This first Brazilian nationwide hepatitis survey showed 1.38% HCV positivity in a representative sample of 19,503 adolescents and adults in all macro-regions. Approximately 1.3 million individuals would be expected to be anti-HCV-positive in the country under the assumption of similar anti-HCV antibody prevalence for the Brazilian population as a whole (population of 169.8 million, as of 2000). The current population-based survey detected a lower prevalence of anti-HCV antibody than that previously reported by the WHO for the country , shifting from intermediate to low prevalence.
The few nationwide population-based studies have reported prevalence of anti-HCV antibodies of 1.3% in the USA  and 1.0% in France , which are similar to that found in this study. An increase in anti-HCV prevalence with age was found in the large census surveys conducted in the USA. Peak HCV prevalence was found among adults aged 30-49 years in the first census survey (1988–1994), shifting to 40–49 years in the later surveys (1999–2002; 2003–2006; 2007–2008) . Although our results also showed increased HCV exposure with age, the peak prevalence was observed in the older age group. Increased HCV infection among the adult population has also been reported in selected regions of Brazil [11, 13] and among first-time blood donors .
The highest prevalence of the HCV marker (2.1%) was in the North macro-region of Brazil, which is below the 2.5% prevalence threshold for low endemicity according to WHO criteria . However, most of the estimated infected individuals resided in the Southeast region (~220,000), since approximately 14 million individuals aged 10 to 69 years live in these highly urbanized State capitals that include the cities of Sao Paulo and Rio de Janeiro. The burden of infection in absolute terms thus has an uneven geographical distribution, concentrated in the South and Southeast macro-regions. According to the official surveillance system, the residents of the southeast and south had higher HCV detection rates (above the national figures) than the residents in the north, northeast and central-west from 2003 through 2009 (unpublished data). One likely explanation for the higher detection rates in the south and southeast may be better access to and organization of the health system of these regions . The surveillance data for HCV may also underestimate the prevalence of HCV infection in the country as a whole, in view of the prevalence found by this population-based survey. Under-reporting of HCV infection by other national surveillance systems has also been discussed for Europe [30, 31] and for the USA , pointing to the need for updated population-based studies for public health purposes.
Interestingly, those participating in this national survey showed similarities in terms of the risk profile for HCV acquisition, with slightly higher frequencies for blood transfusion (7.8%), tattoos (12.2%) and sniffed drugs (5.9%) in the south macro-region (data not shown). In this household survey, injected and sniffed drug use was strongly associated with HCV antibodies in multivariate analysis. These findings accord with the literature regarding injected drug use as the main source of transmission of HCV infection in developed countries and with the reports of sniffed drugs as another important mode of HCV acquisition in developing countries, as previously reported in selected Brazilian populations . In the present survey, a history of previous hospitalization and use of injections with glass syringe not related to drug use were also associated with HCV seropositivity, both of which have been widely reported as epidemiological risk factors for HCV infection worldwide [1, 2]. In our survey, a socioeconomic marker of extreme poverty (no sewage disposal) remained associated with anti-HCV prevalence (OR = 2.53), after adjusting for confounding factors. Although there is no biological plausibility for social deprivation and increased risk of HCV infection, the former may be a surrogate marker for being at risk for HCV transmission. In a countrywide survey in France, social factors remained associated with HCV infection after adjusting for known risk factors as in our study. Nevertheless, having ever used drugs, having tattoos, and having being incarcerated were correlated with social deprivation . The USA HCV prevalence survey also reported a correlation of infection by drug use with low family income, high number of sex partners, and non-injected drug use . Furthermore, studies using cross-sectional designs such as our survey cannot sort out the temporal relationships between the potential risk factors and the outcomes.
The estimated population attributable risk indicates that the risk factors identified explain approximately 40% of all cases of infection. The higher population attributable risk percent found for “ever use of sniffed drugs” (7.2%) than for “ever use of injected drugs” (4.3%) was probably due to the lower frequency of IDUs than of sniffed drug users in the studied population. Although we cannot evaluate to what extent there was an underreporting of injection drug use because of stigma and discrimination in our study, the frequencies we found (data not shown) seem consonant with the national survey conducted by the Brazilian Center of Information on Psychotropic drugs (CEBRID) and a study among hospitalized patients [34, 35]. In relation to IDU, as the population attributable risk is related to the general population (injecting and non-injecting drug users), the attributable risk among the exposed (injecting drug users) will be greater and therefore the proportion of infection in that group that could be prevented by eliminating the exposure would be larger among IDUs.
One limitation of this survey is that it included a sampling of the capital cities of the five Brazilian macro-regions and the Federal District and was not designed as part of a census survey  as was NHANES . Therefore, the low prevalence of anti-HCV detected in all macro-regions may not reflect the levels of infection outside the study settings, such as in rural areas, and specific pockets. Since this investigation was planned as a population study, specific high-risk groups for HCV acquisition, such as institutionalized or incarcerated individuals, were excluded. Therefore, the prevalence reported here should be considered a conservative estimate of anti-HCV prevalence.
It should be noted that, in view of the low frequency of HCV antibodies for the country as a whole, the risk factor analysis was performed for the entire data set and not by macro-region. This had been anticipated by the study protocol .
Another methodological issue regarding the interpretation of our findings is that a sole marker of HCV infection may lead to misclassification, owing to false positive results. However, it would be a non-differential misclassification, underestimating the associations reported. For this reason, the strength of association may also be conservative . We acknowledge that the approximately 40% of HCV-RNA positivity among anti-HCV positive individuals was lower than expected. Potential explanations for this are poor specificity of the anti-HCV antibody assay, poor sensitivity of the HCV-RNA assay, or a true biologically higher HCV-RNA spontaneous clearance rate in the Brazilian setting. In our study, high sensitivity (poor specificity) of anti-HCV was identified because of the cutoff point recommended by the manufacturer; this may have increased false positive rates. In this regard, the ROC curve showed that a cutoff point of OD =4.0 yielded higher specificity on the detection of viremia. The assay used (AMPLICOR™ Roche Version 2.0) for HCV-RNA testing detects 60 International Units per ml. A concern on the detection of hepatitis C virus RNA in a large population survey is the specimen handling and storage. Nevertheless, we have followed the laboratory procedures and storage recommendations and all tests were performed in a reference laboratory where quality control was assured. In addition, data analysis showed that the frequency of viremia found was consistent across macro-regions. In general, about 25% of infected HCV individuals have spontaneous viral clearance. A systematic review of longitudinal studies on spontaneous viral clearance showed that clearance was more frequent in females and in acute clinical hepatitis C infection . Furthermore, a genome-wide association study showed that individuals with favorable IL28B genotypes are more likely to clear HCV infection than are those with unfavorable alleles, which could have implications in viral clearance ability among different ethnic groups . In the South of Brazil, IL28B polymorphism was associated with spontaneous clearance of hepatitis C infection among co-infected HCV and HIV/AIDS patients .
Although this survey covered a large geographical area and enrolled a large population sample in all State Capitals and Federal District, there was low refusal rate in all macro-regions, which was slightly higher in the wealthier regions: South and Southeast. This indicates the feasibility of nationwide-studies.
In this survey, subtypes 1a, 1b, 2b and 3a were detected in the general Brazilian population within the macro-regions. Overall, genotype 1 was predominant, as reported in the literature, which suggests that genotype 1a and 1b are the most frequent HCV genotypes worldwide, accounting for 60% of the infections . The geographical distribution of hepatitis C virus genotypes in Brazil among a large sample of chronic hepatitis C patients from all regions has also shown a predominance of genotype 1 followed by genotype 3 , in accordance with our findings. Among injecting drug-users, the circulating subtypes were 1 and 3 in the studies conducted between 1994–1997 and 1999–2001 in Rio de Janeiro . Our study was not designed to identify the diversity of HCV genotypes among the regions. We are aware that not all anti-HCV positive individuals were tested for HCV-RNA, which may limit the extrapolation of these results.