Automated Platelet Counts and Transfusion in Cancer Patients
Automated Platelet Counts and Transfusion in Cancer Patients
Published reports have shown the performance of the CD61 method to be equivalent in accuracy and more precise compared with the flow cytometric reference method. For the purpose of this study, we validated CD61 as a surrogate reference method using a select number of samples (n = 36). The results showed good agreement between CD61 and the reference manual phase platelet counts, as shown in Figure 1: r = 0.993 (y = 0.95× − 2.6). These results are consistent with the findings of a similar study performed by Gill and colleagues, who also compared CD61 with manual phase microscopy counts: r = 0.980 (y = 0.92× − 0.60).
(Enlarge Image)
Figure 1.
Phase microscopy vs CD61. Dotted line represents identity, solid line represents y = 0.95× − 2.6, and dashed lines represent 95% confidence interval bands. n = 36; r = 0.993. PLT, platelet.
Precision studies of the three hematology analyzers were undertaken for the degree of thrombocytopenia seen in this study. The mean, standard deviation, and CV for the optical and impedance measurements on the various analyzers are shown in Table 1. Method imprecision ranged from 2.6% to 8.1% at a platelet count of approximately 10 × 10/L. The immunoplatelet count showed the lowest CV percentage when comparing the different methodologies.
As described above, platelet measurements were compared across the three analyzers, as well as the different methods within the same analyzer. Table 2 provides the paired t test statistics for each automated method compared with CD61 for all samples and platelet counts (≤50 × 10/L, ≤30 × 10/L, ≤20 × 10/L, and ≤10 × 10/L). Difference plots of samples at 10 × 10/L or less, shown in Figure 2, clearly demonstrate that a positive bias exists in all four test methods. Passing and Bablok regression statistics are provided in Table 3. An increase in proportional error is evident in the 10 × 10/L or less group. An increase in the slopes and decreases in the correlation coefficients suggest a deterioration in analyzer performance at platelet counts of 10 × 10/L or less. The clinical significance of these determinations is especially evident in Table 4, which shows a progressive increase in the rate of undertransfusion as the platelet count thresholds decrease.
(Enlarge Image)
Figure 2.
Difference plots showing instrument/method vs CD61 at platelet count threshold <10 × 10/L. Dashed line represents identity, circles represent data, solid line represents bias, and dotted lines represent 95% limits of agreement. A, Bias = 1.51; 95% limits of agreement = −3.01 to 6.04; n = 92. B, Bias = 2.99; 95% limits of agreement = −2.90 to 8.37; n = 82. C, Bias = 3.30; 95% limits of agreement = −3.75 to 10.35; n = 32. D, Bias = 4.13; 95% limits of agreement = −1.95 to 10.21; n = 30. PLT, platelet.
Accurate enumeration of the platelet count has the potential to affect clinical intervention. Given the current "trigger" values for the administration of platelet transfusions currently in use as standard of care, disparity in the count reported by an analyzer and the "true" count, as determined by a reference method, may result in either undertransfusion (reference ≤10, reported value >10) or overtransfusion (reference >10, reported value ≤10) of platelets. The potential for underprescribing platelet transfusions, given the measurements obtained by various methods/analyzers using decision thresholds of 50, 30, 20, and 10 (×10/L) or less, is outlined in Table 5 and ranges from 9.1% to 22.4% in the 20 × 10/L group and 30.5% to 43.3% in the 10 × 10/L group. The risk for overtransfusion ranges from 0% to 6.1%. To better understand the clinical significance of biased platelet counts among analyzers in severe thrombocytopenia, we investigated the undertransfusion rates at transfusion triggers of 5 × 10/L or less to 15 × 10/L or less. As shown in Table 4, in all four platelet-counting methods, the potential for decision errors dramatically increased as the threshold was lowered. The Abbott optical method was associated with the lowest error rate, followed by Siemens optical and Sysmex impedance, which were fairly consistent at each threshold. The Sysmex optical method demonstrated the highest potential for undertransfusion.
The medical records of 31 patients with optical platelet counts of more than 10 × 10/L and CD61 counts of 10 × 10/L or less were reviewed to determine if a transfusion had been administered. In 20 (64.5%) of 31 patients, prophylactic transfusions were administered within 24 hours of the platelet count being reported, despite an optical measurement of more than 10 × 10/L. The remaining 11 (35.5%) of 31 patients were eventually transfused outside of 24 hours and, in most cases, only after a subsequent optical platelet count of 10 × 10/L or less had been reported. The study design did not include prospective monitoring of individual patients' clinical status, and therefore it is not known if any of the undertransfused patients experienced any significant hemorrhage.
Results
Comparison of Phase Microscopy and CD61 Immunoplatelet Count
Published reports have shown the performance of the CD61 method to be equivalent in accuracy and more precise compared with the flow cytometric reference method. For the purpose of this study, we validated CD61 as a surrogate reference method using a select number of samples (n = 36). The results showed good agreement between CD61 and the reference manual phase platelet counts, as shown in Figure 1: r = 0.993 (y = 0.95× − 2.6). These results are consistent with the findings of a similar study performed by Gill and colleagues, who also compared CD61 with manual phase microscopy counts: r = 0.980 (y = 0.92× − 0.60).
(Enlarge Image)
Figure 1.
Phase microscopy vs CD61. Dotted line represents identity, solid line represents y = 0.95× − 2.6, and dashed lines represent 95% confidence interval bands. n = 36; r = 0.993. PLT, platelet.
Precision Study of the Analyzers
Precision studies of the three hematology analyzers were undertaken for the degree of thrombocytopenia seen in this study. The mean, standard deviation, and CV for the optical and impedance measurements on the various analyzers are shown in Table 1. Method imprecision ranged from 2.6% to 8.1% at a platelet count of approximately 10 × 10/L. The immunoplatelet count showed the lowest CV percentage when comparing the different methodologies.
Comparison of Measuring Degrees of Thrombocytopenia Among Analyzers
As described above, platelet measurements were compared across the three analyzers, as well as the different methods within the same analyzer. Table 2 provides the paired t test statistics for each automated method compared with CD61 for all samples and platelet counts (≤50 × 10/L, ≤30 × 10/L, ≤20 × 10/L, and ≤10 × 10/L). Difference plots of samples at 10 × 10/L or less, shown in Figure 2, clearly demonstrate that a positive bias exists in all four test methods. Passing and Bablok regression statistics are provided in Table 3. An increase in proportional error is evident in the 10 × 10/L or less group. An increase in the slopes and decreases in the correlation coefficients suggest a deterioration in analyzer performance at platelet counts of 10 × 10/L or less. The clinical significance of these determinations is especially evident in Table 4, which shows a progressive increase in the rate of undertransfusion as the platelet count thresholds decrease.
(Enlarge Image)
Figure 2.
Difference plots showing instrument/method vs CD61 at platelet count threshold <10 × 10/L. Dashed line represents identity, circles represent data, solid line represents bias, and dotted lines represent 95% limits of agreement. A, Bias = 1.51; 95% limits of agreement = −3.01 to 6.04; n = 92. B, Bias = 2.99; 95% limits of agreement = −2.90 to 8.37; n = 82. C, Bias = 3.30; 95% limits of agreement = −3.75 to 10.35; n = 32. D, Bias = 4.13; 95% limits of agreement = −1.95 to 10.21; n = 30. PLT, platelet.
Effect of Platelet Measurement on Transfusion Decision and Timing of Transfusion
Accurate enumeration of the platelet count has the potential to affect clinical intervention. Given the current "trigger" values for the administration of platelet transfusions currently in use as standard of care, disparity in the count reported by an analyzer and the "true" count, as determined by a reference method, may result in either undertransfusion (reference ≤10, reported value >10) or overtransfusion (reference >10, reported value ≤10) of platelets. The potential for underprescribing platelet transfusions, given the measurements obtained by various methods/analyzers using decision thresholds of 50, 30, 20, and 10 (×10/L) or less, is outlined in Table 5 and ranges from 9.1% to 22.4% in the 20 × 10/L group and 30.5% to 43.3% in the 10 × 10/L group. The risk for overtransfusion ranges from 0% to 6.1%. To better understand the clinical significance of biased platelet counts among analyzers in severe thrombocytopenia, we investigated the undertransfusion rates at transfusion triggers of 5 × 10/L or less to 15 × 10/L or less. As shown in Table 4, in all four platelet-counting methods, the potential for decision errors dramatically increased as the threshold was lowered. The Abbott optical method was associated with the lowest error rate, followed by Siemens optical and Sysmex impedance, which were fairly consistent at each threshold. The Sysmex optical method demonstrated the highest potential for undertransfusion.
The medical records of 31 patients with optical platelet counts of more than 10 × 10/L and CD61 counts of 10 × 10/L or less were reviewed to determine if a transfusion had been administered. In 20 (64.5%) of 31 patients, prophylactic transfusions were administered within 24 hours of the platelet count being reported, despite an optical measurement of more than 10 × 10/L. The remaining 11 (35.5%) of 31 patients were eventually transfused outside of 24 hours and, in most cases, only after a subsequent optical platelet count of 10 × 10/L or less had been reported. The study design did not include prospective monitoring of individual patients' clinical status, and therefore it is not known if any of the undertransfused patients experienced any significant hemorrhage.
