A comparison of five non-instrumented urine drug-testing devices was performed using a challenging clinical specimen set with drug concentrations close to the immunoassay screening cutoffs. The five devices were Syva RapidTest d.a.u. 8, Syva RapidCup d.a.u. 5, Roche TesTcup 5, Biosite Triage, and Casco-Nerl microLINE Drug Screen Card. Sixty clinical specimens for each of the five SAMHSA-specified drug categories were tested by both a scientist and a non-scientist with each result independently read by both. All specimens were also tested on a benchtop automated immunoassay analyzer (Syva ETS using Emit d.a.u. reagents) for comparison and by gas chromatography–mass spectrometry (GC–MS). The non-instrumented devices demonstrated an overall accuracy of 70% (66–74%), based on standard GC–MS confirmation cutoffs, comparable to the Syva ETS analyzer (80%). There was also little difference in overall accuracy between the scientist (71%) and non-scientist (69%), although the non-scientist reported 10 false-positive results (0.7% of 1490 total results or 3.8% of 260 results for drug-free specimens), and the scientist reported only 1 false-positive result (0.07% of 1490 total results or 0.38% of 260 results for drug-free specimens). When device performance was assessed according to drug presence/absence criteria, accuracy generally improved with all devices demonstrating extremely high positive predictive values (0.98–1).