What information should manufacturers provide on their procedures?

To the Editor: Recently, our laboratory received a bulletin (Technical Bulletin CC06001, Calcium C.f.a.s. Set Point and PNU/PPU Target Value Reassignments: All Clinical Chemistry Systems, January, 2006) from Roche Diagnostics Australia indicating that the calcium value of their calibrator for automated systems (C.f.a.s.) had been reassigned. According to the bulletin, the calibration was performed against atomic absorption spectroscopy (AAS). The bulletin also stated that Roche’s quality assurance system did not reveal any substantial lot-to-lot changes over the 5 years since the last reference standardization. No data were provided on the quality assurance system results or on the standardization. The change for the assigned value for the C.f.a.s. lot being used in our laboratory (lot 166275) was from 2.04 mmol/L to 1.93 mmol/L. We were concerned that the magnitude of this change (5.4%) was greater than the desirable total error for calcium, 2.4% (1 ), and we felt that such a change could have a substantial effect on patient data. Thus, we asked Roche Diagnostics to provide further information regarding the reassignment. We were informed that the Roche Diagnostics in-house standardization procedure included the assays of 10 patient pools on either the Roche AAS or the Hitachi Modular instruments (personal communication, Dr. George Koumantakis, Roche Diagnostics Australia). The results were provided to us as medians, and we assumed that these represented the results from several different determinations (data not provided). Fig. 1 shows the results for these 10 patient pools measured on the Hitachi Modular instrument plotted against the AAS results. Over this range, the Hitachi results provided by Roche Diagnostics Australia for the 10 patient pools were 5% higher than the AAS results, also provided by Roche Diagnostics. We were informed that Roche Diagnostics used this mean to reassign the values for the C.f.a.s. master lot, from which the calcium values for their various commercial lots of C.f.a.s. are derived. The effect of reducing the Hitachi result for the patient pools by 5% is shown as the dashed line in Fig. 1. This method of reassigning was inappropriate for 2 reasons. First, the restandardization was carried out with patient pools with calcium values between 2.16 and 2.46 mmol/L to correct the assigned calcium concentration in C.f.a.s, which is 2.0 mmol/L, well outside the range over which the restandardization was done. Second, the results also showed a proportional bias between the Hitachi method and AAS. We were not sure what effect these 2 issues would have on the reassignment of values outside the range covered by the standardization. To investigate this further, Roche Diagnostics (Australia) kindly provided us with Standard Reference Material 956b (SRM 956b), obtained from the NIST, which are human serum-based materials containing 3 concentrations of calcium, with values determined by inductively coupled plasma mass spectroscopy (ICP-MS). We prepared these materials according to the protocol provided by NIST and assayed them immediately on the Hitachi Modular, using the o-cresolphthalein complexone method with the original assigned value used for the calcium value of the C.f.a.s. Aliquots were also prepared, stored at 30 °C, and subsequently assayed on the Hitachi Modular over 5 days (twice per day) using thawed aliquots. The results are shown in Fig. 1. The error bars represent the errors provided by NIST and those obtained from our assays. The regression line has been fitted by use of linear least-square regression, but other regression methods gave the same results. As with the standardization carried out by Roche Diagnostics, these results show a proportional bias between the results obtained with