When visual inspection of the palpebral conjunctivae falls short …

For anemia screening, physicians can perform a visual inspection of the palpebral conjunctivae, order a complete blood count, spin a hematocrit (Hct), or use a point-of-care hemoglobin (Hb) assay device. Given that visual inspection of the conjunctivae by a physician is at best 75% accurate, measurement of Hb or Hct is clearly preferred.1 Blood collectors measure Hb to determine if prospective donors meet regulatory requirements designed to prevent anemia and to ensure the potency of transfused red blood cell components.2-4 The “gold standard” method, a venous blood sample measured on a laboratory hematology analyzer, is poorly suited for this purpose at many satellite and mobile blood sites. The ideal point-of-care method would be portable, producing rapid results, not requiring technically skilled operators, using only minimal controls or calibration, and would be highly accurate. Hb screening of blood donors historically employed a semiquantitative gravimetric measurement using standardized copper sulfate (CuSO4) solutions. With the introduction of portable devices to measure blood-drop Hb or Hct, many blood collectors have replaced CuSO4 with quantitative Hb and Hct technologies. Regardless of the method used, numerous preanalytic variables affect the accuracy of these results, including body position, capillary sample integrity, and device sample loading.5-10 Recently introduced noninvasive Hb measuring devices may offer improved donor acceptability without the pain (and rare fingerstick-site complications) associated with capillary blood sampling or its technical variability and biohazardous waste. A substantial fraction of blood donors cite the fingerstick as a deterrent to blood donation.11 The arrival of this long-anticipated technology has triggered a new cycle of evaluations of the comparability and accuracy of such devices in the blood donor setting.12-21 Noninvasive Hb measuring devices, however, do not appear to reduce testing time and introduce technical challenges of their own with ambient light and skin or nail coloration interference. The US Food and Drug Administration (FDA) has granted 510(k) clearance for noninvasive spot-check Hb (spHb) testing to the Pronto and Pronto-7 (Masimo) and NBM-200MP and NBM-200 (OrSense) devices. They are indicated for use by trained personnel in noncritical settings like nonintensive care areas of hospital facilities, mobile environments, clinics, physician offices, and ambulatory surgery centers, although only the Pronto-7's approval specifically mentions blood donation facilities. The Haemospect (MBR Optical Systems), a transcutaneous reflection spectroscopy-based device, has not received FDA clearance. An 18-month Haemospect trial in more than 90,000 Irish donors was terminated in November 2015 after small but significant numbers of donors were erroneously accepted with Hb values as low as 8.4 g/dL, with some requiring postdonation transfusion.22, 23 The Pronto-7 is a noninvasive pulse CO-oximeter that spectroscopically measures oxygenated, deoxygenated, carboxy- and met-Hb, each of which differs in visible and infrared light absorption. Multiple wavelengths (500-1300 nm) are transmitted through a reusable finger probe placed over the nail bed. A photodetector captures the attenuated signal, which is analyzed using adaptive filters and multiwavelength calibration equations to derive the spHb. The device has been validated for Hb values between 6 and 18 g/dL.24 The NBM-200 is a portable Hb and oximetry monitor employing occlusion spectroscopy technology. The donor's finger is placed through a reusable ring-shaped sensor with an inflatable cuff and a multiwavelength (600-940 nm) optical measuring system. Brief intermittent arterial occlusion results in changes in blood optical behavior within the finger, which are analyzed and interpreted. Hb values within the range of 7 to 17 g/dL are derived from the signals generated during measurement.25 Regulatory clearance does not preclude cGMP, GLP, and CLIA requirements to evaluate agreement with current methods or simultaneously compare both techniques to the venous Hb. There are, however, no standards for conducting appropriate method comparison studies. Correlation and regression analyses are most frequently reported. However, correlation analysis assesses the linear relationship between two measurements, not their agreement, and more important, not their differences. These analyses are therefore not recommended for judging comparability between methods.26 The Bland-Altman plot is a preferred analytic method to determine agreement by measuring differences of a test method compared with a reference method (reported as the mean bias of test minus reference values at various Hb levels, a measure of accuracy) and the standard deviation of the mean difference (SDD, a measure of precision), which allows estimation of the 95% limits of agreement (LOAs).27, 28 The upper LOA defines a limit of overestimation of Hb outside of which less than 2.5% of test measurements fall within a given sample of donors. Rice and colleagues29 provide guidelines for interpreting Bland-Altman plots. While a large bias implies systematic inaccuracy, a bias of zero may also mask an unacceptably large positive and negative inaccuracy around the mean. Similarly, high precision (a low SDD) may or may not be associated with high accuracy. As opposed to bias and precision, LOAs are a key metric for determination of the utility of a new device, specifically the upper LOA (or its 95% confidence interval [CI], which is less often calculated but represents the highest frequently observed value for the entire population rather than a study sample). The upper LOA should always be viewed from a clinical perspective. A low value that is clinically insignificant suggests that the test may be useful in donor screening. Wide LOAs for noninvasive Hb devices complicate decision-making in individual donors. In a patient setting, Rice and colleagues recommend that Bland-Altman data be separately analyzed in the critical Hb range within which transfusion decisions are made (6.0-10.0 g/dL). This step avoids contamination of mean and SDD calculations with values well within the normal range where small differences in Hb values are insignificant. We similarly recommend subanalysis in the less than 13.5 g/dL range wherein acceptability for donation is critical when considering new technology implementation. It should also be pointed out that apheresis eligibility (and donor health) may be affected at very high Hb values, at which it may also be important to accurately assess new technology accuracy. In this issue of TRANSFUSION, Baart and coworkers12 present data with several specific aims: 1) to assess whether two noninvasive spHb devices provide similar precision and accuracy as invasive capillary Hb measurements and venous sample testing, 2) to identify interfering factors, and 3) to assess practicability of the various devices from donor and staff perspectives. This large study compares the performance of the noninvasive NBM-200 and Haemospect meters and the invasive Hemocue 201 and 301 devices to the reference Sysmex XT-188i venous Hb in almost 2600 blood donors. Values obtained from nonreference devices were analyzed using the following statistical methods: calculation of mean Hb ± SD; mean differences and Pearson correlation coefficients versus the reference technique; Bland-Altman plots of test methods versus venous Hb and noninvasive tests versus the Hemocue 201 test-of-record; and the sensitivity, specificity, positive predictive values (PPVs), and negative predictive values (NPVs) for detection of Hb levels below Sysmex cutoffs for donation. Measurements with all test devices differed significantly from the measurements in venous samples. Compared to measurements in venous samples, Hb levels measured with the invasive devices, HemoCue 201 and 301, were higher in both men and women (Table 1). For the noninvasive devices, Hb levels measured with the NBM-200 were higher compared to measurements in venous samples, whereas measurements with the Haemospect were lower in men and women. The smallest fixed bias (mean of test minus reference values) was seen in noninvasive devices; however, the SDD was smallest for the invasive devices. Using venous values as a reference, measurements like the PPV, NPV, sensitivity, and specificity with invasive devices in both men and women were better than those with noninvasive devices. Measurements with the HemoCue 201 yielded the lowest inappropriate acceptance rates (1 – sensitivity). Measurements performed with the Haemospect were almost uniformly the least accurate. Blood donation is known to lower donors’ Hb by approximately 0.5 g/dL immediately after and by 1 to 2 g/dL after 2 to 3 days.30 In perspective, compared to the venous sample, the NBM-200's 95% upper LOA was 2.0 g/dL for men and 2.3 g/dL for women. Thus, for donors with Hb values of 11.0 g/dL on Sysmex, the NBM-200 could read 13.0 g/dL for a male or 13.3 g/dL for a female donor. The 1 to 2 g/dL Hb decrement would result in a 9 to 10 g/dL postdonation Hb, where early symptoms of anemia are generally considered to develop in people with a relatively healthy cardiovascular status.31 In contrast, a donor at or just below a Hb of 11.0 g/dL is still likely to be deferred with the HemoCue 201. In a meta-analysis of 32 studies of noninvasive hemoglobinometers in 4425 subjects in different clinical settings (including two blood donor studies), Kim and colleagues32 reported an overall pooled random-effects mean difference (noninvasive minus reference hematology analyzer) and SDD of 0.10 ± 1.37 g/dL (LOAs, −2.59 to 2.80 g/dL). While the overall mean difference between noninvasive Hb and central laboratory measurements was small, the wide LOAs suggest that clinicians should be cautious when making clinical decisions based on these devices. A nonexhaustive literature search identified eight published reports in blood donors comparing US FDA-approved noninvasive and invasive devices to venous samples run on hematology analyzers utilizing Bland-Altman analysis (Table 1). The proportion of samples in a clinically relevant range (<13.5 g/dL) was not consistently described, and none performed subset analysis of samples in this range. Only one study reported the 95% CI for the mean bias or upper and lower LOAs.13 As mentioned, the 95% CI of the upper LOA attempts to define the band within which the population limit lies. These shortcomings may limit the studies’ interpretation and generalizability of results, but the LOAs compared to venous reference values in all studies of noninvasive devices are wider than those with invasive devices. Invasive devices were also generally better at preventing inappropriate donor acceptance and deferral. Most studies have concluded with cautionary recommendations regarding noninvasive methods. Alternately, could noninvasive Hb measurement devices be the first part of a stepped Hb screening strategy? For example, Hb levels could be initially measured with a noninvasive device. Values below an appropriate multiple of the device-specific SDD plus cutoff would trigger a second measurement with a more-precise invasive device. Sümnig and coworkers15 suggested that a combined noninvasive-invasive approach with the Pronto-7 rescued donors with Hb values below cutoff (an unusual problem with the HemoCue 301 in their study) or those not yielding test results on that device. With regard to operational issues, Baart and coworkers mention that noninvasive devices are generally easy to use but that finger size complicated NBM-200 measurements, that the Haemospect required frequent resetting, and that donor movement delayed both noninvasive measurements. While 65% of donors unequivocally preferred the noninvasive measurement, 16% preferred the most accurate measurement, whichever it was. A much smaller study comparing the Pronto-7, NBM-200, and Haemospect noted technical errors with the Pronto-7 because of ambient light and donor skin color or nail polish.14 Likewise, the NBM-200 demonstrated sensitivity to ambient light and required the longest time to obtain test results at 115 seconds.14 No studies have reported a no-result rate for the NBM-200, but one did report a 7% unsuccessful rate for the Pronto-7 due to nail polish or artificial nails and cold or callused fingers.15 Despite known limitations, point-of-care devices measuring fingerstick Hb using capillary blood samples have gained wide acceptance. The LOAs in all studies of noninvasive devices are wider than those with fingerstick Hb devices in comparison with venous reference values. This observation hinders adoption of currently approved devices as a sole Hb screening method at lower donor Hb values. Devices without linearity in the higher range of observed Hb also preclude apheresis device programming without another method to assess donor Hb. If noninvasive device accuracy and precision can be improved, however, increased donor comfort without accompanying small decrements in the margin of safety may be realized. When comparative studies uniformly report the upper 95% confidence limit of the upper LOA within clinically relevant (i.e., decision-making) ranges of Hb values, study interpretation will also become substantially easier. The authors have disclosed no conflicts of interest. Hany Kamel, MD e-mail: [email protected] Ralph R.