Immunoassays are commonly used to screen samples prior to confirmation by gas chromatography-mass spectrometry (GC-MS). This serves two purposes: it provides a second method for positive samples, and it allows exclusion of negative samples from further confirmatory testing. In addition, immunoassay results can be used in some cases to determine if dilution of the sample will be required during the confirmatory assay. We used 878 sweat patches worn by 38 subjects receiving treatment for cocaine dependence to compare analysis of the extracts of the patches for cocaine immuno-equivalents by radioimmunoassay (RIA) with determination of cocaine, benzoylecgonine (BE), and ecgonine methyl ester (EME) by GC-MS. Preliminary validation experiments demonstrate that the GC-MS method using positive ion chemical ionization had sufficient specificity and recovery to support a lower limit of quantitation (LLOQ) of 4 ng/patch and was precise and accurate across a linear range up to 500 ng/patch. Cocaine ranging from the LLOQ to 31,900 ng/patch was found in 660 of the samples; BE ranging from the LLOQ to 3470 ng/patch was found in 530 of the samples; and EME ranging from the LLOQ to 2280 ng/patch was found in 476 of the samples. In a subset of 238 samples semiquantitative use of the RIA gave results that agreed with GC-MS with a correlation coefficient of 0.986, but averaged approximately 23% lower. Although this accuracy of the RIA supported its use as a sole quantitative assay, the limited linear range of the RIA (4-200 ng/patch) proved impractical for this purpose. Receiver operator characteristic analysis of the cutoffs of the RIA and GC-MS suggested optimal cutoffs of 5 and 4 ng/patch, respectively. At these cutoffs, the RIA had sensitivity of 90.0% and specificity of 92.2%. For samples that had RIA results greater than the high calibrator (N = 228), various dilution schemes were assessed for their ability to predict retention of either cocaine alone or cocaine and both metabolites within the linear range of the GC-MS. When cocaine was the only analyte of interest, a single 20-fold dilution retained 200 (87.7%) of the samples. This compared to an optimal scheme where a different dilution was selected for each one-tenth ratio (< 0.1, 0.1-0.2, etc.) where 211 (92.5%) of the samples were retained. When trying to retain cocaine and both metabolites, dilution schemes were less successful as BE and EME would often fall below the LLOQ of the GC-MS. A single fivefold dilution of all samples retained 115 (50.4%) compared to an optimum of 143 (62.7%). The optimum could be approached with four dilution sets retaining 142 of the samples. Time expended on performing RIA analysis of all the samples was cost-effective when the results were used to exclude negatives from and predict dilutions required for GC-MS analysis. RIA offers a cost-effective, sensitive, and specific alternative initial test for cocaine determination in extracts of sweat patches.