Local Laboratory Reference Intervals in Clinical Trials

Abstract

Laboratory data are of major importance indrug safety assessments since they are consid-ered objective tools to supplement clinical judg-ments and the transfer of results across studypopulations. Moreover, the spectrum of avail-able biomarkers to assess drug safety and effica-cy is continuously being developed and refined.Although central laboratories are widely usedin clinical trials, local laboratories continue toplay an important role for trials where the use ofsuch central laboratories is not practical or fea-sible. Local laboratories are primarily perform-ing safety testing. The collection and analysis ofthese data remain a challenge in view of differ-ences in reference intervals, analytical methods,and reporting units. These challenges are wide-ly recognized. Other factors that may accountfor differences between local and central labora-tories are differences in population characteris-tics, such as ethnicity or nutrition. While suchdifferences should be reflected by differencesbetween local and central laboratory referenceintervals, analysis of central laboratory data inglobal clinical studies has shown these differ-ences to be insignificant (see next section). Al-though units are easily standardized by mathe-matical conversion, differences in methods andassociated reference intervals continue to raisequestions for statisticians. Chuang-Stein hasproposed a normalization procedure based onthe results of proficiency surveys by devising analgorithm to adjust results by comparing thembetween different laboratories and making theadjusted results comparable across laboratories(1). While such a procedure could indeedimprove result comparability, its feasibility islimited by, inter alia, the difficulties of gainingregular access to proficiency testing data, inter-pretation of the different schemes, especially inglobal studies, and the degree of participationin those schemes by the involved local laborato-ries. Thompson et al. have used the normal re-sult distribution of study populations as the basis for defining a commonly acceptable refer-ence interval (2). However, this proposal is limit-ed by the scope of the observations, which maynot necessarily cover existing differences inmethods in global studies, for example, the well-documented differences for commonly testedanalytes such as alkaline phosphatase or lactatedehydrogenase. Ruvuna et al. have proposed theapplication of generalized lab norms from a“phantom laboratory” by compiling a standard