Performance evaluation of PLT-H (hybrid-channel platelet) under various interferences and application studies for platelet transfusion decisions

Figures & data

Figure 1. Illustration of PLT-H.

Table 1. Regression analysis and deviation analysis in samples without interference.

Interference	Evaluated method	Range (×10^9/L)	Median (P25-P75)	Passing–Bablok regression			Spearman’s correlation (95% CI)	Bland–Altman analysis
				Equation	Intercept 95% CI	Slope 95% CI		Mean Bias	95% LOA
Noninterference	PLT-H vs. PLT-IRM	2–909 (n = 80)	40 (21–88)	y = −1.934 + 1.005x	-3.069 – -0.928	0.9767–1.039	0.993 (0.990–0.996)	−1.26	−14.53–12.01
	PLT-I vs. PLT-IRM			y = −0.4125 + 1.032x	-1.469–0.564	0.9946–1.052	0.992 (0.987–0.995)	2.25	−13.93–18.44
	PLT-O vs. PLT-IRM			y = −0.2349+0.9656x	-1.049–0.2576	0.9463–0.981	0.998 (0.997–0.999)	−2.04	−15.95–11.86
	PLT-H vs. PLT-IRM	2–94 (n = 63)	30 (18–48)	y = −1.679 + 0.9958×	-2.777 – -0.5066	0.9548–1.041	0.987 (0.978–0.992)	−2.00	−9.19–5.18
	PLT-I vs. PLT-IRM			y = −0.2401 + 1.012x	-1.541–1.194	0.9575–1.064	0.983 (0.972–0.990)	0.54	−8.53–9.62
	PLT-O vs. PLT-IRM			y = 0.1837 + 0.9426x	-0.3835–0.6188	0.9185–0.9656	0.997 (0.995–0.998)	−1.77	−6.30–2.77
	PLT-O 8 × vs. PLT-IRM			y = 0.5513 + 0.9414x	-0.8569–1.098	0.9138–0.9815	0.998 (0.996–0.999)	−1.45	−5.91–3.00

Range (minimum-maximum) and median values of PLT-IRM were analyzed; n represents the number of samples evaluated and CI represents the confidence interval.

Table 2. Regression analysis and deviation analysis in samples with various interferences.

Interference	Evaluated method	Range (×10^9/L)	Median (P25-P75)	Passing–Bablok regression			Spearman’s correlation (95% CI)	Bland–Altman analysis
				Equation	Intercept 95% CI	Slope 95% CI		Mean Bias	95% LOA
Small RBCs	PLT-H vs. PLT-IRM	27–617 (n = 51)	262 (221–350)	y = 4.755 + 0.9389x	-9.26 – −23.93	0.846–0.9917	0.963 (0.935–0.979)	−14.56	-63.14–34.00
	PLT-I vs. PLT-IRM			y = 13.5 + 0.9692x	-1.970–37.37	0.8815–1.030	0.940 (0.896–0.966)	2.25	-53.60–63.95
	PLT-O vs. PLT-IRM			y = 9.039 + 0.9165x	-2.481–2 8.13	0.8330–0.9628	0.977 (0.929–0.987)	−15.29	-59.65–29.07
Large PLTs	PLT-H vs. PLT-IRM	10–645 (n = 69)	120 (69–153)	y = 0.6117 + 0.9624x	-3.542–5.173	0.9137–1.040	0.983 (0.972–0.989)	−3.93	-36.35–28.50
	PLT-I vs. PLT-IRM			y = −6.144 + 0.9686x	-11.05 – −0.743	0.9212–1.012	0.964 (0.941–0.978)	−12.22	-44.57–20.13
	PLT-O vs. PLT-IRM			y = 1.178 + 0.9719x	-2.379–4.217	0.9264–1.004	0.985 (0.976–0.991)	−4.43	-27.80–18.93
RBC fragments	PLT-H vs. PLT-IRM	3–1355 (n = 57)	71 (34–221)	y = 1.097 + 0.9475x	-1.324–5.244	0.9107–0.9734	0.992 (0.987–0.996)	−7.86	-52.24–36.52
	PLT-I vs. PLT-IRM			y = 1.465 + 1.062x	-4.539–7.256	0.9772–1.190	0.969 (0.947–0.982)	9.14	-70.07–88.35
	PLT-O vs. PLT-IRM			y = 0.0691 + 0.9455x	-1.949–1.261	0.9251–0.9773	0.997 (0.994–0.998)	−7.95	-44.60–28.70
Multiple interferences	PLT-H vs. PLT-IRM	9–848 (n = 108)	241 (151–299)	y = −2.658 + 0.9959x	-8.048–2.494	0.9731–1.025	0.978 (0.967–0.985)	−2.02	-37.64–33.59
	PLT-I vs. PLT-IRM			y = −7.281 + 1.143x	-24.78–8.960	1.077–1.219	0.938 (0.909–0.958)	30.11	-74.84–135.07
	PLT-O vs. PLT-IRM			y = −2.679 + 0.9922x	-6.286–0.4248	0.9757–1.009	0.990 (0.985–0.993)	−4.77	-30.01–20.46
Samples with interference and PLT<100 × 10^9/L	PLT-H vs. PLT-IRM	3–97 (n = 84)	53 (32–73)	y = −0.5815 + 0.991x	-3.027–3.031	0.9245–1.054	0.975 (0.962–0.984)	0.358	-14.31–15.03
	PLT-I vs. PLT-IRM			y = 2.883 + 0.9135x	-1.169–9.130	0.8225–1.005	0.822 (0.735–0.883)	0.81	-30.31–31.93
	PLT-O vs. PLT-IRM			y = −0.6503 + 0.98×	-2.468–0.9738	0.9359–1.024	0.978 (0.966–0.986)	−0.9	-14.63–12.82
	PLT-O 8× vs. PLT-IRM			y = −0.2470 + 0.9705x	-2.516–1.243	0.9266–1.018	0.981 (0.971–0.988)	−0.66	-14.06–12.77

Range (minimum - maximum) and median values of PLT-IRM were analyzed; n represents the number of samples evaluated and CI represents the confidence interval.

Figure 2. Trend of deviation as MCV, MPV, and FRC% change.

Figure 3. Repeatability and ICC analysis of various platelet count methods.

Figure 4. Confusion matrices at different platelet infusion thresholds.

Table 3. Deviation analysis at different levels of medical decision-making.

Regression Model	Regression Equation	Parameters	Medical decision level	Predicted Y	Mean difference	Relative difference	95% CI	Criteria	Conclusion
Passing-Bablok	y = 0.584 + 0.970x	PLT-I vs PLT-IRM	5	5.43	0.43	8.66%	-12.83–40.31	±25%	deny
			10	10.28	0.28	2.83%	-5.96–17.23	±25%	accept
			20	19.98	−0.02	−0.09%	-3.48–6.11	±25%	accept
			50	49.08	−0.92	−1.84%	-4.77–2.21	±25%	accept
			80	78.17	−1.83	−2.28%	-5.65–1.95	±25%	accept
	y = −1.577 + 1.008x	PLT-O vs PLT-IRM	5	4.9995	−0.0005	−0.01%	-21.58–14.69	±25%	accept
			10	9.79	−0.21	−2.15%	-11.99–4.53	±25%	accept
			20	19.36	−0.64	−3.22%	-7.22 – -0.47	±25%	accept
			50	48.07	−1.93	−3.86%	-5.79–2,17	±25%	accept
			80	76.79	−3.21	−4.02%	-6.04 – -2.05	±25%	accept
	y = −1.362 + 1.000x	PLT-H vs PLT-IRM	5	3.64	−1.36	−27.20%	-46.17 – -6.70	±25%	deny
			10	8.64	−1.36	−13.58%	-21.62 – -4.59	±25%	accept
			20	18.65	−1.35	−6.77%	-9.86 – -2.67	±25%	accept
			50	48.66	−1.36	−2.68%	-4.91–0.33	±25%	accept
			80	78.67	−1.33	−1.66%	-4.21–1.47	±25%	accept