Use of an algorithm applied to urine drug screening to assess adherence to a hydrocodone regimen

Using machine learning to model dose-response relationships

RATIONALE, AIMS AND OBJECTIVES

Establishing the relationship between various doses of an exposure and a response variable is integral to many studies in health care. Linear parametric models, widely used for estimating dose-response relationships, have several limitations. This paper employs the optimal discriminant analysis (ODA) machine-learning algorithm to determine the degree to which exposure dose can be distinguished based on the distribution of the response variable. By framing the dose-response relationship as a classification problem, machine learning can provide the same functionality as conventional models, but can additionally make individual-level predictions, which may be helpful in practical applications like establishing responsiveness to prescribed drug regimens.

METHOD

Using data from a study measuring the responses of blood flow in the forearm to the intra-arterial administration of isoproterenol (separately for 9 black and 13 white men, and pooled), we compare the results estimated from a generalized estimating equations (GEE) model with those estimated using ODA.

RESULTS

Generalized estimating equations and ODA both identified many statistically significant dose-response relationships, separately by race and for pooled data. Post hoc comparisons between doses indicated ODA (based on exact P values) was consistently more conservative than GEE (based on estimated P values). Compared with ODA, GEE produced twice as many instances of paradoxical confounding (findings from analysis of pooled data that are inconsistent with findings from analyses stratified by race).

CONCLUSIONS

Given its unique advantages and greater analytic flexibility, maximum-accuracy machine-learning methods like ODA should be considered as the primary analytic approach in dose-response applications.

Receiver-operating characteristics of adjusted urine measurements of oxycodone plus metabolites to distinguish between three different rates of oxycodone administration

OBJECTIVES

In a study by Couto et al. (Use of an algorithm applied to urine drug screening to assess adherence to an oxycontin regimen. J Opioid Manag 2009;5:359-64), adjusted urine measurements of oxycodone plus metabolites noroxycodone and oxymorphone were determined among volunteer subjects in three groups according to oxycodone administration rates (A: 80 mg/day; B: 160 mg/day; C: 240 mg/day). We performed receiver-operating characteristic (ROC) analyses of the distribution data from this study to determine the ability to correctly categorize individual measurements with respect to each group.

DESIGN AND METHODS

For groups A-C, assumed reference ranges were defined as median-centered intervals encompassing a designated central percentage of the group's distribution. By varying assumed reference ranges across all possible reference ranges, ROC analyses of the ability of each group's reference ranges to appropriately include or exclude members of all groups were calculated. This generated six ROC curves (sensitivity vs. specificity): A vs. (B or C); B vs. (A or C); C vs. (A or B).

RESULTS

Overlaps of distributions A, B, and C were large, such that none of the ROC curves exceeded areas-under-curves of 0.8. The greatest sensitivity-specificity combination had a sensitivity of 74% for C with specificity of 75% for A, for which oxycodone administration rates were different by a factor of 3.

CONCLUSIONS

ROC analyses of data from a previous study demonstrated that, even under experimentally controlled conditions, adjusted urine drug measurements could not be used reliably to correctly categorize individual subjects' results according to their known oxycodone administration rates in the range of 80-240 mg/day. Misclassifications of results were 25% or greater.