| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
 |
ABSTRACT |
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
15 years old (OR = 2.9, P = 0.024). Increasing cattle density was a risk factor in one village but not the other. We were not able to determine the route by which VL entered the villages. Our data demonstrate a new spread of VL in previously unaffected areas. We recommend carefully supervised spraying with DDT, surveillance to pinpoint other affected villages, and efforts to increase availability of diagnostic and treatment facilities. |
INTRODUCTION |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
The vector in India is the female sand fly Phlebotomus argentipes. In southern Asia, humans are the only known infection reservoir. The reported incidence of VL decreased to zero during intensive DDT spraying for malaria in the 1950s and 1960s, but VL began a resurgence focused in the northeastern state of Bihar in the 1970s, with an explosive epidemic in the early 1990s.6 In Bihar, VL most commonly affects older children and young adults7; other factors associated with increased risk of progression to disease include malnutrition and a compromised immune system.8 A case-control study in Nepal identified bed net use and the presence of cattle or buffalo as protective, and damp earthen floors as a risk factor.9 A recent community-based study in Bangladesh identified proximity to a previous case of VL as the strongest predictor of risk, and locally available, untreated bed nets as a strong protective factor.10
In disease-endemic communities of Bihar, transmission seems to be stable over time, with incidence rates of approximately 2.5 per 1,000 per year in one active surveillance site with a population of 26,000 (Singh SP, unpublished data). However, in other states, such as Uttar Pradesh, communities may experience outbreaks with attack rate rates 10%.11 There have been no published epidemiologic studies of VL in India in the past nine years. This study was designed to assess the extent of a reported outbreak in several villages in Uttar Pradesh and evaluate risk factors for disease.
|
MATERIALS AND METHODS |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Based on the history of illness, anyone suspected of having VL was examined by a physician and suspect cases were confirmed using the rK39 dipstick (InBios International Ltd., Seattle, WA). Those believed to have kala-azar were referred to Banaras Hindu University for free treatment. We restricted the case definitions to kala-azar; subclinical infection and post kala-azar dermal leishmaniasis were not ascertained. A current confirmed VL case was defined as an illness characterized by at least 15 days of fever and splenomegaly, plus a positive rK39 dipstick result. Past confirmed VL was defined as an illness characterized by at least 15 days of fever and weight loss or splenomegaly that was treated with either sodium stibogluconate or amphotericin B with clinical resolution of symptoms, or with Leishmania amastigotes in bone marrow or splenic aspirate or tissue with documentation in medical records. Cases of VL with onset during the past five years were included. Past probable VL was defined as an illness with fever lasting at least two months that was treated without effect for some other disease (typically malaria or typhoid) and resulted in death within the past five years. Our comparison group comprised all villagers negative by these case definitions.
We estimated the three-year period prevalence of VL in each village by including only cases that occurred from 2001 onward because the time of onset for earlier cases was often uncertain. However, all cases ascertained were included in the risk factor analysis. Thus, each variable was tabulated against VL status and the estimated cumulative five-year risk of VL was calculated; in most cases variables were dichotomized. These tabulations were done using EpiInfo 2000 version 3.2 (Centers for Disease Control and Prevention, Atlanta, GA). Odds ratios (ORs) and P values were calculated in bivariate analyses (outcome and one exposure, adjusted only for the village) using generalized estimating equations (GEEs) with robust variance estimation to account for intra-household clustering. Exposure variables for which P < 0.2 and those that were suspected based on previous work to be important, including possible confounders, were incorporated into multivariate GEE models and the resulting ORs and confidence intervals examined. Effect modification, confounding, and exposure significance were assessed using a modified version of the hierarchical backward elimination strategy.12 All of these analyses were performed with Intercooled Stata version 8.2 (Stata Corporation, College Station, TX).
Included in the above logistic models were several variables derived from house latitude and longitude: the distance from an individual’s house to each VL household (either confirmed or probable), and the minimum distance to any VL household. Summary distance variables (e.g., the distance to the nearest VL household, the number of VL households within 10 meters) were tabulated against VL status and added to the logistic modeling process described above. The collection of GPS data also allowed us to calculate the cattle density based on the number of cattle (cows, calves, or oxen) within a certain radius of an individual’s house location. Since the positions of animal sheds were not measured, cattle were assumed to be located at the position of the household that owned them.
All analyses were performed for both villages together and each village separately, and for confirmed cases only, as well as confirmed and probable cases combined. In the confirmed-only analyses, the proximity variables corresponding to distance to confirmed VL households were used; for confirmed-and-probable analyses, the proximity variables corresponding to any VL household were used.
The study research protocol was reviewed and approved by the Emory University Institutional Review Board (IRB) and the Kala-azar Medical Research Center (India) Ethics Committee, and ruled exempt by the Centers for Disease Control and Prevention IRB. Informed consent was obtained from the responding member of each household.
|
RESULTS |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
There were important differences between confirmed VL cases and probable cases (defined as deaths from presumed VL). Those who died of presumed VL were more likely to be female (67% versus 35%) and poor (100% in households with an income less than 2,000 Rupees versus 40%). With few exceptions, associations based on confirmed cases and those for all cases were similar; thus, confirmed and probable VL cases were combined for the risk factor analysis that follows. The sex ratio was equal in the two villages, with 52.8% males and 47.2% females overall. Village 1 was approximately two-thirds Muslim and one-third Hindu, whereas village 2 was entirely Hindu. Hindus were far more likely than Muslims to own cows (OR = 25.6, P < 0.0005).
In bivariate analyses, living in the same household with a VL case was associated with a markedly elevated risk of VL, while having a large number of people per sleeping room was significantly protective (Table 1
). Having mud, thatch, or unplastered floors, walls, or roofs was a significant risk factor in village 1, but not in village 2. Ownership of cows was a significant risk factor in village 1, but not in village 2. Sleeping outside and on ground level was associated with increased risk of VL, although only for village 1 (Table 2). Age and religion were important only in village 1.
|
|
presents three multivariate models (for both villages combined and for each village alone), adjusting for confounders and risk factors. Since the interaction between the village and proximity variable was significant, interaction terms were included in the combined-village model with results reported for each village. Proximity to a case remained the strongest risk factor in both villages, and in village 2 remained even outside the same household. Also important in village 1 were sleeping outside and downstairs, being 15 years of age, and the nearby presence of cattle. The only factor important in village 2 other than proximity to a VL case was the number of people per room; high numbers were associated with lower risk of VL. An additional multivariable analysis, identical to the first model in Table 3 except with age as a dichotomous variable, demonstrated that individuals 15 years of age were distinctly more at risk compared with children <15 years of age (OR = 2.1, P = 0.038).
|
|
DISCUSSION |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Some of our results confirm previous findings, but we also encountered contradictions. In contrast to studies in Bangladesh and Nepal, we found no significant protective effect for bed nets or mosquito coils.10,13 However, few villagers used bed nets and we may not have had the statistical power to detect an effect. In addition, most people slept outside in the summer, where nets are more difficult to use and the smoke from coils may disperse too quickly to be worthwhile. We also found a somewhat different age effect compared with previous studies. In India, the literature generally reports that older children and young adults are most at risk, whereas we found equally high risk for all those 15 years of age. However, this finding is consistent with the hypothesis that VL is a recent introduction in our study area. Few residents would therefore be expected to have immunity, and adults and children would be equally susceptible.
We hypothesized that greater household population density (number of people per room) would be more attractive to sand flies and reflective of relative poverty, so that high density might be associated with increased risk. However, we found the opposite: high density was protective. Both of the multiple-case households had unusually low population density; nonetheless, the effect remains strong even with inclusion of the same-household variable, which in effect adjusts for multiple-case households.
The association between cattle and increased VL risk suggests a possible explanation for the puzzle. This finding is consistent with a study in Sudan,14 despite differences in the vector, ecology, and parasite strain. Cow manure is a preferred food source and habitat for sand fly larvae, and its presence might be expected to increase sand fly density.15–18 We hypothesize that in our study setting, a major proportion of the disease transmission occurred among the many people sleeping out of doors in the summer. Sand flies attracted to cows may also take blood meals opportunistically from nearby unprotected humans. This hypothesis is entirely consistent with our findings, although inconsistent with results from studies in Nepal and Bangladesh.10,13 However, in the Bangladesh study community, fewer than 5% of study participants reported ever sleeping outdoors (Bern C, unpublished data), compared with 77% of our study population. There may be other differing factors in Nepal that were not identified in the case-control design. Further research focusing on the entomology of P. argentipes in South Asian communities is needed to definitively address this question.
The differences between the results from the two villages are striking. The primary risk factor in village 2 was proximity to another VL case. That effect was so strong that other associations may have been rendered insignificant. In contrast, in village 1, proximity to VL cases outside the same households was not significant; sleeping outside and proximity to cows seem to be key factors. Evident differences in transmission or risk factors may be related to substantial socioeconomic differences between the two villages: village 2 was a rural farming community with more cattle, and houses often widely separated by fields; village 1 was more prosperous, just off a main road, with more widespread access to electricity, less use of mud and thatch building materials, and more two-story houses. Moreover, although the population of village 2 was stable, the population of village 1 had reportedly been growing rapidly, with many migrants from VL-endemic Bihar. Further community-based studies will be necessary to better understand the dynamics of anthroponotic VL transmission in different settings in South Asia.
The major limitations of our study were the small number of cases and the retrospective ascertainment. We might have missed a few cases because of incomplete recall, but VL and its treatment are sufficiently traumatic that we probably did not miss many cases. Conversely, for the cases we did identify, we relied on typical clinical and hematologic findings plus rK39 dipstick results. We categorized past cases as confirmed without dipstick evidence if they had a parasitologic diagnosis at Banaras Hindu University and were successfully treated. It is possible that some of those cases were not truly VL, this is but unlikely. Misclassification bias was a possibility, but the direction of the bias, if any, is not obvious, and there is no reason to suspect differential bias; we believe that our classification was generally accurate.
We chose to present results based on all cases, not just confirmed ones, to minimize one possible source of misclassification bias. Persons identified as having probable VL (defined as death from presumed VL) were poorer and more likely to be female. This is not surprising in northern India where females have less access to health care and higher rates of malnutrition than males. Consequently, although there is a risk of biasing toward the null by counting unconfirmed people as VL cases, there is also a certain risk of introducing bias in the opposite direction if they are indeed cases but left out of the analysis. The ORs for confirmed cases only for sex (OR = 1.3, P = 0.292) and income (OR = 0.80, P = 0.638) were farther from the null than the results shown in Tables 1
and 2. Otherwise, results from the confirmed-case and all-case analyses were generally comparable, suggesting that our case definitions were generally robust.
The long period for which data were collected also raises doubt about the risk-factor data; would data collected in 2004 have been valid in 1999 or even 2001? In these village settings, the answer is yes, for the most part. Village conditions are stable over time, especially in terms of house construction and personal habits of the people. Animals do come and go, but we asked about both current animals and typical numbers in the past and these data were similar. If our results are biased, it is likely to be toward the null, and the true effects would be even larger than reported here.
What should be done about what is clearly a focal outbreak of VL? We strongly recommend carefully supervised residual insecticide spraying, in combination with surveillance to pinpoint other affected villages. Improvement of the primary health care system and better access to diagnostic and treatment facilities are also clearly needed. We collected data for 148 people with prolonged fever, multiple injections, or VL in the past three years. Not one reported going to the block-level Primary Health Center (PHC) for diagnosis or treatment. Lack of availability of antileishmanial drug therapy in PHCs outside Bihar is a likely contributing factor, but perceived poor quality of care at PHCs may also lead to under use.19 These delays greatly amplify the risk of transmission to people living near untreated VL patients. Effective VL control in Uttar Pradesh, as in other parts of South Asia, will require a combined approach: effective vector control, improved case ascertainment through better surveillance, and access to prompt diagnosis and appropriate treatment to decrease the human infection reservoir.
Received March 27, 2005. Accepted for publication May 25, 2005.
Acknowledgments: We are grateful to our indefatigable field workers (Shaheen Parveen, Zoheb Hasan, Shaziya Hasan, Dawar Ali, and Awadh Narayan) who endured many sweltering days in the field. We also thank Dr. Anil Kumar for dipstick testing, Dr. Aryeh Stein for epidemiologic suggestions, and Dr. Altaf Lal and Dr. Lalit Kant for their support. This study would have been impossible without the active participation and cooperation of the villagers of Tamachabad and Ramdattpur.
Financial support: This study was supported by the Anne E. and William A. Foege Global Health Fund, and the U.S. Department of Health and Human Services Fund for Emerging and Reemerging Infectious Diseases and Disease Surveillance
* Address correspondence to Caryn Bern, Division of Parasitic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Mailstop F-22, Atlanta, GA 30341. E-mail: cxb9{at}cdc.gov 
Authors’ addresses: Paul G. Barnett, % MSF-Holland, Plantage Middenloan 14, 1014 DD Amsterdam, The Netherlands, E-mail: chipbarnett2002{at}yahoo.com. Caryn Bern and Allen W. Hightower, Division of Parasitic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Mailstop F-22, Atlanta, GA 30341, Telephone: 770-488-7654, Fax: 770-488-7761, E-mail: cxb9{at}cdc.gov and awh1{at}cdc.gov. S. P. Singh, Department of Community Medicine, Institute of Medical Sciences, Banaras Hindu University, Varanasi 221005, India. Shyam Sundar, Department of Medicine, Institute of Medical Sciences, Banaras Hindu University, Varanasi 221005, India, E-mail: shyam_vns{at}sify.com.
Reprint requests: Caryn Bern, Division of Parasitic Diseases, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Mailstop F-22, Atlanta, GA 30341. Phone: 770-488-7654. Fax: 770-488-7761. E-mail: CBern{at}cdc.gov
|
REFERENCES |
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  J. H Kolaczinski, R. Reithinger, D. T Worku, A. Ocheng, J. Kasimiro, N. Kabatereine, and S. Brooker Risk factors of visceral leishmaniasis in East Africa: a case-control study in Pokot territory of Kenya and Uganda Int. J. Epidemiol., April 1, 2008; 37(2): 344 - 352. [Abstract] [Full Text] [PDF]
|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
