Despite the initiative by WHO and other international organizations to eliminate HCV in the medium term, hepatitis C infection is still a major public health problem. Even non-injecting drugs users who engage in harmful or addictive drug use are at greater risk of acquiring the infection, when compared to the general population. This study evaluate risk factors for HCV infection in users of crack/cocaine in Brazil, using multilevel models that incorporate variations in the sensitivity and specificity of the respective diagnostic tests. The sample included all the participants of a national survey on street crack cocaine users with serologically reactive result in the rapid test for the HCV as well as 4 non-reactive controls, matched by sex, age category, and major geographic region of residence. Multilevel logistic regression models were used, with and without incorporation of the diagnostic test's sensitivity and specificity values. The odds of HCV infection were 85% higher among polydrug users, 7.81 times higher among injecting drug users, and 3.69 times higher in those reporting to have genital ulcers. Statistical modeling strategies that incorporate the sensitivity and specificity of diagnostic tests in challenging settings are useful for studying the association between risk factors and infection status.