BACKGROUND. Clinical laboratory tests are now being prescribed and made directly available to consumers through retail outlets in the USA. Concerns with these test have been raised regarding the uncertainty of testing methods used in these venues and a lack of open, scientific validation of the technical accuracy and clinical equivalency of results obtained through these services.
METHODS. We conducted a cohort study of 60 healthy adults to compare the uncertainty and accuracy in 22 common clinical lab tests between one company offering blood tests obtained from finger prick (Theranos) and 2 major clinical testing services that require standard venipuncture draws (Quest and LabCorp). Samples were collected in Phoenix, Arizona, at an ambulatory clinic and at retail outlets with point-of-care services.
RESULTS. Theranos flagged tests outside their normal range 1.6× more often than other testing services (P < 0.0001). Of the 22 lab measurements evaluated, 15 (68%) showed significant interservice variability (P < 0.002). We found nonequivalent lipid panel test results between Theranos and other clinical services. Variability in testing services, sample collection times, and subjects markedly influenced lab results.
CONCLUSION. While laboratory practice standards exist to control this variability, the disparities between testing services we observed could potentially alter clinical interpretation and health care utilization. Greater transparency and evaluation of testing technologies would increase their utility in personalized health management.
FUNDING. This work was supported by the Icahn Institute for Genomics and Multiscale Biology, a gift from the Harris Family Charitable Foundation (to J.T. Dudley), and grants from the NIH (R01 DK098242 and U54 CA189201, to J.T. Dudley, and R01 AG046170 and U01 AI111598, to E.E. Schadt).
Brian A. Kidd, Gabriel Hoffman, Noah Zimmerman, Li Li, Joseph W. Morgan, Patricia K. Glowe, Gregory J. Botwin, Samir Parekh, Nikolina Babic, Matthew W. Doust, Gregory B. Stock, Eric E. Schadt, Joel T. Dudley
Percent of the variability explained by each component of the LMM (testing service, subject, blood collection time, and “other”, i.e., residuals, unexplained by other covariates). Subject variability — not shown explicitly — makes up the remaining difference from 100% (e.g., MCHC = 23%, hs-CRP = 98%). The interservice variability can also be interpreted in terms of the intraclass correlation. For example, consider the lab test for MCHC. Variation across testing service explains 65% of the overall observed variance. This result is equivalent to saying that, after correcting for the effects of subject and collection time, the correlation between samples from different services is 65%.