Am. J. Trop. Med. Hyg., 76(3), 2007, pp. 446-449
Copyright © 2007 by The American Society of Tropical Medicine and Hygiene
MULTILEVEL MODELING OF INTRA-HOUSEHOLD SPREAD OF HEPATITIS C VIRUS INFECTION, KARACHI, PAKISTAN
SAEED AKHTAR* AND
TARIQ MOATTER
Department of Community Medicine and Behavioural Sciences, Faculty of Medicine, Kuwait University, Safat, Kuwait; Department of Community Health Sciences, Aga Khan University, Stadium Road, Karachi, Pakistan; Department of Pathology and Microbiology, Aga Khan University, Stadium Road, Karachi, Pakistan
 |
ABSTRACT
|
Several epidemiologic studies have reported the existence of nonsexual intrafamilial hepatitis C virus (HCV) transmission. However, actual routes and their relative efficiency have been controversial. The objective of this study was to investigate whether contacts of HCV seropositive index patient living in the same household have similar probabilities of being HCV seropositive with respect to any of the household-level variables, after taking into account the independent effects of individual-level variables. We analyzed cross-sectional serological data on 341 nonsexual household contacts of 86 HCV-seropositive index thalassemic patients with a multilevel logistic regression model using household contacts at the first level and household characteristics at the second level. Prevalence of HCV seropositivity among household contacts who were tested was 20.5% (70/341). Multilevel analysis of household-level fixed effects indicated that contacts living in families wherein the index thalassemic patient was RNA positive—compared with those contacts living in families wherein the index thalassemic patient was RNA negative—had higher odds of being HCV positive (OR = 2.09; 95% CI: 1.02 to 4.28). Nonetheless, the effect of index patients’ RNA status on the contacts’ HCV serostatus was small in comparison with the unexplained between-cluster variation. The results of this study are pertinent for household members of HCV-infected patients; specifically, their close contact with objects that are contaminated with blood or perhaps saliva of the HCV-seropositive index patient may pose increased risk of HCV transmission. High household intercept variances in different analyses revealed that at there are still unrecognized nonsexual modes of HCV transmission at the household level that need further research.
|
INTRODUCTION
|
The households contacts of hepatitis C virus (HCV) seropositive patients have been shown to have an elevated risk of HCV infection.1,2 However, previous research has been inconclusive on the role nonsexual household exposures entailing an increased risk of HCV transmission.3,4 On the contrary, we demonstrated a significant clustering of HCV seropositivity among nonsexual contacts of HCV-seropositive thalassemic patients.5 When variations in health or health-related behaviors between clusters are investigated, it is relevant to measure the extent to which the phenomena occur in clusters. Furthermore, it is useful to determine whether clusters’ levels of variation can be explained by a given set of individuals and cluster-level factors.6
Because of shared behaviors and living conditions, family members within a household are likely to be more similar to one another than are family members from different households. This potential similarity of familial members within a household often leads to intraclass correlations of HCV seropositivity. The nesting of contacts within families characterizes a two-stage cluster sample in a cross-sectional study of HCV seropositivity and related exposures. Therefore, intra-cluster correlation arising from this type of sampling design needs to be addressed in a statistical analysis of the relationship between exposure and HCV seropositivity.7
We have previously reported a HCV seroprevalence of 20.5% among household contacts of HCV-seropositive thalassemic children in Karachi, Pakistan,8 which was higher than the 16% reported earlier in Pakistan9 and than the 2–8% reported elsewhere.10–12 We identified "sharing toothbrush with carrier" or "bitten by carrier" to be significantly associated with HCV seropositivity among contacts,8 but these findings remain to be verified. Furthermore, our previous analysis of the data and aforementioned studies did not take into account variance in the structure of HCV seropositivity at the household level while evaluating risk factors. We were therefore interested in determining the independent relationship of contacts’ HCV seropositivity with variables measured at the household level. In this contextual approach, we used multilevel logistic regression analysis, which accounts for information at both the individual level and the cluster level within the same model.13 The objective of this study was to investigate whether contacts of HCV seropositive index patient living in the same household have the similar probabilities of being HCV seropositive as related to any of the household-level variables, after taking into account the independent effects of individual-level variables.
|
SUBJECTS AND METHODS
|
Details for enrolling HCV-seropositive thalassemic children as index cases with their corresponding family members are described elsewhere.8 Briefly, index cases for this study were enrolled from a blood bank that is being run by a nongovernmental organization (NGO) in Karachi. Established in 1979, this blood bank provides cost-free freshly screened blood to poor and needy patients with leukemia, hemophilia, thalassemia, and other blood-related diseases. A total of 86 HCV-seropositive thalassemic children were selected as index patients out of 256 consecutive thalassemic children registered at the NGO’s blood bank for routine blood transfusion between November 1998 and January 1999. The probable source of HCV infection among these index children was repeated blood transfusion as second-generation ELISA was used to screen donors at the NGO’s blood bank; some donors may have been incubating or were in an early phase of HCV infection and tested negative at the time of screening. It is also possible that a previously HCV-infected family member might have infected the child, although this is less likely than a HCV-seropositive child transmitting HCV to a family member in our study population.
Eighty-six family units were identified on the basis of the HCV-seropositive thalassemic index children. A family unit was defined as husband, wife, children, and other relatives all living in the same household and sharing a kitchen. Familial contacts of these family units were contacted and invited to participate in the study. For HCV investigations, we studied up to 5 household contacts of age 6 years or more from each selected household between April 1999 and June 1999. In all, 341 family members were interviewed using a pretested questionnaire at the time of blood sample collection by a medical graduate. Children up to 12 years old responded to the interview questions with parental or guardian assistance, and subjects older than 12 years old were interviewed in private. Subjects were informed of their test results and implications at the end of the study. We obtained Institutional Ethical Review Committee approval and informed consent from the participating subjects.
The questionnaire was composed of questions regarding individual-level and household-level variables. Individual-level variables recorded on participating contacts included demographic variables (age, education, relationship to index patient), household behaviors (sharing towels, sharing gum/sweets with carrier, sharing partially eaten food with carrier, sharing toothbrush with carrier, bitten by the carrier, scratched the back of carrier, handling blood-tinged clothes of carrier, kissing of carrier), and extrafamilial factors (including various potential parenteral exposures to blood or blood products, e.g., past hospital admissions, operations, injuries needing hospital interventions, blood, or blood product transfusions, and history of parenteral injections, intravenous drips received, and dental treatments) known to be associated with HCV transmission and were compiled from literature. Household-level variables included mother tongue (proxy for ethnicity), total monthly income, family size, age of index patient, sex of index patient, and RNA status (positive or negative) of the index patient.
HCV serostatus of each participant was determined by an HCV microparticle enzyme immunoassay (MEIA) third-generation kit, used according to the manufacturer’s instructions (Abbot, Chicago, IL). The Amplicor HCV RNA assay was used to detect HCV RNA in the index patient’s serum and in the sera of corresponding HCV-seropositive household contacts in case the index patient was HCV RNA positive (Roche Diagnostic Systems, Indianapolis, IN).
Because of the hierarchical structure of the data, nonsexual contacts of HCV-infected thalassemic index patients nested within households, and the possibility of intrahousehold correlation regarding the likelihood of HCV seropositivity, we used multilevel logistic regression analysis.14 Three consecutive multilevel models were fitted to the data. In model 1 (empty model), the probability of a household contact being HCV seropositive was only a function of contact’s household and modeled by a random intercept. Model 2 included individual contact-level demographic variables and behaviors hypothesized to increase the risk of acquisition of HCV infection from the index patient. Model 3 expanded model 2 to include contextual variables measured at household level. With this analytical approach, we intended to quantify differences between various households (model 1) and to estimate the role played by individual characteristics of contacts from each household (model 2). Finally, in model 3, we examined household-level variables for their influence on HCV serostatus of contacts of the infected index patients after adjustment for individual-level variables.
To estimate fixed effects, we quantified the association of the demographic variables and risk behaviors with HCV serostatus of contact by odds ratios (OR) along with their 95% confidence intervals (CI). To estimate the random effects, the household was examined for a general contextual effect on the HCV serostatus of contact individuals in the household. The purpose was to determine whether individuals living in the same family unit shared a similar probability of being HCV seropositive after adjusting for the individual characteristics included in this study. Variance partition coefficient (VPC) and median OR (MOR) were used to assess the contextual phenomena. We also applied 80% interval OR (80% IOR), which integrates random effects (household variance) in the measurement of fixed effects (i.e., household variables). Detailed mathematics and descriptions of VPC, MOR, and IOR are found elsewhere.15,16 Briefly, VPC is the proportion of the total variance (i.e., the variance at the household level plus the variance at the individual level) at the household level (i.e., VPC = Vh/(Vh + Vi); where Vi = variance between the contacts from the same household [1st-level variance] and Vh is the variance between households [2nd-level variance]). A high VPC value would imply that households are very important in understanding the prevalence of HCV seropositivity among contacts; a low VPC value (VPC 
0) would suggest that households are similar with respect to contacts’ samples and that households are irrelevant for understanding differences (variation) in HCV seropositivity among contacts.
MOR measured the household variance in the OR scale. If MOR = 1 (no household level variance), then there is no difference between households in HCV seroprevalence. However, MOR > 1 signifies the importance of household contextual effects in understanding the individual contact’s probability of being HCV seropositive.
The IOR takes into account the magnitude of variation in HCV seroprevalence between households when interpreting the influence of household variables. IOR as an 80% interval is recommended, but note that IOR is not an ordinary confidence interval. If the IOR contains 1, the remaining unexplained variation between households regarding HCV seropositivity is large compared with the effect of the household variable. However, if the IOR does not contain 1, then the effect of household variables is large compared with unexplained differences between households. MLwiN software, version 1.1, was used for all analyses.17
|
RESULTS
|
In this study, HCV seroprevalence among household contacts was 20.5% (70/341). HCV seroprevalence did not vary statistically significantly across various categories of age (P = 0.331), education (P = 0.337), or type of relationship (P = 0.738) of household contacts with index patients. At the contextual level, although statistically nonsignificant (P = 0.143), the HCV seroprevalence was higher among contacts if the index patient was less than 5 years of age. Also, the HCV seroprevalence among contacts was significantly higher if the index patient was male (P = 0.046), HCV-RNA positive (P = 0.019), or if the household contained 2–4 individuals (P = 0.013). However, house ownership was not statistically significantly (P = 0.864) related with HCV serostatus of household contacts (Table 1
).
View this table:
[in this window]
[in a new window]
|
TABLE 1 Hepatitis C virus seropositivity in household contacts of 86 index HCV seropositive thalassemic patients in a cross-sectional study, Karachi, Pakistan, April 1999–June 1999 (n = 341)
|
|
Multilevel logistic regression analysis of fixed effects revealed that contacts were significantly more likely to be HCV seropositive if they ever shared a toothbrush with the carrier (OR = 2.89; 95% CI: 1.45 to 3.12) or if they had ever been bitten by the carrier (OR = 2.01; 95% CI: 1.14 to 3.55) (Table 2, model 2). Over and above individual-level variables, household contacts living in families wherein the index thalassemic patient was HCV RNA positive compared with those contacts living in families wherein the index thalassemic patient was HCV RNA negative had higher odds of being HCV positive (OR = 2.09; 95% CI: 1.02 to 4.28). However, IOR-80 was very wide (0.23 to 19.25) and contained 1 (Table 2, model 3). Therefore, the contextual effect of the index patient’s RNA status on the contact’s HCV serostatus was not strong enough when compared with the residual household-level heterogeneity. The contextual effects of the sex of the index patient and family size were not significant when we tried to include them in the final model.
View this table:
[in this window]
[in a new window]
|
TABLE 2 Measures of association between individual and household varaibles and the outcome variable and measures of clustering in the HCV seropositivity obtained from multilevel logistic regression models, Karachi, Pakistan, 2002
|
|
Multilevel logistic regression analysis of random effects showed that household intercept variance (± SE) in models 1–3 was 1.844 ± 0.463, 1.78 ± 0.463, and 1.505 ± 0.438, respectively, and was substantially large in each case. Consequently, MORs and VPCs in all three models were also large (Table 2
). In addition, as noted earlier, the 80% IOR included 1 for the index patient’s HCV RNA status, thus confirming the low importance of this contextual effect in the present study.
|
DISCUSSION
|
Research on the role of the familial environment in the spread of HCV has been inconclusive.18 Reportedly, intrafamilial nonsexual HCV transmission is uncommon and related to the length of exposure and the stage of HCV infection in the "source" subjects,19,20 whereas other researchers demonstrated a relatively high incidence of intrafamilial HCV transmission.21 We estimated 20.5% HCV seroprevalence among nonsexual household contacts of HCV-seropositive index thalassemic patients; this value may be is the highest figure yet reported in this risk group.8
Multilevel analyses typically quantify higher-level variations (at cluster level), some of which can be explained by the fixed effects of individual-level and cluster-level factors.22 Our main motivation for this evaluation was to see if, after taking into account the effects of individual-level characteristics, any fixed effect at the household-level could further explain any residual variation in HCV seropositivity in household contacts of HCV-seropositive thalassemic index children.
Multilevel analysis results showed that household-level fixed effects of an index patient’s HCV RNA status was important in explaining variability in HCV seropositivity among contacts. Household contacts living in families wherein the index thalassemic patient was HCV RNA positive compared with those contacts living in families wherein the index thalassemic patient was HCV RNA negative had higher odds of being HCV positive. However, IOR-80 values indicated that the effect of index patients’ RNA status on the contacts’ HCV serostatus was small in comparison with un-explained between-cluster variations. Discussions continue about such factors as length of exposure of contact, stage of HCV infection, and viral load in "source" subjects, all of which may influence the frequency of nonsexual familial HCV transmission.4,19 We are unaware of any published study that has looked at the contextual effects of the index patient’s HCV RNA status in multilevel analysis while discussing HCV seroprevalence among nonsexual familial contacts; therefore, further research is needed to explore this contextual effect.
With regard to household-level random effects, the high importance of the household environment in HCV transmission through nonsexual contact is further supported by the high household variances and resultantly high MORs in the analyses. These measures of clustering along with high VPCs underscore the fact that there are other un-recognized modes of nonsexual intrafamilial HCV transmission.
In conclusion, the results of our previous and current evaluations are pertinent for household members of HCV-infected patients; specifically, their close contact with objects contaminated with blood or perhaps with saliva of the HCV-seropositive index patient may pose an increased risk of HCV transmission. The HCV RNA positive status of the HCV-seropositive index patient seems to play role in household contacts’ HCV seropositivity. However, high household intercept variances in different analyses showed that there are still un-recognized nonsexual modes of HCV transmission at the household-level that need further research.
Received May 4, 2006.
Accepted for publication October 30, 2006.
* Address correspondence to Saeed Akhtar, PhD, Department of Community Medicine & Behavioural Sciences, Faculty of Medicine, Kuwait University, P.O. Box 24923, Safat 13110, Kuwait. E-mail: saeed.akhtar{at}hsc.edu.kw, akhtarsd{at}gmail.com 
Authors’ addresses: Saeed Akhtar, Department of Community Medicine and Behavioural Sciences, Faculty of Medicine, Kuwait University, P.O. Box 24923, Safat 13110, Kuwait, and Department of Community Health Sciences, Aga Khan University, Stadium Road, Karachi 74800, Pakistan, E-mail: saeed.akhtar{at}hsc.edu.kw, akhtarsd{at}gmail.com. Tariq Moatter, Department of Pathology and Microbiology, Aga Khan University, Stadium Road, Karachi 74800, Pakistan.
|
REFERENCES
|
- Tibbs CJ, 1995. Methods of transmission of hepatitis C. J Viral Hepat 2: 113–119.[Web of Science][Medline]
- Ackerman Z, Ackerman E, Paltiel O, 2000. Intrafamilial transmission of hepatitis C virus: a systematic review. J Viral Hepat 7: 93–103.[Medline]
- Everhart JE, Di Bisceglic AM, Murray LM, Alter HJ, Melpolder JJ, Kuo G, Hoofnagle JA, 1990. Risk for non-A, non-B (type C) hepatitis through sexual or household contact with chronic carriers. Ann Intern Med 12: 544–545.
- Velasco M, Poniachik J, Alegria S, Brahm J, 2003. Familial transmission of hepatitis C virus infection in Chilean subjects. Rev Med Chil 131: 1123–1127.[Web of Science][Medline]
- Akhtar S, Moatter T, 2004. Intra-household clustering of hepatitis C virus infection in Karachi, Pakistan. Trans R Soc Trop Med Hyg 98: 535–539.[Web of Science][Medline]
- Merlo J, 2003. Multilevel analytical approaches in social epidemiology: measures of health variation compared with traditional measures of association. J Epidemiol Community Health 57: 550–552.[Free Full Text]
- Korn EL, Garubard BI, 1999. Analysis of Health Surveys. First edition. New York: John Wiley & Sons.
- Akhtar S, Moatter T, Azam I, Rahbar MH, Adil S, 2002. Prevalence and risk factors for intrafamilial transmission of hepatitis C virus in Karachi, Pakistan. J Viral Hepat 9: 309–314.[Web of Science][Medline]
- Pasha O, Luby SP, Khan AJ, Shah SA, McCormick JB, Fisher-Hoch SP, 1999. Household members of hepatitis C virus-infected people in Hafizabad, Pakistan: infections by injections from health care providers. Epidemiol Infect 123: 515–518.[Medline]
- Demelia L, Vallebona E, Poma R, Sanna G, Masia G, Coppola RC, 1996. HCV transmission in family members of subjects with HCV related chronic liver disease. Eur J Epidemiol 12: 45–50.[Web of Science][Medline]
- Saltoglu N, Tasova Y, Burgut R, Dundar IH, 1998. Sexual and non-sexual intrafamilial spread of hepatitis C virus: intrafamilial transmission of HCV. Eur J Epidemiol 14: 225–228.[Web of Science][Medline]
- Brusaferro S, Barbone F, Adrian P, 1999. A study on the role of the family and other risk factors in HCV transmission. Eur J Epidemiol 15: 125–132.[Medline]
- Duncan C, Jones K, Moon G, 1998. Context, composition and heterogeneity: using multilevel models in health research. Soc Sci Med 46: 97–117.[Web of Science][Medline]
- Leyland AH, Goldstein H, 2001. Multilevel Modeling of Health Statistics. Chichester, United Kingdom: Wiley.
- Larsen K, Petersen JH, Budtz-Jorgensen E, Endahl L, 2000. Interpreting parameters in the logistic regression model with random effects. Biometrics 56: 909–914.[Web of Science][Medline]
- Larsen K, Merlo J, 2005. Appropriate assessment of neighborhood effects on individual health: integrating random and fixed effects in multilevel logistic regression. Am J Epidemiol 161: 81–88.[Abstract/Free Full Text]
- Rasbash J, Browne W, Goldstein H, Yang M, Plewis I, Healy M, Woodhouse G, Draper D, Langford I, Lewis T, 2002. A User’s Guide to MLwiN. Version 2.1. Centre of Multilevel Modelling, Institute of Education, University of London.
- Caporaso N, Ascione A, D’Antonio M, DiCostanzo GG, Galeota Lanza A, Tremolada F, Diodati G, Rumi MG, Parravicini ML, Pastore G, 1995. Prevalence of anti-HCV among spouses and offspring of anti-HCV positive subjects: an Italian multicentre study. Ital J Gastroenterol 27: 5–7.[Web of Science][Medline]
- Karmochkine M, Carrat F, Valleron AJ, Raguin G, 1998. Transmission modes of hepatitis C virus. Presse Med 27: 871–876.[Web of Science][Medline]
- Kim YS, Chi HS, Ahn Y, Lee H, Klag MJ, 1998. Lack of familial clustering of hepatitis C virus infection. Int J Epidemiol 27: 525–529.[Abstract/Free Full Text]
- Matyla-Radzewska A, Wysocki J, 2005. Natural course of infection in children and familial transmission of C viral hepatitis. Przegl Epidemiol 59: 475–481.[Medline]
- Goldstein H, Browne W, Rasbash J, 2002. Multilevel modeling of medical data. Stat Med 21: 3291–3315.[Web of Science][Medline]
TOP
ABSTRACT
INTRODUCTION
