Use of Complete Blood Cell Count Components to Screen for Hereditary Spherocytosis in Neonates

The study was approved by the institutional review boards at KPNC (1270417-17) and the University of California, San Francisco (16-19246). We obtained data for this retrospective cohort study from KPNC medical records and demographic, laboratory, and hospitalization databases.

We included all infants born at ≥35 weeks’ gestation from 1995 through 2015 at 15 KPNC hospitals. KPNC is an integrated community medical care delivery system that serves 4.3 million members, nearly one-half of the insured population of Northern California. Neonates were eligible for the primary analysis if they had a CBC with erythrocyte indices drawn within the first 7 days after birth.

CBCs at KPNC hospitals are performed by using Beckman Coulter (Brea, CA) or Sysmex Corporation (Munderlein, IL) hematology analyzers. We included only the first CBC with erythrocyte indices drawn per subject (within the first 7 days) to avoid the possible confounding effects of transfused erythrocytes (because we expected most transfusions would occur after an initial CBC) and declining MCV values over time. We also collected MCV data from CBCs drawn within the first 30 days after birth to determine how the MCV changes over time. The HS Index was calculated by dividing the MCHC (expressed as grams hemoglobin per deciliter RBC) by the MCV (expressed as femtoliters). We obtained the highest total serum bilirubin (TSB) value drawn during the first month after birth for each subject. We excluded any TSB measurements for which corresponding conjugated or direct bilirubin measurement constituted ≥50% of the TSB level because these levels suggest liver disease as a cause of hyperbilirubinemia. Universal bilirubin screening began in some facilities in September 2004 and was fully implemented in all facilities by February 2007.¹⁸  Additional laboratory and clinical characteristics obtained from electronic records included direct antiglobulin test results, sex, gestational age, birth weight, maternal and infant race and ethnicity, receipt of phototherapy, readmission for phototherapy, and receipt of an exchange transfusion or RBC transfusion. Infant race and/or ethnicity was used as a proxy variable to reflect the varying risk of HS on the basis of ancestry.

The outcome variable assessed was the presence or absence of HS. We identified patients with possible HS by searching KPNC electronic records for (1) encounters with International Classification of Diseases, Ninth Revision or International Classification of Diseases, 10th Revision codes for HS (282.0 or D58.0, respectively) and (2) subjects without HS International Classification of Diseases, Ninth Revision or International Classification of Diseases, 10th Revision codes but who had osmotic fragility and/or ektacytometry tests ordered (the 2 most common HS confirmatory tests performed at KPNC) during follow-up. We then verified that these patients have HS by reviewing their electronic and paper charts, with particular attention to pediatric hematology/oncology consult notes. We collected data on the age at diagnosis, timing of RBC transfusion(s), hospital admission for RBC transfusion, hemoglobin values, and splenectomy. We required either (1) a positive diagnostic test (osmotic fragility, ektacytometry, eosin-5'-maleimide [EMA] binding, and/or DNA sequencing) or (2) a positive family history and typical clinical features or laboratory findings (receipt of an RBC transfusion, splenectomy, anemia, elevated reticulocytes, and/or elevated spherocytes on serial CBCs) to confirm a diagnosis of HS. Deidentified study data were collected and managed by using the Research Electronic Data Capture (REDCap) tools hosted at the University of California, San Francisco.

We compared demographic, laboratory, and clinical characteristics of infants with and without HS by using t tests or χ² tests as appropriate. We estimated the prevalence of HS by race and ethnicity to determine the previous probability of HS for each race and ethnicity and compared these previous probabilities via χ² tests. We divided our cohort into 4 separate groups on the basis of highest TSB value (TSB values: <10 mg/dL, 10–14.9 mg/dL, 15–19.9 mg/dL, and ≥20 mg/dL). We performed χ² tests to determine if there is a difference in the previous probability of HS between TSB groups. We expected the prevalence of HS to increase with rising peak TSB values.

We quantified the discrimination of the HS Index by using the area under the receiver operating characteristic curve (AUROC). We estimated LRs for different HS Index intervals irrespective of peak TSB level. HS Index intervals were chosen on the basis of inflection points in the observed receiver operating characteristic curve. LRs were estimated as the proportion of HS cases divided by the proportion of noncases for a particular HS Index interval.¹⁷  We estimated LRs for the various peak TSB categories irrespective of HS Index values so that we could take into account severity of hyperbilirubinemia when determining the probability of HS. We compared the discrimination of the HS Index by peak TSB levels by comparing the AUROC. Although not specified a priori, we evaluated the discrimination of the red cell distribution width (RDW) and estimated LRs for various RDW intervals. We performed all analyses by using Stata/SE 14 (Stata Corp, College Station, TX), supplemented by a Stata command that we developed to calculate interval LRs (available from the authors).

Of the 670 272 infants included in the study on the basis of their year of birth and gestational age, 79 infants were diagnosed with HS (Fig 1), resulting in a prevalence of 1.2 per 10 000 infants. There were 5 infants whose diagnosis was unclear because of loss to follow-up; these infants were excluded from the analysis. Of the subjects with HS, 50 infants had a CBC with erythrocyte indices performed within 7 days of birth, 63 infants had a TSB within 30 days of birth, and 48 infants had both a CBC and TSB. Characteristics of the newborns with and without HS included in the study are shown in Table 1. The prevalence of HS was higher among infants of white and “other” (presumably multiple) race and ethnicity categories and increased with higher peak serum bilirubin levels (P < .001 for linear trend; Table 1). The risk of HS was higher among infants needing phototherapy, requiring readmission for phototherapy, with TSB values that crossed the exchange transfusion threshold and who received an exchange transfusion or an RBC transfusion during the first month of life.

The majority of infants diagnosed with HS (85%, n = 67) had confirmatory testing performed. A diagnosis was made because of a positive family history of HS and typical clinical or laboratory findings in 15% (n = 12) of HS subjects (Supplemental Table 5). Among subjects with HS, 78% had a positive family history, 80% were diagnosed by 5 years of age, 59% received phototherapy, 29% received an RBC transfusion, and 30% underwent splenectomy during follow-up (Table 2). The median follow-up time for the HS subjects was 7.2 years (interquartile range: 3.1–14.4). The majority of RBC transfusions occurred during the first year after birth and led to admission to the hospital (Table 2).

There were significant differences in hemoglobin, hematocrit, RBC indices, HS Index values, and RDWs between neonates with and without HS (Table 3). The mean HS Index value of neonates with HS was higher than that of the reference group at 0.358 (SD: 0.023) vs 0.327 (SD: 0.029). However, there is a substantial amount of overlap when comparing HS Index distribution by HS status (Fig 2). HS infants had a mean MCV of 99.1 fL vs 104.8 fL among non-HS infants. If we included all CBCs (n = 163) drawn per HS subject within the first 30 days of birth, the mean MCV decreased to 95.3 fL. Discrimination of the HS Index as a continuous variable was good, with an AUROC of 0.84 (95% confidence interval [CI] 0.78–0.90; Fig 3). The LRs for the HS Index ranged from 10.1 for an HS Index ≥0.380 to 0.1 for an HS Index <0.310 (Table 4). Dichotomized at 0.36, the HS Index had sensitivity of 56%, specificity of 93%, and positive predictive value of 0.3%.

Discrimination of the HS Index was not significantly different among infants who did and did not receive phototherapy (P = .9). The HS Index appeared to perform best among infants with TSB levels <10 mg/dL and was similar among the other categories (P < .001). LRs for the peak TSB categories are shown in Table 4. When we dichotomized the HS Index (at ≥0.355) and the TSB levels (at ≥15 mg/dL), the LRs for all 4 combinations of results were similar to those predicted from the products of the LRs for each test result, suggesting that the HS Index and TSB are independent tests for HS (Supplemental Table 6).

Because the difference in mean RDW between neonates with and without HS was substantial (20.0% vs 16.9%), we also evaluated the discrimination of RDW, which was excellent at 0.91 (95% CI 0.86–0.96) and better than that of the HS Index (Fig 3). LRs for the RDW categories are shown in Table 4.

In this retrospective cohort study, we found that the sensitivity and specificity of the HS Index were lower than previously reported and its usefulness limited by the low previous probability of HS. Consider a white infant with no family history of HS who has a peak serum bilirubin of 20 mg/dL and an HS Index of 0.38. Assuming the peak TSB and HS are independent tests for HS, we can multiply the pretest odds of HS among white infants in our population (1.67 per 10 000) by the LR for TSB ≥20 mg/dL (9.1) and LR for HS Index ≥0.380 (10.1) to obtain the post-test probability. Even with both tests abnormal, the probability of HS in this infant remains low at ∼1.5% (0.000167 × 9.1 × 10.1) and unlikely to prompt us to pursue further diagnostic testing.

On the other hand, obtaining an HS Index or RDW costs little and, if elevated, may prompt a physician to treat jaundice more aggressively. For example, one might treat an infant with a high HS Index and/or high RDW, high bilirubin, and elevated reticulocytes as having hemolytic disease (particularly if there is a family history of HS), thus starting phototherapy at a lower bilirubin level.

The HS Index may be more useful in ruling out HS in neonates who have a family history of HS. If, for example, an infant has both a low HS Index of <0.310 and a TSB <10 mg/dL, the probability of HS would decrease from 50% (assuming autosomal dominant HS) to ∼1% (previous odds of HS × LR for HS Index <0.310 × LR for TSB <10 = 1 × 0.1 × 0.1). Therefore, we could be reassured that the infant is unlikely to have inherited HS and would defer confirmatory studies.

There are several possible explanations for the lower utility of the HS Index in our study compared with that of the report by Christensen et al.^6,11  Our population had a lower prevalence of HS, at 1.2 per 10 000. This is partially explained by the diversity of KPNC members, with a lower proportion of white members (41%), who have the highest incidence of HS. However, even among white infants, the prevalence of HS at ∼1.7 in 10 000 was lower than the expected 2 to 5 in 10 000. It is possible that we missed diagnoses among subjects with mild, asymptomatic HS. Furthermore, infants born at the later end of the cohort may not have undergone workup for HS, either because they had not yet become clinically symptomatic or because diagnostic testing is usually performed after 12 months of age. Newer DNA sequencing panels were rarely ordered in our cohort, which may have resulted in missed cases because of equivocal or negative results on other confirmatory tests or misdiagnosis as another type of hereditary anemia.^9,19  We also did not investigate whether subjects with substantially elevated HS Index and/or RDW values had clinical findings consistent with an undiagnosed case of HS. However, even in the unlikely event that we missed as many as 20% of HS cases, the main effect on our results would be only a slight (∼20%) underestimation in the previous and posterior probability of HS, which would not affect our conclusions.

Our decision to include only the first CBC drawn per infant in the first week after birth had a large impact on our disparate findings. Because neonatal hyperbilirubinemia peaks during the first week, CBCs to evaluate jaundice are likely to be drawn then as well, making them the most useful in screening for HS. In Christensen’s cohort, the mean MCV among HS infants was 89.9 fL (compared with 107 fL among the reference group),¹¹  whereas, in our cohort, HS infants had a mean MCV of 99.1 fL (compared with 104.8 fL among non-HS infants). As expected, the mean MCV among HS infants in our cohort fell to 95.4 fL when we included all CBCs drawn per HS subject within the first 30 days and would likely decrease further had we included CBCs out to 90 days. This supports our hypothesis that the inclusion of multiple CBCs for HS infants and only the initial CBC for the reference group contributed to the difference in HS Index values between the 2 groups.

In previous studies, it has been revealed that an elevated RDW (reflective of anisocytosis) is typical among individuals with HS.⁸  In children >1 year of age, an RDW >14% had an 85% sensitivity and 97% specificity for identifying HS, with a LR of 28.3.¹⁴  However, the utility of the RDW among neonates with HS has not, to our knowledge, been previously described. Among 13 neonates in the Intermountain Healthcare System with HS who had a TSB in the high-risk zone²⁰  and a CBC drawn within the first week of life, the mean RDW was elevated at 20.7% (SD: 1.5%),⁷  but the authors reported RDW had poor discrimination for identifying HS. In our cohort, the mean RDW among HS infants was similar at 20.1% (SD: 2.0%) but had better discrimination than the HS Index. It is important to note that RBC indices, in particular, the MCV (which is used to calculate the hematocrit and MCHC), vary slightly on the basis of the make and model of hematology analyzer used.^21–23  Differences between instruments may affect the precision of the intervals used to calculate LRs for the HS Index and RDW.

The discrimination of the HS Index was modest in neonates when obtained from a CBC drawn within the first week after birth. Although the RDW performed better than the HS Index as a screening tool in neonates, its usefulness is also limited by the low previous probability of HS, which is only modestly raised by hyperbilirubinemia. However, because both of these screening tests have minimal cost, they may be helpful in situations in which even a low post-test probability would alter management, such as prompting more aggressive treatment of jaundice, while performing further workup.

We thank Hamid Niki for constructing the study’s background data sets from electronic data.

AUROC

area under the receiver operating characteristic curve

CBC

complete blood count

EMA

eosin-5'-maleimide

HS

hereditary spherocytosis

KPNC

Kaiser Permanente Northern California

LR

likelihood ratio

MCHC

mean corpuscular hemoglobin concentration

MCV

mean corpuscular volume

RDW

red cell distribution width

TSB

total serum bilirubin