Significant Variability in the Identification and Reporting of Band Neutrophils by Participants Enrolled in the College of American Pathologists Proficiency Testing Program: Time for a Change

A CAP HCMC subcommittee (study authors) designed the study and created survey materials. The goals of the study were 2-fold: (1) to understand current white blood cell (WBC) manual differential reporting practices among CAP PT participants, with a specific focus on band neutrophil reporting, and (2) to investigate and compare laboratories' performance in identifying band and segmented neutrophils on a peripheral blood smear via digital microscopy.

To address the first goal, the CAP HCMC subcommittee designed an 11-question survey that was distributed to CAP PT participants in the 2020-A mailing of the multiple Hematology Automated Differential (FH) and Virtual Peripheral Blood Smear (VPBS) surveys to assess laboratory size, practice setting, patient population, and WBC reporting practices (Supplemental Table 1; see supplemental digital content, containing 2 tables and supplemental material, at https://meridian.allenpress.com/aplm in the June 2024 table of contents). Survey respondents who provided complete responses to survey question (SQ) 1 (“What automated WBC differentials are reported by your instrument?”) or SQ2 (“Does your laboratory perform manual differentials?”) were included in the analysis (details in the supplemental digital content). These SQs were included in all 9 CAP Hematology PT Surveys (ie, FH1, FH2, FH3, FH4, FH6, FH9, FH10, FH13, and VPBS) from the 2020-A mailing. Given that the same laboratory could participate in multiple surveys and therefore answer SQs multiple times, a deduplication process was used to ensure that each laboratory would provide only one set of laboratory practice answers to the SQs (supplemental digital content). Percentages of responses were calculated based on the number of total respondents for a given SQ, excluding laboratories that provided no response.

To assess the second goal of evaluating the reproducibility of identifying bands, a virtual peripheral blood smear with 12 cells premarked for identification was distributed through the VPBS-B-2020 CAP PT survey. Given that a band count is often clinically requested in the setting of neonatal sepsis, the HCMC members chose a smear from a 40-week 3-day-old neonate admitted shortly after birth for decreased movement and depressed respiratory rate with Apgar scores of 3, 7, and 8 at 1, 5, and 10 minutes after birth, respectively. The complete blood cell count (CBC) was as follows: WBC 15.8 × 10⁹/L, red blood cells 4.75 × 10¹²/L, hemoglobin 17.3 g/dL, hematocrit 47.6%, mean corpuscular volume 100 fL, and platelets 202 × 10⁹/L. A Wright-Giemsa–stained peripheral blood smear was scanned and viewed using DigitalScope technology (DigitalScope, https://www.digitalscope.org/). Cells were selected by the CAP HCMC subcommittee members by consensus review of the scanned slide, using definitions from the CAP HCMC glossary (supplemental digital content).²⁰ 

A total of 12 cell identifications (cell IDs) were selected: 3 cells thought to represent easily identifiable (“easy”) segmented neutrophils because of the presence of clear multilobation and threadlike filaments lacking chromatin (virtual differential [VDIFF] -05, -09, and -13); 2 cells representing easy bands because of C-shaped configuration of the nuclei indented to more than half the distance to the farthest nuclear margin yet without condensation of chromatin to a single filament (VDIFF-02 and -10); 2 cells representing moderately challenging (“moderate”) bands with constriction connecting 2 lobes yet clearly visible chromatin between the dark, parallel nuclear margins (VDIFF-06 and -07); 2 cells representing difficult-to-classify (“difficult”) bands given the presence of tight constriction but still-discernible nuclear chromatin and thus meeting the CAP glossary definition (VDIFF-03 and -11); and 3 cells representing metamyelocytes with nuclei indented to less than half of the maximal nuclear diameter (VDIFF-04, -08, and -12). In addition, the VPBS participants and the HCMC subcommittee members were asked to perform a 100-cell WBC differential on the VPBS, enumerating band forms separately from segmented neutrophils.

Laboratory performance on the cell IDs and 100-cell WBC differential from the 2020-B VPBS Survey were linked with the supplemental question response data from the 2020-A mailings (supplemental digital content, Supplemental Table 2). Chi-square tests of association were performed to assess differences in cell ID challenges, with a Bonferroni correction applied to the critical value for these tests. All data editing, summarizations, and statistical testing were performed using SAS (version 9.4, SAS, Cary, North Carolina).

Among the 9359 kits sent in the 2020-A mailing of the CAP Hematology PT surveys, 7880 survey respondents (84.2%) provided answers to SQ1 and/or SQ2. The deduplication process from 7880 respondents, to provide only one set of practice answers per laboratory, resulted in inclusion of 6444 unique participant laboratories (81.8%). The Results section focuses on key data. The complete survey responses are presented in Tables 1 and 2. Hematology laboratory workflows, use of automated and digital imaging systems, and monthly volumes of CBC with differential are discussed in the Supplemental Results.

Among 6422 respondents to SQ2 (“Does your laboratory perform manual differentials?”), 5314 (82.7%) performed manual differential counts (manual differentials); of note, 22 of 6444 total participating laboratories did not respond to SQ2. Table 3 highlights that manual differentials are performed in nearly all practice settings. The practice settings with the lowest proportion of laboratories performing manual differentials were those classified in the “other” category, with 833 of 1335 (62.4%) performing manual differentials. This category includes blood banks, research and development institutions, rehabilitation facilities, military hospitals, Native American clinics, student centers, occupational health centers, and laboratories that did not provide an answer to their institution type, that is, no response to SQ9 (n = 182). On the other hand, the practice settings with the highest proportion of laboratories reporting manual differentials were in community practice (2829 of 3047; 92.8%) and veterans hospitals (202 of 210; 96.2%).

Among 5314 laboratories that performed manual differentials, 5165 responded to SQ3 (“What percentage of all reported differentials are manual differentials?”), with 149 providing no response. Of the 5165 responders, the reported percentage of manual differential counts (% manual diff rate) were as follows: less than 5% manual diff rate, 1540 of 5165 (29.8% of laboratories), 6% to 10%, 1346 of 5165 (26.1% of laboratories), 11% to 20%, 1247 of 5165 (24.1% of laboratories) and 21% or more, 1032 of 5165 (20.0% of laboratories). The % manual diff rate varied based on practice setting. A low % manual diff rate of less than 5% was most commonly reported by independent/commercial laboratories (227 of 518; 43.8%), followed by “other” laboratories (333 of 792; 42.0%), veterans hospital laboratories (79 of 200; 39.5%), and national/regional reference laboratories (106 of 291; 36.4%). For most practice settings (eg, community, independent/commercial, veterans hospital, “other”), the number of laboratories performing lower % manual diff rates exceeded those performing progressively higher % manual diff rates. For example, of the 2751 community setting laboratories, 693 (25.2%) reported %manual diff rates of less than 5%; 790 (28.7%) reported a 6% to 10% manual diff rate; 749 (27.2%) reported an 11% to 20% manual diff rate; and 519 (18.9%) reported a 21% or higher manual diff rate. The exception to this was university/academic clinical laboratories, where the largest category fell in the high % manual diff rate; of the 613 total laboratories in this category, 102 (16.6%) reported a less than 5% manual diff rate, 128 (20.9%) reported a 6% to 10% manual diff rate, 176 (28.7%) reported a 11% to 20% manual diff rate, and 207 (33.8%) had a 21% or higher manual diff rate.

Interlaboratory differences were noted in response to question 4 (“Who performs manual differentials in your laboratory?”) (Table 2). Among 5271 responding laboratories, in 4787 (90.8%) manual differentials were performed by all technologists, whereas in the remaining 484 (9.2%) only a subset of technologists performed manual differentials.

Table 3 details band and reference range reporting practices by laboratory setting. Of the 5314 laboratories that performed manual differentials, 5268 responded to SQ6 (“Does your lab report band neutrophil counts/percentages on manual differentials?”), with 46 providing no response. Band neutrophils were always reported in manual differentials by 4029 of 5268 responding laboratories (76.5%), and sometimes, per laboratory policy and/or per specific patient population or clinical scenario, in an additional 525 of 5268 (10.0%). Combining these 2 categories, bands were reported in more than 80% of all laboratories regardless of practice setting (total n = 4554 of 5268; 86.4%), ranging from 428 of 521 independent/commercial laboratories (82.1%) to 178 of 200 veterans hospital laboratories (89.0%).

Of the 4554 laboratories that always or sometimes reported bands, 4412 laboratories (96.9%) responded to SQ8 (“Does your laboratory provide a reference range when reporting band counts?”), with 142 providing no response. A significant subset (1190 of 4412; 27.0%) did not provide band reference ranges. Overall, 3222 of 4412 (73.0%) of laboratories provided defined band reference ranges with band counts (ie, absolute count reference ranges alone, percentage reference ranges alone, or both absolute and percentage reference ranges), varying from 154 of 233 (66.1%) national/regional reference laboratories to 396 of 522 (75.9%) university/academic laboratories.

Of 6444 total laboratories, 6262 responded to SQ9 (“What institution type is your laboratory affiliated with (choose one that best describes your practice)?”), with 182 providing no response. Community (n = 3051 of 6262; 48.7%), independent/commercial (n = 720 of 6262; 11.5%), and university/academic (n = 737 of 6262; 11.8%) settings were the largest survey participants, together representing 72.0% of all responding participants (Table 1). Responses among these 3 laboratory settings were compared for SQ6 (“Does your laboratory report band neutrophil counts/percentages?”) and for SQ7 (“If your laboratory reports bands for specific patient populations, clinical scenarios and/or at clinician request, which departments send the specimens?”) (Figure 1; Tables 2 and 3). There were no significant differences in band neutrophil reporting among different practice settings. In each of these settings, a majority always reported bands (Figure 1, A, orange slices): community, 2213 of 2816 (78.6%); independent local/commercial, 355 of 521 (68.1%); and university/academic, 472 of 628 (75.2%). An additional subset of these laboratories sometimes reported bands per policy or request (Figure 1, A, blue slices): community, 258 of 2816 (9.2%), independent local/commercial, 73 of 521 (14.0%), and university/academic, 68 of 628 (10.8%). Only a minor fraction of the laboratories never reported bands (Figure 1, A, gray slices): community, 345 of 2816 (12.3%); independent local/commercial, 93 of 521 (17.9%); and university/academic, 88 of 628 (14.0%).

There were slight differences among requesting departments (SQ7; overall n = 525 of 5268 [10.0%]) (Table 2). For example, in university/academic laboratories, band reporting requests were highest from neonatal intensive care units (ICUs). In the community hospital setting, neonatal ICU, emergency department/urgent care, and general pediatrics were the most common clinical services. On the other hand, in the independent/commercial settings, adult hematology/oncology and general pediatrics were the primary requesting departments, though other departments followed closely after (Figure 1, B).

Of 6444 total laboratories, 6021 responded to SQ11 (“Where is your laboratory located?”), with 423 providing no response. Of the 6021 respondents, 4983 (82.8%) were from the United States, 481 (8.0%) were from Canada, and 557 (9.3%) were from other countries (Table 1). Manual differentials were performed in 4130 of 4983 US laboratories (82.9%), 410 of 481 Canadian laboratories (85.2%), and 423 of 557 non-US/non-Canadian laboratories (75.9%). Interestingly, striking differences in band reporting practices were present between US and Canadian laboratories and between US and non-US/non-Canadian laboratories: only 368 of 4130 US laboratories (8.9%) never reported bands, compared with 92 of 423 non-US/non-Canadian laboratories (21.7%) and 214 of 410 Canadian laboratories (52.2%) (P < .001) (Figure 2, A; Table 4). Laboratories that sometimes performed band counts per policy and/or clinical request included 341 of 4130 US laboratories (8.3%), 49 of 410 Canadian laboratories (12.0%), and 89 of 423 non-US/non-Canadian laboratories (21.0%). The neonatal ICU appeared to be the top requesting department in the United States and Canada (Table 4). Differences in frequency of requests from the adult ICU and adult and pediatric hematology/oncology departments existed. In comparison with the United States, for example, the adult hematology/oncology department was the top requester in non-US/non-Canadian laboratories (44 of 89; 49.4%), whereas US (101 of 341; 29.6%) and Canadian (3 of 49; 6.1%) laboratories received fewer requests from this department (P < .001).

VPBS-B responses ranged from 1321 to 1328 per cell ID challenge overall (Figure 3). The best performance for intended responses by laboratory participants was observed for easy segmented neutrophils (99.7%, VDIFF-05; 99.6%, VDIFF-13; and 99.5%, VDIFF-09) (Figure 3, A) and one of the easy band neutrophils (98.3%, VDIFF-10) (Figure 3, B). There was a wide spectrum of performance for the intended response for the rest of the bands, ranging from 3.1% to 39.0% (for moderate to difficult band neutrophils VDIFF-03, VDIFF-11, VDIFF-06, and VDIFF-07) to 78.0% (for one of the easy band neutrophils, VDIFF-02) (Figure 3, B). Notably, when not classified as band neutrophils, bands were almost exclusively classified as segmented neutrophils, such that when responses for band and segmented neutrophils were combined as “mature neutrophils,” performance was excellent (98.1%–100.0%). Finally, laboratories displayed a spectrum of performance for identification of metamyelocytes (Figure 3, C). For VDIFF-04, 88.0% (n = 1166) gave the intended identification of metamyelocyte, whereas 12.0% (n = 159) identified this as a band. For VDIFF-12, 54.3% (n = 732) of participants replied with the intended identification of metamyelocyte, 19.8% (n = 262) classified the cell as a band neutrophil, and 24.9% (n = 330) classified it as a segmented neutrophil. Similar results were present for VDIFF-08, with 53.6% (n = 708) giving the intended identification of metamyelocyte versus 31.7% (n = 420) as band neutrophil and 14.6% (n = 193) as segmented neutrophil.

Of the 1328 total laboratories that responded to the morphologic cell ID VPBS 2020-B mailing, 1165 (87.7%) had additionally participated in the VPBS 2020-A survey questionnaire. The morphologic cell ID and survey questionnaire response data sets were linked together to conduct further analysis and statistical testing.

Statistical tests for differences in cell ID challenges with less than 85% agreement (5 cell IDs) were performed against 8 selected SQs, for a total of 40 specific tests analyzed. Although 5 associations tested displayed numeric statistical significance, no consistent pattern emerged (Supplemental Results). Laboratory practices and characteristics (namely, whether a laboratory had high versus low CBC with differential testing volumes, had generalist versus specialist technologists reviewing blood smears, had digital imaging systems, reported band neutrophils, reported band reference ranges, was located in the United States versus Canada versus other countries, or was an academic/university versus community versus independent/commercial laboratory setting) did not reliably predict better performance for the intended cell IDs.

Moreover, regardless of laboratory characteristics, there was excellent performance in identifying segmented neutrophils (98.2%–100% agreement rate) yet increased variability in identifying band neutrophils (ranges from 72.7%–100.0% on the “easy” identifications to 0.0%–60.0% for “moderate” to “difficult” identifications). Interestingly, metamyelocyte identifications were similarly variable, with performance ranging from 18.2% to 100.0% (Supplemental Table 2).

A 100-cell manual differential count was performed by 1311 VPBS challenge participants and 9 CAP HCMC subcommittee members. Figure 4, A, displays interlaboratory variability among CAP PT survey participants in enumerating segmented neutrophils as a percentage of all WBCs (mean, 66.9%; range, 27.0%–93.0%; coefficient of variation [CV], 15.0%) versus band neutrophils (mean, 15.2%; range, 0%–50%; CV, 55.8%). Variability was significantly improved when segmented and band forms were grouped together (mean, 82.0%; range, 48.0%–96.0%; CV, 6.2%). Similar improvement in variability was seen among the CAP HCMC subcommittee members when segmented neutrophils were grouped with bands (Figure 4, B): segmented neutrophils, mean 47.7% (range, 28.0%–61.0%; CV, 24.1%); band neutrophils, mean 32.6% (range, 18.0%–53.0%; CV, 32.9%); and combined segmented and band forms, mean 80.2% (range, 74.0%–87.0%; CV, 5.0%).

Position papers spanning several decades have highlighted inaccuracies in band neutrophil enumeration, questioned its clinical utility, and advocated against its use. Moreover, with the availability of hematology CBC analyzers that generate 6-part automated differentials and the current availability of other biomarkers for sepsis, it had been believed that many laboratories had discontinued reporting bands separately from segmented neutrophils. However, the CAP HCMC has had to address multiple inquiries about band counts in recent years. Therefore, a goal of this study was to evaluate the current state of laboratory practices and determine how many laboratories had instituted the recommendations to discontinue band reporting. The inquiries received by the CAP HCMC suggested that band reporting persists and that clinical providers continue to request band enumeration. Thus, we also sought to identify which clinical departments request band counts. Another study objective was to assess interlaboratory variability in band reporting and evaluate whether certain factors (ie, practice setting, laboratory workload volumes, proportion of differential counts performed and reported as manual differentials, general versus specialty technologists performing manual differentials) were associated with better performance in band enumeration. Our study found that a striking number of laboratories currently report band counts and that interlaboratory and interobserver concordance of band identification is poor, regardless of any of the factors we surveyed.

Band counting in clinical sepsis evaluation is entrenched in medical practice and further propagated by its inclusion in several clinical assessment algorithms. The 1991 systemic inflammatory response syndrome criteria, for example, include more than 10% bands as a criterion,¹⁶  as does the St John sepsis protocol.^17,18  An additional study showed that a subset of patients with culture-proven sepsis have normal WBC counts but elevated band counts.²¹  However, studies have shown that band counts are unhelpful in evaluating sepsis,^1,22,23  particularly in patients with normal WBC counts,⁵  and have little added benefit when absolute neutrophil counts and cells more immature than bands (ie, immature granulocytes [IGs], now available as a category on some modern CBC analyzers) are taken into consideration.⁴ 

There are many known causes of variability in band classification, whether due to interobserver or intraobserver variation, and they include the inherent statistical fallibility of a 100-cell enumeration (well explained by the Rumke statistic¹⁵), different definitions of bands,^11–13,24  poor interobserver agreement despite training using a single strict definition,^2,25  and fundamental challenges in classifying cells that fall within a morphologic spectrum.²⁶  One of the limitations of our study is that we did not collect data to account for intraobserver variation. However, it was demonstrated in prior studies^4,27  that even when manual counting was extended to a 500-cell differential or to a higher band count threshold (eg, >15% bands), band counts still showed minimal value for identifying culture-proven bacterial infection in patients with a normal total leukocyte count. Others^2,25  demonstrated that using strict definitions for bands and pretraining did not improve interobserver variability. Furthermore, neutrophilia and bandemia are not specific to infection and/or sepsis, as they are common in other inflammatory states, including tissue damage, stress responses, and drug effects.¹  Thus, monitoring changes in band counts over time in a patient more frequently than not may represent the inherent imprecision of the method rather than actual changes in the patient's status.

Prior to this study, the HCMC subcommittee members had assumed that many, if not most, laboratories had discontinued enumeration of bands.^9,10  The results of this survey are surprising, as they revealed that an overwhelming majority of survey respondents who performed manual differentials still reported band counts (4554 of 5268; 86.4%). Moreover, a significant subset (4029 of 5268; 76.5%) always enumerated band neutrophils separately from segmented neutrophils, whereas only 525 of 5268 (10.0%) did so under specific clinical circumstances. It is unclear whether the laboratories that always enumerated band forms continued to do so for historical reasons, or whether they had attempted discontinuation of this practice and were unable to do so because of clinical demands. Of note, when participants were stratified by location, we observed that fewer than half of participating laboratories in Canada reported bands. This suggests wider acceptance of recommendations to discontinue band reporting in Canadian laboratories.

One of the strengths of this study is that more than 1300 participants had access to the same digitally scanned blood smear for band enumeration, whereas previous band studies had shown only a selection of cells for identification. The inability to consistently identify bands translated to the high interobserver variability seen in the band count (CV, 55.8%), despite access to the same blood smear. Even among the CAP HCMC members who used a single morphologic definition,^12,20  significant imprecision persisted in the interobserver variability of the band count, although it was somewhat less (CV, 32.9%) than in the CAP PT survey respondents.

One limitation of our study is that we did not mandate a single definition for band counting by survey participants, yet this likely provides a more accurate assessment of current band reporting practice in clinical laboratories. Furthermore, as previously discussed, even with training on a single band definition, high interobserver variability persists.^2,25  We did not assess which band definition each laboratory used. Regarding morphologic definitions of bands,^11–13,24  on the restrictive end of the practice spectrum, any constriction thinner than one-third of the nuclear width would be defined as a segmented neutrophil, resulting in fewer enumerated bands. In contrast, the liberal end of the definition spectrum holds that any nuclear material wider than a clear-cut visible strand devoid of chromatin between nuclear lobes should be classified as a band neutrophil, thus resulting in many more bands. In this study, we attempted to limit interobserver variability by using the CAP glossary definition as a standard.²⁰  However, the poor interobserver/interlaboratory comparability in the evaluation of the moderate and difficult band neutrophil identifications was in part attributable to different morphologic definitions used by laboratories to identify bands.

Methods with high interlaboratory variability may be partially mitigated with appropriately verified reference ranges; yet among laboratories reporting band counts, only 73.0% reported accompanying reference ranges. In addition, for methods prone to interobserver variability, laboratories need to closely assess interobserver comparability within the laboratory to assure the clinical performance of the results. By way of illustration, a study of febrile infants used the same reference range for absolute band counts between 2 different institutions in different states.²⁴  One laboratory's band count practice resulted in a 10-fold higher count than the other laboratory. This would have led to misclassification in the first institution of infants as having elevated band counts with no other risk factor for serious bacterial infections. In this survey, we did not assess how the laboratories reporting band reference ranges verified the clinical performance of the reference range. For the laboratories that did not report band reference ranges, we did not assess whether an appropriate comment, such as “The reference interval(s) has not been established,” is used. However, CAP-accredited laboratories reporting band counts must include accompanying verified reference intervals, interpretive ranges, or interpretive comments to be in compliance with CAP checklist item HEM.36820.

Identification of segmented neutrophils that the HCMC subcommittee classified as easy achieved near-perfect correlation among surveyed laboratories. However, for moderate to difficult band neutrophil identifications, performance was poor, with only 3.1% to 39.0% of the respondents providing the intended response. Interestingly, no consistent trends in laboratory characteristics reliably predicted better performance on the intended morphologic cell IDs. The accuracy of band identification was no better in laboratories that had only specialists performing WBC differentials than in laboratories where generalist technologists performed the differentials. In fact, even when CAP HCMC subcommittee members used the CAP glossary definition to separately enumerate bands in the 100-cell manual differential virtual microscopy challenge, significant interobserver variability remained. These findings highlight the inherent difficulties in band identification and enumeration: neither a laboratorian's experience level nor application of a uniform band definition yielded improvements in interobserver agreement.

In this study, when band counts were performed by clinician request, the most frequent requests came from neonatal ICUs, general pediatrics, and adult hematology oncology. Although the survey did not specifically identify the reason for these clinical requests, it is likely that they were made in the setting of evaluation for sepsis. This begs the question: what alternative methods are available for rapid identification of sepsis, if band counts are unreliable? The most recent sepsis initiative (Sepsis-3) focuses on Sequential Organ Failure Assessment scores; this score supplants the Sepsis-2/systemic inflammatory response syndrome classification, which had included band enumeration as a criterion.^28–30  Sepsis-3 is designed for adult patients in ICU settings. However, in neonatology, rapid evaluation of sepsis is also an especially common and clinically important scenario.^31,32  Antibiotic intervention is essential to positive outcomes; but conversely, antibiotic use in nonseptic, bacterial culture–negative neonates carries significant risks.³³ 

For these reasons, band neutrophil trending as a marker of sepsis is particularly embedded in neonatology literature.^34–38  In particular, the immature to total neutrophils (I:T) ratio requires enumeration of bands (with immature forms including bands, metamyelocytes, and myelocytes, whereas total neutrophils include segmented forms, bands, metamyelocytes, and myelocytes), with a neonatal I:T ratio less than 0.2 considered as unlikely to indicate sepsis. However, some neonatology groups no longer use the I:T ratio (and thus, manual band neutrophil trends) for this indication because of poor interobserver concordance in the laboratory and because of specific clinical scenarios where neutrophilic left shift is not robust in the setting of sepsis, for example, low WBC/neutrophil counts in preterm babies of mothers with preeclampsia/hemolysis, elevated liver enzymes, and low platelets (HELLP) syndrome^38,39  (Joern-Hendrik Weitkamp, MD, written communication, December 13, 2021). Indeed, predictive biomarkers not reliant on the band neutrophil count have emerged, including C-reactive protein⁶  and procalcitonin, endorsed by American Academy of Pediatrics policy (Joern-Hendrik Weitkamp, MD, written communication, December 13, 2021), as have other biomarker combinations.^7,40  Procalcitonin and C-reactive protein not only are appropriate detectors of sepsis but also have the benefit of monitoring and guiding discontinuation of antibiotics.^41,42 

Several studies have evaluated automated IG counts—a parameter combining circulating metamyelocytes, myelocytes, and promyelocytes—and found them to be of some utility in sepsis assessment.^43–46  Overall, no single biomarker is independently sensitive or specific. Automated IG counts, however, offer the advantage of better precision, decreased turnaround time, decreased cost, and decreased interinstrument/interlaboratory variability.⁴⁷  Additional advances in hematology instrumentation have led to several studies on cell population data as sepsis markers, including some that have shown promise.^48–50  Because cell population data parameters are instrument specific, they lack widespread applicability.

There is evidence that the limitations of manual morphologic band identification are not well understood. As an example, a published study evaluated “automated versus manual band counts” in the diagnosis of invasive bacterial infections in febrile infants.⁵¹  It is unclear whether “automated band counts” refers to IG counts or to classification of band neutrophils using digital images. This study highlights continued misperceptions in clinical medicine about the performance of band enumeration: that it is fully automated, accurate, and rapid. The reality is that it is manually performed, leads to prolonged test turnaround times, and has significant interobserver variability.

In summary, band neutrophils continue to be commonly reported by a majority of CAP PT–participating laboratories despite interobserver variability in cell identification, ample literature documenting their limited value and reliability, and the availability of alternative methods for biomarkers for identifying sepsis. Together with the fact that manual band counts are laborious and expensive,²⁶  and the reality of an aging and diminishing laboratory technologist workforce with the experience and skill set necessary to perform manual methods, the findings of this study support the CAP HCMC's strong recommendation to discontinue reporting band neutrophils separately from segmented neutrophils. This study, among others, can be used in discussions with clinical providers to set evidence-based institutional policies that eliminate band neutrophil reporting.

We thank Joern-Hendrik Weitkamp, MD, and William Walsh, MD, Neonatology, Monroe Carell Jr Children's Hospital at Vanderbilt University Medical Center, for helpful discussions and context.