mg%) 40-250 128.352 28.642 3.053 0 0 0 69-192 Alb(% Tp) 52-68 67.436 7.046 0.751 2.27 51.13 53.46 45.7-85.1 N(%) 60-65 56.125 5.858 0.624 65.9 6.81 72.72 42-68 L(%) 20-40 33.318 4.797 0.511 0 7.95 7.95 24-48 E(%) 1-3 8.148 4.268 0.455 0 87.45 87.45 1-19 M(%) 2-8 2.614 1.309 0.14 13.63 0 13.63 1-6 Hb(gm%) 14-16 11.063 1.231 0.131 100 0 100 8-12.8 Tc(%) 70-75 63.5 6.447 0.687 80.68 0 80.68 42-75 Bc(%) 15-20 24.08 5.126 0.546 1.13 70.45 71.59 12-38 Tp(gm%) 6-8 7.142 0.387 0.041 0 0 0 6.5-8 Tch(mg%) 150-280 126.773 22.951 2.447 82.95 0 82.95 80-184 Fc(mg%) 50-70 42.045 13.597 1.449 78.4 2.27 80.68 20-94 Ec(mg%) 95-210 85.886 22.21 2.368 69.31 0 69.31 26-140 RBC(m/cmm) 4000-10000 3.794 0.474 0.051 98.86 0 98.86 2.31-4.78 From the above table, we briefly comment on the average health condition as follows: BP: Due to the simplicity of living and diet, the community under study maintained acceptably good BP levels. 369 RBC, Hb: Lower values indicate that most of the people are anaemic. Tp, Tch, Fc, Ec: Lower values than the normal range is indication of low fat diets coupled with normal protein levels. E: High concentration of eosinophils is generally correlated with parasitic infection in the people. WBC, N, L, M: The levels of these parameters are acceptable in terms of facilitating immnuno-defence mechanism. Summary Table: 1A Class Sample Statistic P BP/sys BP/dia TP Tch Fc Ec RBC WBC size (count/min) (mm Hg) (mm Hg) (gm%) (mg%) (mg%) (mg%) (m/cumm) (no/count) mean 91.6 112.6 77.7 7.08 136.2 39.2 102.5 4.062 7707.5 1 20 sd 16.28005 8.9241 8.97273 0.32031 21.7154 11.26765 14.42047 0.28971 816.13035 se 3.64033 1.9955 2.00636 0.07162 4.85572 2.51952 3.22451 0.06478 182.19229 mean 94.06667 108.67 7606 6.94666 96 38.63333 57.46667 3.72133 6736.66667 2 15 sd 15.48533 13.656 12.04326 0.23907 10.0133 14.50455 13.9421 0.48780 772.97405 se 3.99829 3.526 3.10955 0.06172 2.58543 3.74506 3.59983 0.12595 199.58104 mean 100.4 106.36 72.92 7.156 136.16 45.72 90.12 3.6264 7084 3 25 sd 15.09967 14.982 10.9724 0.43458 20.9908 13.92413 19.17982 0.43472 953.90984 se 3.01993 2.9965 2.19448 0.08691 4.19816 2.74482 3.83596 0.08694 190.78197 mean 92.78571 115.04 78.25 7.27857 128.143 42.67857 85.45429 3.79071 7144.64286 4 28 sd 15.065 11.975 10.35659 0.39402 14.8413 13.40589 17.97713 0.52002 898.54438 se 2.84701 2.263 1.95721 0.07446 2.80475 2.53347 3.39735 0.09827 169.80892 Summary Table: 1B Class Sample Statistic N L E M Hb Tc Bc size (%) (%) (%) (%) (gm%) (%) (%) mean 55.45 32.85 9.65 2.05 11.82 65 23.1 1 20 sd 4.225 3.454 3.825 0.805 0.621 4.278 4.3 se 0.945 0.772 0.855 0.18 0.139 0.957 0.962 mean 53.667 36.4 5.667 4.267 10.647 63.733 23.6 2 15 sd 5.907 4.514 3.32 1.289 1.447 5.234 5.414 se 1.525 1.165 0.857 0.333 0.374 1.351 1.398 mean 588 31.56 7.64 2.64 10.704 62.16 25.32 3 25 sd 6.203 5.375 4.906 1.353 1.144 7.22 5.732 se 1.241 1.075 0.981 0.271 0.229 1.444 1.146 mean 55.536 33.571 8.857 2.107 11.064 63.5 23.929 4 28 sd 5.635 4.346 3.71 0.673 1.257 7.287 4.705 se 1.065 0.821 0.701 0.127 0.238 1.377 0.889 370 Summary Table: 1C Class Sample Statistic IgG IgA IgM Alb α1α2β γ size (mg%) (mg%) (mg%) (%) (%) (%) (%) (%) (%) mean 1519.3 199.65 114.35 73.985 3.095 3.135 4.83 14.885 1 20 sd 199.158 19.132 28.079 5.513 1.28 1.042 1.36 4.62 se 44.533 4.278 6.279 1.233 0.286 0.233 0.304 1.003 mean 2870.333 182.6 129.067 58.74 3.68 4.353 8.267 24.867 2 15 sd 277.906 19.231 24.845 6.706 1.574 2.513 3.374 6.182 se 71.755 4.965 6.415 1.732 0.407 0.649 0.871 1.596 mean 2777.96 190 135.84 67.844 3.792 3.74 5.084 19.48 3 25 sd 410.31 15.773 33.308 3.86 1.406 1.484 1.606 3.417 se 82.062 3.155 6.662 0.772 0.281 0.297 0.321 0.683 mean 2608.393 199.07 131.286 67.054 3.604 4.136 6.971 18.339 4 28 sd 267.147 16.873 22.247 5.046 1.125 1.527 2.125 3.693 se 50.486 3.189 4.204 0.954 0.213 0.289 0.402 0.698 B: Analysis of Variance In this section, we will study whether or not the means of 21 blood parameters, BP (Systol/Diastol) and pulse rates are different for 4 classes of people described above. For this study, we consider the following two way analysis of a variance model using class and sex as two classifying factors for each of the 21 blood parameters, pulse rate and BP (Systol/Diastol): yi ; j ; k ( t ) = m + a ( i ) + b ( j ) + c ( i ; j ) + ei ; j ; k (1) where yi ; j ; k ( t ) = response obtained from the kth individuals for the jth ( j = 0 ; 1 ) sex of the ith ( i = 1 ; 2 ; 3 ; 4 ) class, corresponding to the parameter, t. m = general effect a ( i ) = effect due to ith class ( i = 1 ; : : : ; 4 ) a ( j ) = effect due to jth sex ( j = 0 ; 1 ) a ( i ; j ) = interaction effect between ith class and jth sex Assumptions of the Analysis of variance model: (a) ∑ ia ( i ) = ∑ jb ( j ) = ∑ ic ( i ; j ) = ∑ jc ( i ; j ) = o 371 (b) e ( i ; j ; k ) = independently and identically distributed as normal variate with mean zero and variance σ2. The assumptions of the model imply that yi ; j ; k ( t ) ’s are independently normally distributed with mean m + a ( i ) + b ( j ) + c ( i ; j ) and variance σ2. Since the number of observations of the 4 classes (20, 15, 28) are different, unbalanced analysis of varince techniques are used. This is clearly explained with an example by Kshirsagar (1983). Following the analysis of variance tables, the effects of β,γ, IgG, Alb, Tp, Tch and Ech were found to be significant. This implies that the mean of each of the parameters β,γ, IgG, Alb, Tp,Tch and Ech is different for 4 classes. The computations for these parameters are shown in the tables (2A -2G). Table 2A: ANOVA ( β β β ) Source ss df ms F-value P-value Class 123.106 3 41.035 8.141 0 Sex 0.303 1 0.303 0.06 0.807 Interaction 1.754 3 0.585 0.116 0.951 Error 403.26 80 5.041 Table 2B: ANOVA ( γ γ γ ) Source ss df ms F-value P-value Class 627.839 3 209.28 10.509 0 Sex 39.157 1 39.157 1.966 0.165 Interaction 35.099 3 11.7 0.588 0.625 Error 1593.121 80 19.914 Table 2C: ANOVA (IgG) Source ss df ms F value P value Class 14318260 3 4772753.3 47.721 0 Sex 68669.954 1 68669.954 0.687 0.41 Interaction 65112.199 3 3360328.2 33.599 0.844 Error 8001077.2 80 100013.47 372 Table 2D: ANOVA (Alb) Source ss df ms F value P value Class 1236.765 3 412.255 14.108 0 Sex 0.011 1 0.011 0 0.984 Interaction 31.364 3 10.455 0.358 0.784 Error 2337 80 29.222 Table 2E: ANOVA (TP) Source ss df ms F-value P-value Class 2.212 3 0.737 3.892 0.012 Sex 0.033 1 0.033 0.172 0.679 Interaction 0.465 3 0.155 0.817 0.488 Error 15.159 80 0.189 Table 2F: ANOVA (TCH) Source SS df ms F-value P-value Class 13386.909 3 4462.303 13.466 0 Sex 1.986 1 1.986 0.006 0.938 Interaction 1477.278 3 492.426 1.486 0.225 Error 26509.82 80 331.373 Table 2G: ANOVA (EC) Source ss df ms F-value P-value Class 14398.88 3 4799.627 15.454 0 Sex 272.46 1 272.46 0.877 0.352 Interaction 249.905 3 83.302 0.268 0.848 Error 24846.723 80 310.584 373