Point-of-care haematological monitoring during treatment with clozapine

Introduction

Therapeutic treatment with clozapine is associated with a 0.4% risk of agranulocytosis [1]. Patients treated with clozapine are therefore required to provide regular venous blood samples to monitor changes in both white blood cell (WBC) and neutrophil counts. These tests are performed by a central laboratory, or an in-clinic analyser. The frequency of these tests depends on the stage of treatment, with patients starting on clozapine required to have blood samples taken weekly for the first 18 weeks. This then moves to fortnightly until month 12, then monthly thereafter [2]. This requirement for venous blood sampling can lead to patients declining clozapine drug therapy [3], and can also lead to patients discontinuing clozapine [4, 5]. The additional logistical demands of monitoring can also deter clinicians from prescribing clozapine, particularly as regular blood monitoring is not required for other antipsychotic medications.

An alternative to venous blood sampling is to collect a much smaller volume of capillary blood from a fingerprick. Capillary sampling is more acceptable to patients than providing a venous sample [6], and the procedure is easier for clinicians to perform. Devices are now available which can measure a full blood count using this approach [7], [8], [9], but their utility has yet to be evaluated in the context of heamatological monitoring for patients being treated with clozapine.

We addressed this issue by evaluating a device that permits point of care testing (POCT) for WBC and neutrophil counts using fingerprick samples: OLO^® (Sight Diagnostics LTD, Derech Begin 23, Lewinstein Tower, Tel Aviv, Israel). It’s accuracy was directly compared with that of a standard laboratory method, using capillary and venous samples collected from patients with schizophrenia who were being treated with clozapine.

Materials and methods

Results

Analytical method comparison

We analysed 215 samples by both methods; all patients that were asked agreed to double sampling. The WBC range of results as determined by ADVIA 2120i on venous blood was 3.12–16.52×10⁹/L, and the neutrophil count ranged from 1.38–11.72×10⁹/L. Very close correlations were determined by regression analysis between venous blood measurements by ADVIA 2120i and capillary blood measurements by the OLO device. Correlation coefficient for WBC count was 0.94, with a slope of 0.91 (95% CI=0.868 to 0.960) and an intercept of −0.26 (95% CI=−0.592 to 0.052) (Figure 1). Correlation coefficient for neutrophil count was 0.95 with a slope of 0.898 (95% CI=0.861 to 0.936) and an intercept of −0.082 (95% CI=−0.238 to 0.087) (Figure 2).

Figure 1:

OLO^® WBC (capillary whole blood) vs. Advia 2120i (venous whole blood) WBC counts (n=215).

Figure 2:

OLO^® Neutrophil (capillary whole blood) vs. Advia 2120i (venous whole blood) neutrophil counts (n=215).

Very close correlations were also determined by regression analysis between venous blood measurements by ADVIA 2120i and venous blood measurements by the OLO device. Correlation coefficient for WBC count was 0.98, with a slope of 0.95 (95% CI=0.923 to 0.978) and an intercept of −0.025 (95% CI=−0.212 to 0.159) (Figure 3). Correlation coefficient for neutrophil count was 0.98 with a slope of 0.93 (95% CI=0.906 to 0.957) and an intercept of −0.055 (95% CI=−0.167 to 0.051) (Figure 4).

Figure 3:

OLO^® WBC (venous whole blood) vs. Advia 2120i (venous whole blood) WBC counts (n=215).

Figure 4:

OLO^® Neutrophil (venous whole blood) vs. Advia 2120i (venous whole blood) neutrophil counts (n=215).

In total six patients (2.8%) recorded counts below standard thresholds on the OLO device by capillary sample. Of these, 4 had WBC counts <3.5×10⁹/L (but normal neutrophil counts), 1 had a neutrophil count <1.5×10⁹/L (but a normal WBC count), and 1 had both a WBC count <3.5×10⁹/L and a neutrophil count <1.5×10⁹/L. When assessed by the standard reference method, 2 out of 6 patients were below the cutoffs. One had WBC count <3.5×10⁹/L and neutrophil count >1.5×10⁹/L, and the other had both a WBC count <3.5×10⁹/L and neutrophil count <1.5×10⁹/L.

Bland-Altman plot generated for venous blood measurements by ADVIA 2120i and capillary blood measurements by the OLO device for WBC showed a mean of the differences (bias) of −0.91×10⁹/L (−11.5%) and a standard deviation of the differences of 0.85 with 95% CI = −2.58 to 0.76 (Figure 5). Bland-Altman plot generated for venous blood measurements by ADVIA 2120i and capillary blood measurements by the OLO device for neutrophil count showed a mean of the differences (bias) of −0.53×10⁹/L (−11.1%) and a standard deviation of the differences of 0.59 with 95% CI=−1.68 to 0.63 (Figure 6) Bland-Altman plot generated for venous blood measurements by ADVIA 2120i and venous blood measurements by the OLO device for WBC count showed a mean of the differences (bias) of −0.42×10⁹/L (−5.3%) and a standard deviation of the differences of 0.49 with 95% CI=−1.38 to 0.55 (Figure 7). Bland-Altman plot generated for venous blood measurements by ADVIA 2120i and venous blood measurements by the OLO device for neutrophil count showed a mean of the differences (bias) of −0.41×10⁹/L (−8.5%) and a standard deviation of the differences of 0.39 with 95% CI=−1.18 to 0.38 (Figure 8).

Figure 5:

Bland-Altman plot for the difference between OLO capillary whole blood WBC count and Advia 2120i venous whole blood WBC count against the mean of both methods for WBC count (n=215).

Figure 6:

Bland-Altman plot for the difference between OLO capillary whole blood neutrophil count and Advia 2120i venous whole blood neutrophil count against the mean of both methods for neutrophil count (n=215).

Figure 7:

Bland-Altman plot for the difference between OLO venous whole blood WBC count and Advia 2120i venous whole blood WBC count against the mean of both methods for WBC count (n=215).

Figure 8:

Bland-Altman plot for the difference between OLO venous whole blood neutrophil count and Advia 2120i venous whole blood neutrophil count against the mean of both methods for neutrophil count (n=215).

Discussion

The close correlations between the measures of both WBC and neutrophil counts from the OLO device and the standard ADVIA 2120i laboratory method indicate that the OLO device could be used clinically in patients being treated with clozapine.

In the U.K, the supplier of clozapine must provide a monitoring platform as well as a database that can identify patients’ WBC and/or neutrophil counts when they fall below defined clinical cutoffs and clinical intervention is required. These thresholds are presently <3.5×10⁹/L for WBC and <1.5×10⁹/L for neutrophils [12]. Assuming that these cut-off points apply, the OLO device would be acceptable as a frontline screen to measure these parameters at the point of care in clinical practice. Patients with results falling below these cutoffs by fingerprick capillary blood sample would be required to have a further venous blood sample taken for a confirmation measurement using the ‘gold standard’ laboratory method. This is because of the degree of negative bias that was found with WBC and neutrophil counts, whereby the OLO device generally undercounts compared with the standard method.

The Bland-Altman plots for capillary blood sampling (Figures 5 and 6) show small differences at these decision limits. These differences lie mainly at zero or slightly below, and then diverge as the measuring ranges increase giving rise to increased measuring bias. These findings do not affect patient outcomes concerning decision making at the clinical cutoff points. Reassuringly, they prevent missing results that are below these clinical cutoffs, but may falsely identify results that sit at, or slightly above the thresholds as being below these cutoffs. This bias should ensure that no cases of neutropenia or agranulocytosis are missed. The same findings can be seen for venous blood sampling (Figures 7 and 8), but with less divergence and scatter as demonstrated by the 95% confidence intervals.

In the present study, of 215 patients tested, only 6 (2.8%) tested would have needed to have a confirmatory FBC using the standard reference laboratory method. However, only one of these 6 patients would have needed clinical intervention, where the neutrophil count was <1.5×10⁹/L as confirmed by venous blood analysis by the standard reference method. Thus 2.8% of our sample would have needed a venous confirmatory FBC, and only 0.5% would have required clinical intervention. This would mean, on average, 3 out 100 patients tested by this point-of-care finger prick method would require a confirmatory venous sample to be taken, of whom at most one person would need to have adjustments to their clozapine treatment. These findings are similar to a previous study where we evaluated the Hemocue^® which found 4–5 patients in every 100 needing a venous blood confirmatory FBC, of which one would require changes to their clozapine treatment [13].

Evaluating the differences between slope, intercept, and correlation coefficient of the regressions of these two devices (OLO and Hemocue), for the parameter WBC, we find that the intercept (p=0.005) and correlation coefficient (0.015) are significantly different, whereas the slope is not (p=0.407). For neutrophils, we find that the slope (p=0.01) is significantly different, whereas the intercept (p=0.89) and correlation coefficient (p=0.15) are not. Overall, these findings would not alter the decision-making process in practice if using either device – the differences are not clinically significant. The main difference here is that OLO would likely require fewer confirmatory tests.

A similar phenomenon was demonstrated with the venous blood samples tested on the OLO device, where 5 (2.3%) out of 215 patients tested would need to have a confirmatory FBC using the standard reference laboratory method. Of these, 2 had WBC counts <3.5×10⁹/L (but normal neutrophil counts), 1 had both a WBC count <3.5×10⁹/L and a neutrophil count <1.5×10⁹/L, and 2 had a neutrophil count <1.5×10⁹/L (but normal WBC counts). When assessed by the standard reference method, 2 out of the 5 patients were below the cutoffs. One had WBC count <3.5×10⁹/L and neutrophil count >1.5×10⁹/L, and the other had both a WBC count <3.5×10⁹/L and neutrophil count <1.5×10⁹/L.

Given that we observed these phenomena with both capillary and venous bloods, this may point to the bias being due to the analytical method itself. However, the manual application of blood into the test cartridge may also be a possible source of this bias. Standardisation of practice, such as that through automation, could eliminate such a possibility.

These findings show that OLO device could be adopted into clinical practice for routine monitoring of patients on clozapine via either capillary finger prick or venous blood samples. This improves upon the previous study that evaluated only the capillary testing method [13], and shows the feasibility of providing patient choice on which blood collection technique they prefer. Practically however, the capillary sample collection method is somewhat complicated by the need to collect two blood drops into two separate pipettes which are then dispensed individually into the test cartridge. Also, the device is not portable which limits its use to clozapine clinic or ward settings. Community-based clozapine monitoring for agranulocytosis might still require venous blood collections to be analysed by standard laboratory method in the absence of a portable point of care alternative.

Limitations to this evaluation is that other parameters such as haemoglobin, platelets, haematocrit amongst others, were not considered. Although vital for physical health monitoring, are not essential for monitoring of agranulocytosis for patients on clozapine. Also, inter- and intraassay coefficients of variation were not established, however this has been investigated in an earlier study [14].

Conclusions

Adoption of a point of care device for clozapine monitoring of WBC and neutrophil counts would make the monitoring process less onerous for patients and clinicians, and could lead to an increase in the utilisation of clozapine in the treatment of schizophrenia. Clozapine is the most effective treatment for the disorder [15] and is the only treatment that is effective in patients who do not respond to other antipsychotic medications [16, 17]. The impact of using a POCT method for haematological monitoring could be further enhanced by combining this with the use of a POCT approach for monitoring clozapine plasma concentrations, which has recently been validated [18].