Materials and Methods
IQMH POC glucose and laboratory glucose PT data obtained between September 2009 and June 2011 were used for the assessments. IQMH is a not-for-profit corporation wholly owned by Ontario Medical Association. IQMH proficiency testing programs are accredited against International Organization for Standardization/International Electrotechnical Commission (ISO/IEC) 17043:2010: Conformity Assessment—General Requirements for Proficiency Testing. Clients in Canada and other countries may voluntarily apply for accreditation and/or participate in PT surveys through the IQMH.
IQMH accreditation requirements are based on ISO 15189 and ISO 22870:2006, which contain general medical laboratory and POC testing requirements, respectively. All participants in the POC and the laboratory glucose surveys included in this study were accredited Ontario laboratories, and as a regulatory requirement, all POC testing devices in the facilities were managed by the institutions' respective laboratory regardless of the location of the devices within each institution.
The IQMH provides POC glucose and laboratory glucose surveys (ie, glucose tests performed on an automated analyzer in a hospital laboratory). The POC glucose survey consists of two shipments per year with three PT items in each shipment. The laboratory glucose test PT program consists of three shipments with three levels of PT material each time.
Commercially prepared bovine plasma products (Eurotrol WB Glucose, Ede, the Netherlands) and pooled human serum samples were used as PT items in the POC and laboratory glucose surveys, respectively. It is a challenge for PT providers to provide capillary whole blood or a commutable PT item that is compatible with all different brands of POC glucose devices. PT items for the POC glucose survey contain a physiologic buffered salt solution with a small amount of purified plasma of bovine origin. Because of the high probability of a matrix effect in this artificial PT item, only imprecision within instrument peer groups is evaluated, not method bias.
Peer group means and coefficients of variation (CVs) were estimated and compared. These statistical parameters were calculated using the robust algorithm recommended in ISO/IEC 13528(E). This nonparametric statistical algorithm consists of an iterative statistical function that minimizes the effects of the outliers on peer group mean, SD, and CV. Calculations start with a median (x*) as a statistical parameter for central tendency and s* = 1.483 × median |xi − x*| as a parameter for the dispersion around the median. The next step is that all values in the original data set that are lower than x* − 1.5 s* and higher than x* + 1.5 s* are trimmed to x* − 1.5 s* and x* + 1.5 s*, respectively. In the following step, new x* and s* are estimated as the mean and SD of a new data set using conventional parametric statistics. The trimming step is repeated several times until there is no change from one iteration to the next in the third significant figure of mean and SD.
Laboratories' performances were assessed using predefined allowable performance limits (APLs) (also known as performance specifications, analytic quality specifications, or analytical goals) and reflect total allowable error. APLs that are used in the survey assessment are determined by the IQMH scientific committees. Committees use one or more of the following methods in the determination of an APL: clinical outcomes in specific clinical settings, biological variation, recommendations from professional expert groups and the regulatory bodies, and current state of the art as demonstrated by PT program providers. Table 1 shows the allowable performance limits used at the time of the PT challenges included in this study.
The standard uncertainty of assigned value (UAV) was also estimated using the method recommended in the ISO 13528, and when peer groups' UAV were higher than 10% of the allowable performance range, peer groups were excluded from the study. The number of participants in each peer group was changing from 5 to 1,500 in the POC glucose and 5 to 47 in the laboratory glucose PT surveys. The Mann-Whitney U test was used for statistical comparison of the laboratory and POC glucose method CVs.
Participants are required to submit formal discordant investigation forms when their results exceed allowable limits at a given level, indicating large variations from the assigned mean. This investigation must include the cause of the discordant result and its corresponding corrective action. The program's advisory or expert committee reviews the submitted investigations and determines if the corrective action satisfactorily addresses the discordant result.
The number of institutions participating in POC glucose surveys was between 189 and 194, and the number of glucose meters within each institution ranged from 1 to 149 devices. A total of 51,379 POC glucose results were obtained from four PT surveys and 12 PT items. These results were assessed from 15 to 17 different POC devices. Laboratory glucose data included 2,661 submitted results within 179 peer groups. A total of 163 to 179 laboratories participated in six PT surveys that comprised 16 samples. Instruments used by participant laboratories are listed in Table 2.