Permanent clinical chemistry program of the National Institutes of Health. II. Sources of variation in 2 imprecise analyzers

BACKGROUND: In a previous report of the program of interlaboratory concordancy in the MNIH (Mexican National Institutes of Health) we observed imprecision in nearly half of the analytes assayed by participants B and J. OBJECTIVES: 1. To identify possible sources of variation in the analyzers. 2. To take corrective actions and see their effect on precision. MATERIAL AND METHODS: Analyzer B had been operating for 2.5 years and analyzer J for more than ten years (Lab J switched to a new analyzer from a different supplier in the last three months of the study). Both participants submitted their daily results in commercial controls (normal and high). Lab J furnished results for a period of six months (Jul-Sep/92 using the old analyzer, and Oct-Dec using the new one) and Lab B reported five months (it used Jul to change its operating procedure). Lab B assayed 16 analytes (two electrolytes, four enzymes and ten organic analytes) and Lab J 17 (four electrolytes, four enzymes and nine organic). They also agreed to maintain a logbook of any change in calibrators, controls, reagents, analysts, or operating procedures. The information of the first two months was used in a multiple analysis of variance (MANOVA) using the control assays as the dependent variable, and the logbook notations as the independent variables. Precision was expressed as a monthly mean of means of CVs of all analytes. RESULTS: The MANOVA identified two sources of variation in analyzer B: 1) recalibrations (done fortnightly in an effort to improve performance); 2) control flasks. A restriction of calibrations led to an improvement in precision, i.e. from a CV of 5-7% in Aug-Sep to 3-4% in Oct-Dec. In the old analyzer of Lab J, the MANOVA was unable to identify causes of the high imprecision (CVs of 14-19% in Jul-Sep). With the new analyzer, precision improved (CV of 6% in Nov-Dec). Two observations of interest in the new analyzer were: a) a period of familiarization with the apparatus was apparently needed (CV was 11% in Oct); b) imprecision persisted in three analytes (CVs of 11-14% in BUN, calcium, and triglycerides in Nov-Dec). A second MANOVA in this lapse also failed to identify sources of variation for these three analytes. CONCLUSIONS: 1. Analyzer B is operating at an acceptable level of precision and can fully participate in our program. 2. The new analyzer of J can participate in the assays of 14 analytes.