Performance of 6 routine coagulation assays on the new Roche Cobas t711 analyzer

Objectives For new analyzers or tests, analytical evaluation is required before implementation in the clinical laboratory. We evaluated the novel Roche Cobas t711 analyzer with six newly developed coagulation assays: the activated partial thromboplastin time (aPTT), prothrombin time (PT), international normalized ratio (INR), fibrinogen, d-dimer and anti-Xa. The evaluation included imprecision experiments, method comparison with the currently used Stago STA-R Evolution, monitoring of unfractionated heparin (UFH) with aPTT, a fast centrifugation protocol to improve turn-around time, and determination of sample stability in whole blood and plasma. Design and methods Imprecision and method comparison were assessed using commercial quality control samples and patient samples, respectively. For dose monitoring of UFH with the aPTT, samples from patients treated with UFH were used. Samples from healthy volunteers were collected for evaluation of the fast centrifugation protocol (5’ 2750×g) and for investigating sample stability over 6–8 h. Results Results for between-run precision were within the desirable specification. Method comparison showed an excellent agreement for fibrinogen, d-dimer and anti-Xa. For aPTT, PT and INR, a good correlation was found, but results were significantly lower on the t711 compared to the STA-R Evolution, which is caused by different coagulation activators. Results from the fast centrifugation protocol differed not significantly from the standard protocol (15’ 2500×g). Blood and plasma samples were stable at room temperature up to 6 and 8 h, respectively. Conclusions The t711 coagulation analyzer with 6 novel tests is suitable for routine use in clinical laboratories.


Introduction
Laboratory accreditation according to the international standard ISO15189:2012 demands that laboratories validate each novel assay and instrument before it can be used in clinical practice [1]. When methods developed by certified in vitro diagnostic companies are implemented without any modification, verification (instead of validation) of the method will suffice (ISO15189 paragraph 5.5.1.2). This implies that the laboratory should confirm that the performance characteristics as stated by the manufacturer are met. Although the Abbreviations: aPTT, Activated partial thromboplastin time; PT, Prothrombin time; INR, International normalized ratio; UFH, Unfractionated heparin; LMWH, Low molecular weight heparin.
ISO15189 standard does not specify which performance characteristics need to be evaluated, common verification procedures include at least analysis of imprecision and method comparison [2,3]. Additional performance characteristics can be included in the verification process based on national or local guidelines, or the laboratory specialist's professional opinion.
In this study, a verification was performed of six coagulations assays on the novel Roche Cobas coagulation analyzer t711, including the activated partial thromboplastin time (aPTT), prothrombin time (PT), international normalized ratio (INR), d-dimer, fibrinogen, and anti-Xa for monitoring of both unfractionated heparin (UFH) and low molecular weight heparin (LMWH) therapy. Results were compared with the corresponding methods on the Stago STA-R Evolution.
The aims of this study were first to reproduce previous results on imprecision and method comparison [4][5][6][7], second to evaluate the performance of the anti-Xa assay (not previously published), third to determine whether the aPTT can be used to monitor UFH therapy, fourth to investigate a fast centrifugation protocol in order to shorten turn-around time, and finally to study sample stability over time, since this is known to be method and reagent dependent [8].

Analyzers and choice of reagents
The Roche Cobas t711 (Roche Diagnostics GmbH, Mannheim, Germany) is a fully automated, continuous random-access coagulation analyzer. Liquid and lyophilized reagents are available and the latter are automatically reconstituted by the t711. The test menu comprises both general and specialized assays based upon optical clotting, chromogenic and immunoturbidimetric detection. Results were compared with the Stago STA-R Evolution coagulation analyzer (Diagnostica Stago S.A.S., Asni eres-sur-Seine, France). An overview of the investigated assays, reagents, detection methods and corresponding reference intervals on both systems is provided in Table 1.
For determining the INR on the t711, two different reagents are available (PT-rec and PT-Owren) with two modes of INR calculation each: uploading the lot-specific ISI and MNPT using e-barcode or analytical calibration with the PT Cal Set (calibrated against National Institute for Biological Standards and Control RBT/05). Before selection of a specific PT-reagent and INR calculation method for future use, we first determined the INR results from 1629 patient samples from the local thrombosis services using both methods and using both INR calculations, and compared these with the INR results of the Hepato-Prest reagent on the STA-R Evolution. The combination of the PT-Owren reagent with INR calculation using the PT Cal set corresponded best with the Hepato-Prest reagent (data not shown) and was therefore selected for all subsequent experiments.

Reproducibility
The reproducibility (i.e. intermediate precision or within-lab variation) was determined by measuring 2 or 3 levels of commercial quality control samples (Roche) in duplicate on 20 separate days, except for anti-Xa which was performed in duplicate on 10 days. Between-run precision results were compared with the performance data specified by the manufacturer and with the desirable specification of imprecision based on within-and between-subject biological variation according to Westgard [10,11].

Method comparison
For this experiment, only excess material of pre-existing and anonymized patient samples collected for routine clinical care was used. Results were used for analytical evaluation only and not reported in the patient file, so patient-informed consent was not required. Samples were collected via venipuncture into evacuated blood tubes containing 0.109 mol/L (3.2%) sodium citrate (BD Vacutainer, Becton Dickinson and Company, Franklin Lakes, NJ) and were measured directly on both the t711 and STA-R Evolution. For anti-Xa measurements, samples from dialysis patients anticoagulated with LMWH or UFH were centrifuged twice to obtain platelet poor plasma (platelet count <10 Â 10 9 /L) and immediately frozen until analysis. Because anti-Xa was previously not performed in our laboratory, results from the anti-Xa assay on the t711 were compared with the anti-Xa results determined by a reference laboratory (Radboud University Medical Centre, Nijmegen, the Netherlands) using the STA-R Evolution. The number of samples and the investigated range for each assay is included in Table 3. Data were analyzed using the Passing-Bablok regression method [3,12] in Analyze-it (Analyze-it Software Ltd, Leeds, UK).

Sensitivity to unfractionated heparin
After verification of the anti-Xa assay on the t711, the dose-response of the aPTT assay to UFH was investigated in 41 samples from patients anticoagulated with UFH, in order to determine whether aPTT results can be used to monitor UFH therapy.
The dose-response curve of the aPTT versus anti-Xa activity was empirically fitted using an exponential function in Microsoft Excel 2016, which resulted in a better correlation coefficient than a linear fit.

Sample centrifugation
The standard centrifugation protocol of 15 min at 2500Âg as recommended by the manufacturer, was compared with centrifugation for 5 min at 2750Âg. For this experiment, two citrated blood samples were obtained from 10 healthy volunteers and from 20 patients treated with vitamin K antagonists. In all samples, platelet count, aPTT, PT, INR, fibrinogen and d-dimer were measured. Anti-Xa was not investigated in this experiment, as a double spin to obtain platelet poor plasma is always required for this assay in order to avoid heparin neutralization by platelet factor 4. Data from both centrifugation protocols were compared using a paired t-test in Analyze-it.

Stability in plasma and whole blood
To simulate a delay in sample processing, for instance due to transportation, whole citrated blood samples were drawn from five healthy volunteers and incubated for 0, 2, 4 and 6 h at room temperature. After each incubation step, samples were centrifuged and the aPTT, PT, INR, fibrinogen and d-dimer were determined. Stability of anti-Xa was not investigated since in our hospital anti-Xa activity is always determined within 1 h after blood collection. Similarly, in order to determine the stability in plasma, the citrated plasma samples from the same volunteers were incubated for 0, 2, 4, 6 and 8 h at room temperature. Data were analyzed in Analyze-it using a Dunnett comparison of the average at each time point versus the average at t ¼ 0. Table 2 shows the reproducibility of each assay for the different levels of quality control samples. Results are compared to the specification of performance provided by the manufacturer and to the desired imprecision based on biological variation derived from the Westgard database [10,11]. For aPTT, PT, INR, d-dimer and fibrinogen, the reproducibility was comparable to the manufacturer's specifications. Additionally, results for all assays easily met the desired imprecision based on biological variation for all these assays, which is an important criterion for clinical application. For anti-Xa, no information on biological variation was available and the specifications claimed by the manufacturer were only met for LWMH. No explanation was found for the higher CV% of the anti-Xa for UFH measured in the current study.

Reproducibility
Our results for reproducibility were similar to previous studies. In a study by Kitchen et al. analytical performance was determined on the Cobas t711 for PT-rec, aPTT and PT-Owren [4]. Although they determined within-run and not between-run reproducibility, their results for the CV of PT-rec (0.2-0.7%), PT-Owren (0.3-0.8%) and aPTT (0.3-1.2%) were comparable to our results and well within the desired imprecision (based on biological variation). Additionally, our results were comparable to the study performed by Lippi et al. [7], who described between-run CV's of 0.6% and 0.5% for PT-rec at 7.97 s and 27.8 s, and 1.7% and 1.5% for aPTT at 25.8 s and 40.8 s, respectively.
For the D-dimer and fibrinogen assays, our results also compared well to studies performed by Salvagno et al. and Lippi et al. [6,7].
Salvagno et al. determined between-run CV's for D-dimer of 3.0%, 0.7% and 0.6% at 350 μg/L, 1348 μg/L and, 8026 μg/L, respectively, whereas our results (for 2 levels) were 1.0% and 0.6% at 883 μg/L and 3677 μg/L. For fibrinogen, our results of 1.6% and 3.2% at 1.19 g/L and 2.82 g/L compared well to the results from Lippi et al. [7] of 1.7% and 2.0% at 1.21 g/L and 2.56 g/L, respectively. Table 3 and Supplemental Figs. 1 and 2 show the results of the method comparison studies between the t711 and the STA-R Evolution. Correlation coefficients varied from 0.92 to 0.99 and were comparable to (or higher than) determined in previous studies for PT (r ¼ 0.97), aPTT (r ¼ 0.81) and fibrinogen (r ¼ 0.97) 7 . Patient results for both the aPTT and PT were significantly lower on the t711 compared with the STA-R Evolution. This was also found in a previous study and can be explained by the differences in coagulation activators and detection methods (Table 1) [7]. Bland-Altman plots showed a consistent difference between the t711 and the STA-R Evolution for the PT and aPTT assays (Supplemental Fig. 2a/b). Therefore, before routine utilization of the t711, clinicians should be well informed about the lower aPTT and PT results and corresponding reference ranges [4].

Method comparison
INR-values derived from the PT-Owren on the t711 were on average 6% lower than Hepatoprest values over the entire range, despite international normalization using the INR calibration set to determine the ISI (of 0.904) and MNPT (20.2 s) for this lot. This was also reflected by a negative bias in the Bland-Altman plot (Supplemental Fig. 2c). This difference was also observed in a previous study, where the Owren PT assay on the Cobas t711 was also compared to the HepatoPrest assay on the STA-R MAX [4]. A slope of 0.853 and an intercept of 0.132 with a correlation of 0.994 was found, reflecting a good correlation but significantly lower INR values of the Owren PT on the t711.
To determine the relevance of these lower INR values, the performance of the INR was further investigated by measuring 17 proficiency testing samples with INR-values ranging from 0.99 to 3.40. INR results from the t711 were consistently lower than the mean Table 3 Results of the method comparison experiments. The 95% confidence intervals for slope and intercept were determined using Passing-Bablok regression and are shown in parentheses. Statistically significant differences between the Roche Cobas t711 and Stago STA-R Evolution are highlighted in bold (i.e. a slope different from 1 and/or an intercept different from 0).  consensus value for 14 samples, but all results were within the acceptable analytical range (i.e. mean AE 2 Â standard deviation; data not shown). The INR was therefore approved for clinical use and thrombosis specialists were informed on the assay differences to allow adequate dosing of vitamin K antagonists. For d-dimer, fibrinogen and anti-Xa, no significant differences or biases were found. Unfortunately, low fibrinogen levels could not be included because patient samples with decreased fibrinogen were unavailable during the evaluation period. As an alternative, we used six samples from a proficiency testing program with low fibrinogen levels (<2 g/L). This showed excellent agreement (data not shown) and the t711 fibrinogen was therefore approved for the low range samples as well. Fig. 1 shows the dose-response curve of the t711 aPTT versus anti-Xa activity of unfractionated heparin measured in 41 patient samples. A good correlation with an R [2]-value of 0.87 was obtained. The aPTT therapeutic window was subsequently determined by calculating the aPTT corresponding with anti-Xa activities of 0.3 and 0.7 U/mL, which is the common therapeutic range [13]. This resulted in a calculated aPTT range of 42-82 s, which we considered a sufficient window for UFH monitoring.

Sensitivity to unfractionated heparin
Interestingly, Kitchen et al. found a considerably more narrow therapeutic window of 46-57 s, which was determined using plateletpoor plasma from 117 patients [4]. A possible explanation is that Kitchen et al. applied a linear fit of their data, resulting in a moderate correlation coefficient of 0.41 (corresponding to an R 2 of 0.17). In contrast, the R 2 of 0.87 obtained using an exponential fit, indicates that our model describes the relationship between anti-Xa and aPTT more accurately, even though fewer samples were included. Table 4 shows the results for the standard and faster centrifugation protocol. Although the number of platelets was significantly higher using the faster centrifugation, no significant differences were found for PT, INR, fibrinogen and d-dimer between both centrifugation protocols. Although aPTT results were on average 0.74 s higher using the standard centrifugation protocol, this difference is clinically irrelevant. The faster sample centrifugation of 5 min at 2750Âg is therefore considered suitable for all routine coagulation assays on the t711. This is important for the optimization of sample throughput and allows for a clinically relevant reduction of the turnaround time.

Stability in plasma and whole blood
Supplemental Figure 3 shows the stability at room temperature for the aPTT, PT, INR and fibrinogen on the t711 in whole blood and plasma obtained from five healthy volunteers. Only one volunteer had detectable d-dimer levels, thereby hampering further analysis. These stability studies showed that additional coagulation tests on the t711 can be safely performed in plasma until 8 h after centrifugation. When sample transportation is required (e.g. from an outpatient clinic to a central laboratory), a delay of maximally 6 h between venipuncture and centrifugation is acceptable. A drawback of the current study is that the experiments were performed on samples obtained from healthy volunteers and that no patient samples were included. Additional stability studies using patient samples with abnormal results are required for complete evaluation.

Conclusions
The Roche Cobas t711 coagulation analyzer is a user-friendly instrument suitable for routine application in 24/7 clinical laboratories. All investigated assays showed adequate analytical performance in the reproducibility and method comparison experiments. No implications for patient diagnosis and monitoring were experienced upon implementation of the t711 in routine clinical care. Additionally, we found that whole blood and plasma samples are stable for up to 6 and 8 h at room temperature, respectively. This is important for transportation (e.g. from outpatients or from general practice) or when additional coagulation tests are requested in stored plasma. The fast centrifugation protocol of 5' at 2750Âg was found suitable for all routine tests on the t711, thereby allowing improved sample throughput and reduced turn-around times.
Additional experiments to verify potential interferences from hemolysis, icterus or lipemia should be conducted for the t711, since all assays are based on optical detection methods. Furthermore, future studies should include determining the sensitivities of the aPTT and PT to coagulation factor deficiencies and to direct oral factor IIa or factor Xa inhibitors (e.g. dabigatran, rivaroxaban and apixaban).

Author contributions
MO and JBdK designed the research, analyzed the data and wrote the paper. RLN and JDN performed the research and analyzed the data. All authors drafted or revised the manuscript critically for intellectual content and approved the final version. Table 4 Average platelet counts and results of the coagulation tests obtained in paired samples centrifuged for 15 min at 2500Âg or 5 min at 2750Âg. Results are presented as mean AE standard deviation. P-values <0.05 as determined using a paired t-test are considered statistically significant. For d-dimer, results are shown for 10 samples only, as the other 20 samples had d-dimer levels below the limit of detection.