Analytical Approach

Analytical Approach

This study assessed how well the algorithm that adjusted raw urine drug levels for urine pH, urine specific gravity and lean body mass discrimination among the three clinically relevant doses of OxyContin® (80, 160 and 240 mg/d) when compared with raw values of the oxycodone LC-MS-MS substrate. All LC-MS-MS values reported in this study are the aggregated total of the noroxycodone, oxymorphone and oxycodone levels detected in the urine.

First, the concordance was tested1,2 between each pair of raw and adjusted LC-MS-MS values for each study participant between their third and fourth day values. The rationale for this was two-fold: (1) to confirm that participants achieved steady state at the given drug dose at days 3 or 4, and (2) to assess which of the two methods (raw or adjusted LC-MS-MS) achieved better concordance. In the case of the former, whether a patient achieves steady state is of clinical importance because during the accumulation phase, interpatient variability in elimination and accumulation can have exaggerated effects on observed serum and urine levels. In the case of the latter, it can be hypothesized that the method which achieves better concordance is also likely to show better discrimination between doses, as a direct result of lowered variability within each dose.

Second, an analysis of medians was conducted for each of the dosage groupings using Bonett-Price confidence intervals3-5 for both raw and adjusted LC-MS-MS values. In this study, an analysis of medians is a more appropriate choice than an analysis of means, given that (a) the distribution of values in this relatively small sample appears skewed at each dose level (see Figures 1 and 2), (b) no information as to the true population distribution of LC-MS-MS values is available, and (c) the analysis of medians is robust to almost any type of non-normality that would likely be encountered in practice. The Bonett-Price confidence interval method was chosen in particular because of its superior performance in simulation experiments of small samples.3 To establish even more conservative estimates, a Bonferroni adjustment was applied to the confidence intervals.

Fig 1. Distribution of raw LC-MS-MS values (in ng/mL) by dose.

Ameritox Figure 1

Fig 2. Distribution of adjusted LC-MS-MS values (in ng/mL) by dose.

Ameritox Figure 2