Why the study?
Does the readjusted standardization of the high-sensitivity cardiac troponin T assay alter the diagnostic accuracy for acute myocardial infarction compared to affected lots?
Does the readjusted standardization of the high-sensitivity cardiac troponin T assay alter the diagnostic accuracy for acute myocardial infarction compared to affected lots?
Answers are grounded in this paper and cite their sources.
Readjustment of the high-sensitivity troponin T assay standardization corrected a low-end downward shift without affecting the clinical diagnosis of acute myocardial infarction at the 14 ng/L threshold.
To the Editor: Apple and Jaffe recently discussed potential implications of low-end adjustment of the Roche Diagnostics Elecsys® Troponin T high-sensitivity assay (TnT hs)1 (1). Because we represent the product's manufacturer, we reply to their questions. This fifth-generation assay was launched outside the US in 2009 and is not yet commercially available in the US. In recent studies of the reference interval, we observed a downward shift in the TnT hs assay at low concentrations of the measurement interval (3–10 000 ng/L; Fig. 1A). This value was confirmed to be 2.7 ng/L in internal lot-to-lot comparisons (Fig. 1B) and reported as 3–5 ng/L in recent customer reports. This shift caused up to 88% of the samples from a healthy cohort to be measured with values below the limit of the blank (3 ng/L), rather than ≤50% of samples, as was observed in the initial multicenter evaluation (MCE) study (2). Apple and Jaffe detected the shift by using maximum values of individual samples rather than mean values. A review of Fig. 1A may help clarify this aspect. Lot 163704 (old standardization) was compared with lot 167345 (readjusted standardization). The difference plot shows the greatest variation at the lower limit of the measurement interval and a deviation of about 3 ng/L at the cutoff value of 14 ng/L. For affected lot 163704, 623 samples (84%) had TnT results below the limit of the blank, and 1 sample had a value of 29.8 ng/L (Δ, 2.1 ng/L). For readjusted lot 167345, 288 samples (39%) had TnT hs results below the limit of the blank. (B), With 127 human samples with TnT hs values of 3–500 ng/L, internal lot-to-lot comparisons confirmed the shift at the low end of the measurement interval to be 2.7 ng/L (y = 0.99x + 2.70 ng/L, Passing–Bablok regression). We retrospectively identified the following 6 affected lots: 157120, 160197, and 163704 for the TnT hs 18 min assay and lots 158268, 161334, 164773 for the TnT hs STAT assay produced between 2009 and 2011. The criteria for lot release were fulfilled for all lots produced since the introduction of the assay. A subsequent analysis of the raw data (signal counts) confirmed that this shift was not caused by a reagent abnormality but by the acceptance criteria of standardization at the very low end of the calibration curve, which were too broad. To prevent such shifts in the future, we adjusted the standardization procedure by instituting more-stringent release criteria (see point 4 below). This optimization was first introduced in 2012 with lots 167345 (TnT hs 18 min assay) and 167650 (TnT hs STAT assay) to ensure that the low-end recovery was similar to that of the original MCE, in which 57% of the samples from the healthy population had measured values ≥3 ng/L (limit of the blank) and 32% had values ≥5 ng/L, the limit of detection. Specifically, we obtained a 99th percentile value of 14.4 ng/L in a single-center study that used lot 167345 (n = 739; median age, 49 years). This standardization also ensures that the 99th percentile decision threshold remains at 14 ng/L, as originally specified in the MCE (2). The identification of approximately 50% of the healthy individuals with TnT values ≥3 ng/L has been confirmed in other studies (3, 4). Apple and Jaffe might have overlooked the facts that some of the cited studies had used the precommercial assay that produced preliminary performance data and that stable cardiovascular disease was the basis for selecting the study population in the PEACE (Prevention of Events with Angiotensin Converting Enzyme Inhibition) study (5). Our answers to the specific questions are as follows: The well-established 99th percentile value of approximately 14 ng/L is preserved for all lots released since 2012. For the 6 lots with the downward shift, the 99th percentile was lower by 3–5 ng/L (for additional information, please refer to our recent product management letter to customers no. 57/2012). The readjustment of the standardization does not lead to a percentage increase in the detectable results (i.e., a limit of detection greater than that obtained in the MCE), but only compared with the affected lots. All lots introduced since 2012 present the original recovery concentration and the MCE study's 99th percentile value of approximately 14 ng/L. Compared with the affected lots, the restandardization caused a percentage increase in the detectable results that was greater than the limit of detection (5 ng/L) for 27.6% (n = 1033) of the patients presenting to the emergency department with symptoms of suspected acute coronary syndrome. The use of troponin to aid in the diagnosis of acute myocardial infarction (AMI) requires troponin results to be assessed in conjunction with the patient's medical history and a clinical examination. Especially at low troponin concentrations, guidelines require that the diagnosis of AMI not be based on a single troponin measurement but rather on the detection of an increase/decrease in the troponin concentration (i.e., a change, Δ) so that AMI can be distinguished from chronic troponin increases and biological variations. We thus tested and compared the diagnostic accuracy of an affected lot with that of a lot with the readjusted standardization for 1033 patients presenting to the emergency department with suspected acute coronary syndrome. Single troponin measurements with the 2 lots 3 h after emergency department admission identified identical numbers of patients (i.e., n = 94) with suspected acute coronary syndrome, a diagnosis of AMI, and a troponin T concentration >14 ng/L. In contrast, population and risk-stratification studies of apparently healthy cohorts might be affected by the downward shift and must be assessed individually. Having identified the cause of the downward shift, we have implemented more-stringent criteria in the QC process, including the use of much narrower intervals and limits for the standardization process. Moreover, additional control samples lower than and close to 14 ng/L have been added to prevent such downward shifts in the future. The shift was not due to a change in reagent formulation but to the standardization of lots against the master lot. It is possible to apply the adjusted standardization procedure to former measurements and to recalculate data that have been measured with one of the 6 affected lots, provided that both the raw data (signal counts) for the samples and a complete set of calibration data are available. Recalculation based on concentrations is also possible, but this approach cannot recover values originally displayed as being below the assay limit (3 ng/L). Roche Diagnostics Elecsys® Troponin T high-sensitivity assay multicenter valuation Prevention of Events with Angiotensin Converting Enzyme Inhibition (study) acute myocardial infarction.
No takes yet. Share an insight, caveat, or question.
Hallermayer et al. (2013) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: