Bayesian hierarchical models for disease mapping applied to contagious pathologies.

Disease mapping aims to determine the underlying disease risk from scattered epidemiological data and to represent it on a smoothed colored map. This methodology is based on Bayesian inference and is classically dedicated to non-infectious diseases whose incidence is low and whose cases distribution is spatially (and eventually temporally) structured. Over the last decades, disease mapping has received many major improvements to extend its scope of application: integrating the temporal dimension, dealing with missing data, taking into account various a prioris (environmental and population covariates, assumptions concerning the repartition and the evolution of the risk), dealing with overdispersion, etc. We aim to adapt this approach to model rare infectious diseases proposing specific and generic variants of this methodology. In the context of a contagious disease, the outcome of a primary case can in addition generate secondary occurrences of the pathology in a close spatial and temporal neighborhood; this can result in local overdispersion and in higher spatial and temporal dependencies due to direct and/or indirect transmission. In consequence, we test models including a Negative Binomial distribution (instead of the usual Poisson distribution) to deal with local overdispersion. We also use a specific spatio-temporal link in order to better model the stronger spatial and temporal dependencies due to the transmission of the disease. We have proposed and tested 60 Bayesian hierarchical models on 400 simulated datasets and bovine tuberculosis real data. This analysis shows the relevance of the CAR (Conditional AutoRegressive) processes to deal with the structure of the risk. We can also conclude that the negative binomial models outperform the Poisson models with a Gaussian noise to handle overdispersion. In addition our study provided relevant maps which are congruent with the real risk (simulated data) and with the knowledge concerning bovine tuberculosis (real data).

Comparative population genomics of the Borrelia burgdorferi species complex reveals high degree of genetic isolation among species and underscores benefits and constraints to studying intra-specific epidemiological processes.

Lyme borreliosis, one of the most frequently contracted zoonotic diseases in the Northern Hemisphere, is caused by bacteria belonging to different genetic groups within the Borrelia burgdorferi species complex, which are transmitted by ticks among various wildlife reservoirs, such as small mammals and birds. These features make the Borrelia burgdorferi species complex an attractive biological model that can be used to study the diversification and the epidemiology of endemic bacterial pathogens. We investigated the potential of population genomic approaches to study these processes. Sixty-three strains belonging to three species within the Borrelia burgdorferi complex were isolated from questing ticks in Alsace (France), a region where Lyme disease is highly endemic. We first aimed to characterize the degree of genetic isolation among the species sampled. Phylogenetic and coalescent-based analyses revealed clear delineations: there was a ∼50 fold difference between intra-specific and inter-specific recombination rates. We then investigated whether the population genomic data contained information of epidemiological relevance. In phylogenies inferred using most of the genome, conspecific strains did not cluster in clades. These results raise questions about the relevance of different strategies when investigating pathogen epidemiology. For instance, here, both classical analytic approaches and phylodynamic simulations suggested that population sizes and migration rates were higher in B. garinii populations, which are normally associated with birds, than in B. burgdorferi s.s. populations. The phylogenetic analyses of the infection-related ospC gene and its flanking region provided additional support for this finding. Traces of recombination among the B. burgdorferi s.s. lineages and lineages associated with small mammals were found, suggesting that they shared the same hosts. Altogether, these results provide baseline evidence that can be used to formulate hypotheses regarding the host range of B. burgdorferi lineages based on population genomic data.

Classification method for disease risk mapping based on discrete hidden Markov random fields.

Risk mapping in epidemiology enables areas with a low or high risk of disease contamination to be localized and provides a measure of risk differences between these regions. Risk mapping models for pooled data currently used by epidemiologists focus on the estimated risk for each geographical unit. They are based on a Poisson log-linear mixed model with a latent intrinsic continuous hidden Markov random field (HMRF) generally corresponding to a Gaussian autoregressive spatial smoothing. Risk classification, which is necessary to draw clearly delimited risk zones (in which protection measures may be applied), generally must be performed separately. We propose a method for direct classified risk mapping based on a Poisson log-linear mixed model with a latent discrete HMRF. The discrete hidden field (HF) corresponds to the assignment of each spatial unit to a risk class. The risk values attached to the classes are parameters and are estimated. When mapping risk using HMRFs, the conditional distribution of the observed field is modeled with a Poisson rather than a Gaussian distribution as in image segmentation. Moreover, abrupt changes in risk levels are rare in disease maps. The spatial hidden model should favor smoothed out risks, but conventional discrete Markov random fields (e.g. the Potts model) do not impose this. We therefore propose new potential functions for the HF that take into account class ordering. We use a Monte Carlo version of the expectation-maximization algorithm to estimate parameters and determine risk classes. We illustrate the method's behavior on simulated and real data sets. Our method appears particularly well adapted to localize high-risk regions and estimate the corresponding risk levels.

Dyslipidaemia in HIV-infected women on antiretroviral therapy. Analysis of 922 patients from the Spanish VACH cohort.

BACKGROUND: Information concerning lipid disturbances in HIV-infected women on antiretroviral therapy (ART) is scarce. The objective of the study is to describe the lipid profile in a large cohort of HIV-infected women on contemporary ART and analyse differences between regimes and patient's characteristics. METHODS: Observational, multicentre, cross-sectional study from the Spanish VACH Cohort. 922 women on stable ART without lipid-lowering treatment were included. RESULTS: Median age was 42 years, median CD4 lymphocyte count was 544 cells/mm3, and 85.6% presented undetectable HIV-1 viral load. Median total cholesterol (TC) was 189 mg/dL (interquartile range, IQR, 165-221), HDL cholesterol 53 mg/dL (IQR, 44-64), LDL cholesterol 108 mg/dL (IQR, 86-134), and triglycerides 116 mg/dL (IQR, 85-163). Mean accumulated time on ART was 116 months; 47.4% were on NNRTI-based regimes, 44.7% on PI, and 6.7% on only-NRTI therapy. 43.8% were also hepatitis C (HCV) coinfected. Patients on PI treatment presented higher TC/HDL ratio than those on NNRTI (p < 0.001). Significantly higher HDL values were observed in NNRTI-treated patients. HCV-coinfected patients presented lower TC/HDL ratio than the non HCV-coinfected. In multivariate analysis, factors independently associated with TC/HDL ratio were age, triglyceride levels and HCV co-infection. PI treatment presented a non-significant association with higher TC/HDL ratio. CONCLUSIONS: In HIV-infected women, the NNRTI-based ART is associated with a better lipid profile than the PI-based. Factors unrelated to ART selection may also exert an independent, significant influence on lipids; in particular, age, and triglyceride levels are associated with an increased TC/HDL ratio while HCV co-infection is associated with a reduced TC/HDL ratio.

Characteristics and outcome of HIV infection in gypsies in the Spanish VACH Cohort

Objective To study the characteristics of HIV infection in the gypsy (Roma) population in Spain, as compared with those of the Caucasian, non-gypsy majority. Design Cross-sectional, historical cohort study from the Spanish VACH Cohort. Methods Patients attending VACH clinics between 1 June 2004 and 30 November 2004 were classified according to their racial and ethnic origin as “gypsies”, Caucasian non-gypsy Spanish natives (CNGN), and “other” (the last being excluded from this study). Their sociodemographic and clinico-epidemiological characteristics were compared, as well as the Kaplan–Meier curves of time to AIDS, or death, or disease progression (either of the 2 outcomes).Results4819 (48%) of 10,032 cases included in the VACH database were eligible: 210 (4.2%) were gypsies and 4252 (84.8%) were CNGN. Differences were observed in age, household, academic, inmate, marital, and employment history. Injecting drug use had been the most frequent mechanism of transmission in both groups, but to a greater extent among gypsies (72% versus 50%; P<0.000). Sex distribution, CD4 cell counts, and viral loads at the first visit were similar in the 2 groups, as was the percentage of patients with previous AIDS, percentage receiving antiretrovirals, and percentage subsequently starting antiretroviral therapy. Up to 1 April 2005, 416 new AIDS cases and 85 deaths were recorded. The percentage of these outcomes did not differ between groups, but log-rank test showed a shorter time to AIDS and disease progression among gypsies. Conclusions The sociodemographic characteristics of gypsies, the largest minority in the VACH Cohort, show differences relative to those of CNGN. HIV-related outcomes suggest that gypsies have a poorer prognosis (AU)

Objetivo estudiar las características de la infección por el VIH en gitanos en España, en comparación con las de la mayoría caucásica no gitana (CNG).Métodos estudio transversal y de cohortes históricas en la Cohorte VACH. Clasificamos a los pacientes que acudieron a las clínicas participantes en VACH entre el 1 de junio de 2004 y el 30 de noviembre de 2004 de acuerdo a su raza y etnia, como «gitanos», «nativos españoles CNG» u «otros» (estos, excluidos de este estudio). Comparamos sus características sociodemográficas y clinicoepidemiológicas, así como sus curvas de Kaplan–Meier del tiempo hasta sida, muerte o progresión de la enfermedad (cualquiera de ambos).Resultados4819 (48%) de 10.032 casos recogidos en la base de datos de VACH fueron incluidos en el estudio: 210 (4,2%) eran gitanos y 4.252 (84,8%) eran nativos CNG. Observamos diferencias en sus distribuciones por edad, domicilio, estudios, antecedentes penales, situación laboral y marital. La inyección de drogas había sido el mecanismo de transmisión del VIH más frecuente en los dos grupos, pero más marcadamente en los gitanos (72% frente a 50%; p<0,000). La distribución por sexos, los recuentos de linfocitos CD4 y las cargas virales en la primera visita fueron similares en ambos grupos, así como las proporciones de pacientes con sida previo y las de quienes estaban ya en, o iniciaron entonces, tratamiento antirretroviral. Hasta el 1 de abril de 2005 se registraron 416 nuevos casos de sida y 85 muertes. La proporción de ambos resultados fue similar en ambos grupos, pero la prueba del rango logarítmico demostró una evolución más rápida a sida y a progresión de la (..) (AU)

Characteristics and outcome of HIV infection in gypsies in the Spanish VACH Cohort.

OBJECTIVE: To study the characteristics of HIV infection in the gypsy (Roma) population in Spain, as compared with those of the Caucasian, non-gypsy majority. DESIGN: Cross-sectional, historical cohort study from the Spanish VACH Cohort. METHODS: Patients attending VACH clinics between 1 June 2004 and 30 November 2004 were classified according to their racial and ethnic origin as "gypsies", Caucasian non-gypsy Spanish natives (CNGN), and "other" (the last being excluded from this study). Their sociodemographic and clinico-epidemiological characteristics were compared, as well as the Kaplan-Meier curves of time to AIDS, or death, or disease progression (either of the 2 outcomes). RESULTS: 4819 (48%) of 10,032 cases included in the VACH database were eligible: 210 (4.2%) were gypsies and 4252 (84.8%) were CNGN. Differences were observed in age, household, academic, inmate, marital, and employment history. Injecting drug use had been the most frequent mechanism of transmission in both groups, but to a greater extent among gypsies (72% versus 50%; P<0.000). Sex distribution, CD4 cell counts, and viral loads at the first visit were similar in the 2 groups, as was the percentage of patients with previous AIDS, percentage receiving antiretrovirals, and percentage subsequently starting antiretroviral therapy. Up to 1 April 2005, 416 new AIDS cases and 85 deaths were recorded. The percentage of these outcomes did not differ between groups, but log-rank test showed a shorter time to AIDS and disease progression among gypsies. CONCLUSIONS: The sociodemographic characteristics of gypsies, the largest minority in the VACH Cohort, show differences relative to those of CNGN. HIV-related outcomes suggest that gypsies have a poorer prognosis.

Anthropogenic factors and the risk of highly pathogenic avian influenza H5N1: prospects from a spatial-based model.

Beginning in 2003, highly pathogenic avian influenza (HPAI) H5N1 virus spread across Southeast Asia, causing unprecedented epidemics. Thailand was massively infected in 2004 and 2005 and continues today to experience sporadic outbreaks. While research findings suggest that the spread of HPAI H5N1 is influenced primarily by trade patterns, identifying the anthropogenic risk factors involved remains a challenge. In this study, we investigated which anthropogenic factors played a role in the risk of HPAI in Thailand using outbreak data from the "second wave" of the epidemic (3 July 2004 to 5 May 2005) in the country. We first performed a spatial analysis of the relative risk of HPAI H5N1 at the subdistrict level based on a hierarchical Bayesian model. We observed a strong spatial heterogeneity of the relative risk. We then tested a set of potential risk factors in a multivariable linear model. The results confirmed the role of free-grazing ducks and rice-cropping intensity but showed a weak association with fighting cock density. The results also revealed a set of anthropogenic factors significantly linked with the risk of HPAI. High risk was associated strongly with densely populated areas, short distances to a highway junction, and short distances to large cities. These findings highlight a new explanatory pattern for the risk of HPAI and indicate that, in addition to agro-environmental factors, anthropogenic factors play an important role in the spread of H5N1. To limit the spread of future outbreaks, efforts to control the movement of poultry products must be sustained.

Bacterial meningitis in HIV-1-infected patients in the era of highly active antiretroviral therapy.

BACKGROUND: The burden that spontaneous bacterial meningitis (SBM) currently represents among HIV-1-infected patients is poorly known. METHODS: We prospectively evaluated 32 episodes of SBM in HIV-1-infected patients from the VACH (VIH-Aplicación de Control Hospitalario) Cohort and compared findings with those of 267 episodes in uninfected persons, matched by age and year of infection. A group of 13,187 HIV-1-infected patients from the VACH Cohort were used to identify predictors for acquiring SBM. RESULTS: Between 1997 and 2006, we found 32 episodes of SBM among HIV-1-infected patients for an annual incidence rate of 62.0 cases per 100,000 population compared with 3.2 (3.0 to 3.4) per 100,000 population for uninfected patients (P < 0.001). The last CD4 >or=200/mm count was the only predictor for developing SBM. Compared with uninfected, HIV-1-infected patients with SBM had a greater prevalence of primary extrameningeal infection, especially pneumonia (P = 0.02), bacteremia (P = 0.02), focal neurologic signs (P = 0.005), seizures (P = 0.06), a lower cerebrospinal fluid to blood glucose ratio (P = 0.02), and a lower prevalence of nuchal rigidity (P = 0.005). Streptococcus pneumoniae was the most frequent etiologic agent among HIV-1-infected patients. HIV-1-infected patients had neurologic complications more frequently (P = 0.02), a higher overall case fatality rate (P = 0.004), and greater incidence of neurologic sequelae (P = 0.001). CONCLUSIONS: Even in the highly active antiretroviral therapy era, the risk of developing SBM is 19 times higher among HIV-1-infected patients than among uninfected ones. It tends to present in severely immunosuppressed patients not previously vaccinated and off antiretroviral therapy, with a concomitant extrameningeal infection, bacteremia, and focal neurologic signs, and is caused by S. pneumoniae. SBM in HIV-1-infected patients carries a worse prognosis than in uninfected ones both in terms of lethality and sequelae.

Effectiveness and safety of didanosine, lamivudine and efavirenz versus zidovudine, lamivudine and efavirenz for the initial treatment of HIV-infected patients from the Spanish VACH cohort.

BACKGROUND: Preliminary data suggest that a once-daily combination of lamivudine, didanosine and efavirenz is an effective alternative regimen for antiretroviral-naive HIV-1-infected patients. However, data from randomized trials comparing this combination versus standard first-line regimens are not available yet. In an observational study, we analyse the efficacy and tolerability of didanosine plus lamivudine and efavirenz versus zidovudine plus lamivudine and efavirenz in a cohort of therapy naive patients. METHODS: We performed an observational study on prospectively collected data from patients participating in a multicentre Spanish treatment-naive cohort (VACH cohort). Efficacy was assessed comparing time to therapeutic failure and CD4 cell recovery. Safety was analysed comparing the proportion of patients who discontinued therapy for toxicity or any other reason. RESULTS: Overall, 219 patients treated with once-daily didanosine/lamivudine/efavirenz and 409 patients receiving twice-daily zidovudine/lamivudine (Combivir) plus efavirenz were evaluated. By intent-to treat analysis (non-completers and therapeutic change=failure), time to treatment failure was similar in both groups of treatment: 40.0 months (95% CI 23.3-56.8 months) among patients on didanosine/lamivudine/efavirenz and 33.3 months (95% CI 25.6-41.1 months) in patients treated with zidovudine/lamivudine/efavirenz (P=0.253). The risk of failure due to treatment change was almost double among patients treated with zidovudine/lamivudine/efavirenz compared with those who received didanosine/lamivudine/efavirenz. CONCLUSIONS: Our data suggest that didanosine/lamivudine/efavirenz is a combination with an efficacy comparable to zidovudine/lamivudine/efavirenz as first-line therapy for HIV infection. The risk of treatment change was significantly higher among patients treated with zidovudine/lamivudine/efavirenz than in those starting therapy with didanosine/lamivudine/efavirenz.

First-line antiretroviral therapy with efavirenz or lopinavir/ritonavir plus two nucleoside analogues: the SUSKA study, a non-randomized comparison from the VACH cohort.

BACKGROUND: Efavirenz and lopinavir/ritonavir are both recommended antiretroviral agents for combination first-line therapy, although information on direct comparisons between them is scarce. A retrospective longitudinal study from the VACH cohort comparing both regimens was performed. METHODS: Efficacy was examined comparing time to virological failure, CD4 recovery and clinical progression. Tolerability was examined comparing time to treatment discontinuation for any reason and for toxicity. Survival analysis was conducted using the Kaplan-Meier method, and standard and weighted Cox regression models. RESULTS: A total of 1550 antiretroviral-naive patients starting a two-nucleoside reverse transcriptase inhibitor regimen plus either efavirenz (n = 1159) or lopinavir/ritonavir (n = 391) were included in the study. At baseline, patients starting lopinavir/ritonavir had higher HIV-1 RNA and lower CD4+ cell counts. There was no difference in the adjusted hazards of virological failure [efavirenz versus lopinavir/ritonavir hazard ratio (HR) = 0.93, 95% confidence interval (CI): 0.77-1.12, P = 0.43], CD4 recovery (HR = 1.11, 95% CI: 0.95-1.30, P = 0.19) and clinical progression (HR = 0.71, 95% CI: 0.39-1.31, P = 0.27). There was an increased risk of discontinuation for any reason or for toxicity for lopinavir/ritonavir (HR = 2.10, 95% CI: 1.40-3.15, P = 0.0003). CD4 recovery with both drugs was also similar in the lowest CD4 strata. A higher risk of early hypertriglyceridaemia was associated with lopinavir/ritonavir-based regimens. CONCLUSIONS: Our study suggests similar virological efficacy for efavirenz- or lopinavir/ritonavir-based first-line antiretroviral regimens, but an increased risk of discontinuation because of toxicity in case of lopinavir/ritonavir-based therapy. Immunological outcome appeared similar with both regimens.

Bovine spongiform encephalopathy and spatial analysis of the feed industry.

In France, despite the ban of meat-and-bone meal (MBM) in cattle feed, bovine spongiform encephalopathy (BSE) was detected in hundreds of cattle born after the ban. To study the role of MBM, animal fat, and dicalcium phosphate on the risk for BSE after the feed ban, we conducted a spatial analysis of the feed industry. We used data from 629 BSE cases as well as data on use of each byproduct and market area of the feed factories. We mapped risk for BSE in 951 areas supplied by the same factories and connection with use of byproducts. A disease map of BSE with covariates was built with the hierarchical Bayesian modeling methods, based on Poisson distribution with spatial smoothing. Only use of MBM was spatially linked to risk for BSE, which highlights cross-contamination as the most probable source of infection after the feed ban.

Impact of lamivudine on the risk of liver-related death in 2,041 HBsAg- and HIV-positive individuals: results from an inter-cohort analysis.

BACKGROUND: The impact of lamivudine (3TC) as part of combination antiretroviral therapy (cART) on the risk of liver-related death (LRD) in HIV/hepatitis B virus (HBV)-coinfected patients has not been extensively studied. METHODS: We performed an analysis involving HIV/HBV-coinfected patients in 13 cohorts who initiated cART. The end-point was LRD--that is, death with concomitant decompensated liver disease (DLD) or hepatocellular carcinoma--as the main cause. Incidence rates of LRD after initiation of cART were expressed as number of events per 100 person-years of follow-up (PYFU). A Poisson regression model adjusted for cohort, gender, mode of HIV transmission, CD4+ T-cell count at cART initiation, liver disease pre-cART, duration of 3TC before cART, and hepatitis C virus was used to assess the association between use of 3TC and risk of LRD. RESULTS: We analysed 2,041 patients. Follow-up after starting cART was 7,648 PYFU (5,569 spent on 3TC-containing regimens) with a median per person of 48 months (range: 2-91). Of the total, 217 subjects died; 57 deaths were liver-related resulting in a rate of 7.5 per 1,000 PYFU [95% confidence intervals (CI): 5.6-9.7]. The relative risk of LRD per extra year of 3TC use was 0.73 (95% CI: 0.59-0.90, P = 0.004). CONCLUSION: The use of 3TC was associated with a reduced risk of LRD over 4 years of follow-up. This study supports the current view that the use of 3TC as part of cART should be considered in patients who are tested positive for HBsAg.

Hepatitis C virus and human immunodeficiency virus coinfection in Spain.

OBJECTIVE AND METHODS: In a cross-sectional study, based on a cohort composed of HIV-infected patients of fifteen tertiary level institutions of Spain, the main data of the entire cohort are described, characteristics of patients with or without hepatitis C coinfection are compared, and the possible association of hepatitis C virus coinfection with socioeconomic, HIV-related, and hepatitis B-related variables is assessed. RESULTS: A total of 4,709 patients are studied. Median of age is 37 years, 78.3% are male. HIV risk behaviours are: parenteral drug use in 63.8% of patients, heterosexual in 22.3%, and homosexual in 10.8%. Serology of hepatitis C is positive in 69.2% of participants. The following variables are associated with increased prevalence of hepatitis C coinfection, both in univariate and in multivariate analysis: HIV risk behaviour, positive anti-HBs, longer time elapsed since HIV infection diagnosis, younger age, lower social status, lower CD4 cell count increase between nadir and last available result, and lower educational level (all P<0.001). Patients with heterosexual behaviour are more frequently coinfected than patients with homosexual behaviour (P<0.001). CONCLUSION: This study highlights that, in Spain, more than two thirds of patients with HIV infection are coinfected with hepatitis C virus.