Advertisement

Real-world data from Europe and Africa suggest that accuracy of systems for self-monitoring of blood glucose is frequently impaired by low hematocrit

Open AccessPublished:May 11, 2021DOI:https://doi.org/10.1016/j.diabres.2021.108860

      Highlights

      • Hematocrit outside common glucose meter performance limits is not uncommon.
      • 3% of outpatients in Europe have hematocrit outside commonly specified ranges.
      • This is especially the case for older subpopulations with higher diabetes prevalence.
      • In South Africa, low hematocrit is even more common, and across all age strata.
      • Patients diagnosed with diabetes should be carefully matched with SMBG instruments.

      Abstract

      Aims

      Certain systems for self-monitoring of blood glucose (SMBG) demonstrate inaccuracy at low and high hematocrit (HCT). Manufacturers define HCT ranges for accurate performance. Our objective was to assess the frequency of HCT values that can lead to clinically relevant errors.

      Methods

      In this cross-sectional study, we collected real-world data representing over 360,000 outpatients from the Netherlands (NL), the Czech Republic (CZ), and South Africa (ZA). These were subsequently stratified by sex and age and compared to commonly specified HCT range limits, reference intervals, and data from 1780 healthy Czech subjects.

      Results

      HCT values were comparably distributed in NL and CZ. Outpatients had a higher dispersion of values than healthy subjects. Low HCT values in Europe were common in age groups with a high prevalence of diabetes. All ZA age groups showed a higher prevalence of low HCT than in Europe.

      Conclusions

      Real-world data indicate that SMBG systems specified to perform only within the frequently used 30–55% HCT range would leave 3% of outpatients in Europe and 18% in South Africa at risk of false SMBG results, with individual age strata being substantially higher. This could affect their diabetes management. Adequate SMBG systems should thus be chosen.

      Keywords

      Abbreviations:

      COPD (Chronic Obstructive Pulmonary Disease), CZ (Czech Republic), HCT (Hematocrit), NL (The Netherlands), SANHANES (South African National Health and Nutrition Examination Survey), SMBG (Self-Monitoring of Blood Glucose), TB (Tuberculosis), ZA (South Africa)

      1. Introduction

      Studies have demonstrated that the inaccuracy of systems for self-monitoring of blood glucose (SMBG) may result in overlooking hypoglycemia, increased frequency of hypo- and hyperglycemia, increased glycemic variability and increased HbA1c [
      • Kilpatrick E.S.
      • Rumley A.G.
      • Myint H.
      • Dominiczak M.H.
      • Small M.
      The effect of variations in haematocrit, mean cell volume and red blood cell count on reagent strip tests for glucose.
      ,
      • Tang Z.
      • Lee J.H.
      • Louie R.F.
      • Kost G.J.
      Effects of different hematocrit levels on glucose measurements with handheld meters for point-of-care testing.
      ,
      • Ginsberg B.H.
      Factors affecting blood glucose monitoring: sources of errors in measurement.
      ,
      • Breton M.D.
      • Kovatchev B.P.
      Impact of blood glucose self-monitoring errors on glucose variability, risk for hypoglycemia, and average glucose control in type 1 diabetes: an in silico study.
      ,
      • Lyon M.E.
      • Lyon A.W.
      Patient acuity exacerbates discrepancy between whole blood and plasma methods through error in molality to molarity conversion: “Mind the gap!”.
      ,
      • Pfützner A.
      • Schipper C.
      • Ramljak S.
      • Flacke F.
      • Sieber J.
      • Forst T.
      • et al.
      Determination of hematocrit interference in blood samples derived from patients with different blood glucose concentrations.
      ]. Therefore, SMBG systems need to be accurate. It is well-known that a low hematocrit (HCT) can lead to inaccurately high blood glucose values in certain SMBG systems. This can result in insulin overdosing, which in turn can lead to hypoglycemia, even in a hospital setting [

      Baltaro RJ, Dangott BJ, Tanenberg RJ, Thombare A, Lin SF. Evaluation of Hematocrit and Hypoglycemia in a University Hospital: Blood Glucose Monitoring Experience Using Retrospective Data Analysis. Am. Diabetes Assoc. ADA 75th Sci. Sess., Boston, Massachusetts; 2015. https://doi.org/10.1177/193229681300700122.

      ]. Conversely, a high HCT can lead to an underestimation of blood glucose values and potentially to hyperglycemia.
      We wanted to assess the likelihood of real-world hematocrit distributions having an impact on the accuracy of SMBG systems.
      Manufacturers of SMBG systems provide specified HCT ranges within which these devices operate reliably. However, these intervals differ substantially between SMBG systems: while many have a specified HCT range from 30 to 55%, others can reliably measure at an HCT as low as 25%, 20%, or even 10% or lower. Additionally, a previous study has shown that even within these self-specified ranges, the specified limits are frequently not maintained [
      • Hattemer A.
      • Wardat S.
      Evaluation of Hematocrit Influence on Self-Monitoring of Blood Glucose Based on ISO 15197:2013: Comparison of a Novel System With Five Systems With Different Hematocrit Ranges.
      ]. Understanding the distribution of low and high HCT is key to deciding what effect a given specified HCT range has on patient safety and choosing a suitable SMBG system accordingly.
      Anemia prevalence differs significantly across countries [

      Kassebaum NJ, GBD 2013 Anemia Collaborators. The Global Burden of Anemia. Hematol Oncol Clin North Am 2016;30:247–308. https://doi.org/10.1016/j.hoc.2015.11.002.

      ,
      • World Health Organization
      The global prevalence of anaemia in 2011.
      ]. While anemia is defined by the hemoglobin concentration, low hemoglobin is usually associated with a low HCT. It is thus likely that HCT varies similarly and with it, the risk of inadequate SMBG systems being used by people with diabetes. It is, therefore, not possible to verify the distribution in one or two locations only and make global generalizations.
      We set out to verify the frequency of low and high HCT across several countries, including one with a significantly higher prevalence of anemia than is common in Europe. That way, we estimated the risk of patients having HCT values outside the HCT ranges commonly specified by manufacturers of SMBG systems. In addition, to verify our methodology and to compare our results with reference intervals published elsewhere, we compared the HCT values of outpatients with those of a healthy cohort tested in one of the participating centers.

      2. Subjects

      We evaluated outpatient samples from three different countries with varying prevalence of anemia. Due to the intended scope, using healthy people would have been logistically and ethically problematic. On the other hand, the HCT of inpatients is often subject to temporary variation due to injury, surgery, or medical therapy, and as a result, does not allow conclusions on the general population. While the prevalence of people with diabetes among outpatients may not be entirely representative of the general population, we are not aware of any reasons why outpatient cohorts should contain less people with diabetes than the general population. We thus decided to focus on analyzing outpatients.
      We compiled retrospective real-world data of measurements without any further selection criteria than the availability of HCT measurements and an available status as in- or outpatient from three clinical laboratories in the Netherlands (NL), the Czech Republic (CZ), and South Africa (ZA). The latter was included to verify the influence of a high prevalence of anemia compared to the European locations [

      Kassebaum NJ, GBD 2013 Anemia Collaborators. The Global Burden of Anemia. Hematol Oncol Clin North Am 2016;30:247–308. https://doi.org/10.1016/j.hoc.2015.11.002.

      ]. In NL, all full blood count requests from 6/2016-7/2017 were collected, in CZ all hematology samples containing HCT from 01/2015-12/2016 were assessed and in ZA, all full blood counts for outpatients at Tygerberg Hospital from 10/2015-01/2018 were included. The data were pseudonymized and only retraceable to actual patients by the providing institution. Data from patients not clearly recorded as being outpatients (e.g., patients that shifted from outpatient status to inpatient status the same or the following day) were excluded. Table 1 lists the participating institutions, the number of HCT values received, and the number of outpatient samples evaluated from each center.
      Table 1Participating Institutions and their Contributions.
      InstitutionTotal number receivedMeasurements evaluatedIndividual outpatients evaluated
      Certe Medical Diagnostics and Advice, Leeuwarden, The Netherlands579,296421,045:

       241,394 f

       179,637 m

       14 ns
      163,950:

       95,133 f

       68,804 m

       13 ns
      General University Hospital, Prague, Czech Republic500,000283,860:

       167,883 f

       115,977 m

       0 ns
      104,251:

       62,940 f

       41,311 m

       0 ns
      Tygerberg Hospital, Cape Town, South Africa258,323249,810:

       136,471 f

       112,882 m

       457 ns
      92,811:

       50,613 f

       42,082 m

       116 ns
      Overview of contributing institutions and samples evaluated. Samples of uncertain gender were included in general analysis but excluded from gender-specific analysis. In no case was their number high enough to have an influence on the results (f = female, m = male, ns = not specified).
      In CZ, 283,860 measurements were recorded. For NL, the data set contained 421,045 measurements, and lastly, in ZA, there were 249,810 measurements. As outpatient status from external sources could not be reliably ascertained for South Africa, only confirmed outpatients from the hospital hosting the laboratory itself were included.

      3. Materials and methods

      Instruments used were CellDyn Sapphire (Abbott, Chicago, USA, used in NL), Sysmex XN (Sysmex, Kobe, Japan, used in CZ) and ADVIA 2120i (Siemens, Erlangen, Germany, used in ZA).
      We considered all measurements tagged with the date of the measurement and basic patient information (gender, age in years, pseudonymized patient identifier, and clear outpatient status). To achieve a more stringent, though conservative, estimate for the frequency of out-of-range HCT values, samples with HCT < 10% were excluded as unlikely to be physiological, e.g., mislabeled samples that were not blood but other body fluids. The data from each institution were evaluated independently of each other to account for possible differences in allocation criteria and outpatient populations. In the process, we followed the STROBE checklist for cross-sectional studies wherever applicable and feasible [
      • von Elm E.
      • Altman D.G.
      • Egger M.
      • Pocock S.J.
      • Gøtzsche P.C.
      • Vandenbroucke J.P.
      The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.
      ].
      In order to not include multiple measurements of a single medical episode, after counting an HCT value for a given outpatient, all following measurements of the same patient within a period of 28 days, starting with the day of the measurement, were excluded from further analysis; the total number of episodes can thus be greater than the number of individual outpatients if patients had multiple episodes over the time examined. The data were then evaluated on the frequency of values outside HCT ranges commonly specified for SMBG systems, and on the likelihood of patients presenting with such values.
      In addition, we calculated reference intervals for men and women using HCT data from 1780 healthy subjects from the Czech site, which were collected during employee health checkups. These were compared to the literature and the outpatient measurements.
      Statistical analysis was conducted in R 3.6.0 using the dplyr package. Density plots were generated using the ggplot2 package, with outpatient values as specified above, based on patients presenting with medical episodes. In addition, the percentage of medical episodes and patients below certain common lower measurement thresholds of SMBG systems were calculated.
      Since information on the department originating the complete blood count request was available for South Africa, we used that information to verify the influence of including or excluding hematology departments in the analysis, where hematological parameter measurements might be significantly distinct from other specialties. As the effect was marginal, we decided that neither having distinct exclusion criteria for South Africa nor requesting the respective information from the other sites was necessary or advisable.
      As the prevalence of both diabetes and anemia vary by sex and age, samples were likewise stratified by sex and age. To be able to compare with the frequency of diabetes/high HbA1c, we grouped patients into the strata used by EUROSTAT[

      Eurostat: European Health Interview Survey (EHIS) hlth_ehis_cd1e 2014. Available from http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=hlth_ehis_cd1e&lang=en [accessed 20 Nov 2020].

      ] and in part by SANHANES-1 [
      • Shisana O.
      • Labadarios D.
      • Rehle T.
      • Simbayi L.
      • Zuma K.
      • Dhansay A.
      • et al.
      South African National Health and Nutrition Examination Survey (SANHANES-1).
      ]: <15-years-old, 15–24, 25–34, 35–44, 45–54, 55–64, 65–74, and 75+ years of age.

      4. Results

      We assessed the frequency of medical episodes with HCT values outside commonly used specified HCT ranges by evaluating outpatient samples from the Czech Republic, the Netherlands, and South Africa. The frequency was assessed both overall (Table 2) and stratified by sex and age (Fig. 1). The results were also compared to healthy subjects from CZ (Table 3).
      Table 2Frequency of patients outside common Certified Measurement Range Limits.
      Individual outpatients<20% HCT<25% HCT<30% HCT>50% HCT>55% HCT>65% HCT
      CZ104,251229

      (0.22%)
      949

      (0.91%)
      3375

      (3.24%)
      2512

      (2.41%)
      177

      (0.17%)
      6

      (0.01%)
      NL163,950236

      (0.14%)
      1121

      (0.68%)
      4721

      (2.88%)
      2575

      (1.57%)
      229

      (0.14%)
      14

      (0.01%)
      ZA92,8111818

      (1.96%)
      5635

      (6.07%)
      13,854

      (14.93%)
      7798

      (8.40%)
      3138

      (3.38%)
      477

      (0.51%)
      Number and percentage of patients presenting at any point with a HCT below common lower and above common upper certified measurement range limits in the three locations over all patients.
      Figure thumbnail gr1
      Fig. 1Density Function of HCT values (a, c, e) and percentage of outpatients presenting with HCT <30% stratified by age and sex (b, d, f) in CZ (a, b), NL (c, d) and ZA (e, f). Light bars represent female patients, dark bars represent male patients.
      Table 3Czech Healthy Subjects and Outpatients – Reference Intervals and Relative Standard Deviation.
      2.5–97.5 percentile healthy (Reference interval)2.5–97.5 percentile outpatientsReference interval in the literatureRelative Standard Deviation (CV) healthyRelative Standard Deviation (CV) medical episodes
      Male40.7–49.7%26.9–50.4%41–50%5.4%12.8%
      Female35.7–45.9%28.5–46.4%36–46%6.3%10.7%
      Nonparametric reference limits (2.5–97.5 percentile) calculated from healthy Czech subjects; relative standard deviation (coefficient of variation) of the healthy subjects with outpatients of the same sex. Literature: Reference interval as per Heil and Erhard 2008 (12).
      In CZ, the eligible data comprised HCT values of 104,251 individual outpatients. 3.24% of the outpatients (~1 out of 31) presented with an HCT of <30% at least once throughout the observation period, 0.22% (~1 out of 460) had an HCT of <20%. The NL data could be traced back to 163,950 outpatients. Of these, 2.88% (~1 out of 35) had an HCT of <30%, 0.14% (~1 out of 700), even presented with an HCT <20%. The ZA data could be traced back to 92,811 outpatients. Of these, 14.93% (~1 out of 7) had an HCT of <30% and 1.96% (~1 out of 51), even one of HCT <20%.
      Comparison with healthy subjects from Prague (Table 3) showed that 20% of Czech female outpatients and 29% of Czech male outpatients were below the lower reference limits calculated from the measurements of the local healthy subjects, which closely matched those reported in the literature [

      Heil W, Ehrhardt V. Reference Ranges for Adults and Children - Pre-Analytical Considerations. 9th ed. Mannheim; 2008.

      ]. The distribution of HCT values in both male and female outpatients was much wider than that for corresponding healthy subjects (Fig. 1A, C, E).
      Further stratification of the individual measurement results by age (Fig. 1B, D, F) showed that among adult patients, the lowest frequency for HCT values <30% among women was found in the youngest group of 15–24 years, both in CZ (1.08%) and in NL (1.43%), though unlike in CZ, juveniles under 15 years in NL presented with a comparable frequency of low HCT. For males, the lowest frequency for adults was found in the adjoining age group, 25–34 years old, again both in CZ (0.60%) and NL (0.20%).
      More importantly, the percentage of patients with an HCT <30% increased markedly with advancing age, peaking at 7.09% and 6.78% for Dutch male and female patients aged 75+, respectively. Similarly, we found the highest frequency in Czechs of the same age group, with 7.66% for males and 6.92% for female outpatients. All in all, in this age group, 1 in 14 patients presented with an HCT <30% in both countries, and 1 in 21 (NL) or 18 (CZ) medical episodes started with an HCT <30%. This finding is noteworthy since the prevalence of diabetes also increases among older people, with a maximum of almost 20% for men aged 75+ in NL and over 25% for men and women aged 75+ in CZ [

      Eurostat: European Health Interview Survey (EHIS) hlth_ehis_cd1e 2014. Available from http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=hlth_ehis_cd1e&lang=en [accessed 20 Nov 2020].

      ].
      In ZA, the distribution of low HCT among age strata differed from those in Europe, as low HCT was more prevalent in younger subjects. In women, the frequency was highest among patients aged 15–34 years at 18.89–19.99%, and for adult women, lowest among 55–64 year-olds, with 10.15%. The overall lowest values were observed for girls of 6–10 years. Men had little overall age-related variation, with the highest prevalence for adults of 12.69% among 35–44-years-olds, but generally a frequency of over 9% post-puberty.
      Patients with HCT >55% were comparatively rare in Europe and usually confirmed by the laboratory as patients suffering from polycythemia with correct HCT results. In ZA, high HCT was substantially more frequent.
      Based on our initial question, we investigated whether HCT values outside ranges commonly specified for SMBG systems have a real-world importance as a confounding factor for SMBG measurements, and on what basis specified ranges for HCT can be derived that do not leave large numbers of patients at risk of erroneous SMBG measurements. We could confirm that reference intervals, being based on healthy subjects and excluding the lower and upper 2.5% in that situation, are not a good basis. The distribution of HCT in outpatients differs greatly from healthy subjects, with low HCT being a particularly common occurrence, especially in South Africa.

      5. Discussion

      Our results suggest that SMBG systems with a specified range of 30% HCT and above leave a significant number of patients at risk of clinically relevant inaccurate readings, while specified ranges of ≥20% HCT and especially ≥10% HCT minimize that risk. We could demonstrate that in both CZ and NL, at least one in 35 patients presented with an HCT <30% at some point. This was increasingly more probable with advancing age, occurring in up to 1 out of 18 medical episodes and 1 out of 14 patients 75+ years of age when diabetes also becomes increasingly prevalent – to the point of one in five (NL) or one in four (CZ) people presenting with diabetes [

      Eurostat: European Health Interview Survey (EHIS) hlth_ehis_cd1e 2014. Available from http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=hlth_ehis_cd1e&lang=en [accessed 20 Nov 2020].

      ]. These groups seem to have a substantial risk of presenting with both a low HCT and diabetes and are at risk of erroneous SMBG results. This could lead to incorrect insulin dosage, potentially resulting in hypo- or hyperglycemia.
      Low HCT is omnipresent across age groups in ZA, including the working-age population. Conditions leading to low HCT such as iron deficiency and anemia of chronic disease (ACD), e.g., caused by chronic infections (HIV, schistosomiasis, hookworm infections, malaria etc.), are significantly more common than in most of Europe [
      • World Health Organization
      The global prevalence of anaemia in 2011.
      ,
      • Kassebaum N.J.
      • Jasrasaria R.
      • Naghavi M.
      • Wulf S.K.
      • Johns N.
      • Lozano R.
      • et al.
      A systematic analysis of global anemia burden from 1990 to 2010.
      ,

      Moonasar D (coordinator), National Department of Health Republic of South Africa Malaria Elimination Strategic Plan 2019 – 2023 2019. Available from: http://www.health.gov.za/index.php/2014-08-15-12-54-26/category/532-2019-strategicdocuments?download=3570:malaria-elimination-strategic-plan-for-south-africa-2019-2023 [accessed 13 Nov 2020].

      ]. Furthermore, the City of Cape Town is one of the Tuberculosis (TB) hotspots in South Africa [

      South Africa National AIDS Council (SANAC). South Africa’s national strategic plan for HIV, TB and STIs 2017-2022. Pretoria: South Africa National AIDS Council; 2017. Available from: https://sanac.org.za/thenational-strategic-plan/ [accessed 10 Aug 2020].

      ]. TB frequently leads to anemic conditions, contributing to the observed frequency of low HCT. At the same time, in a smaller portion of cases, it has also been reported that pulmonary TB may lead to elevated HCT [
      • Abay F.
      • Yalew A.
      • Shibabaw A.
      • Enawgaw B.
      Hematological Abnormalities of Pulmonary Tuberculosis Patients with and without HIV at the University of Gondar Hospital, Northwest Ethiopia: A Comparative Cross-Sectional Study.
      ,
      • Kahase D.
      • Solomon A.
      • Alemayehu M.
      Evaluation of peripheral blood parameters of pulmonary tuberculosis patients at St. Paul’s Hospital Millennium Medical College, Addis Ababa, Ethiopia: comparative study.
      ], which could contribute to the higher frequency of elevated HCT discussed below. As per the South African National Health and Nutrition Examination Survey SANHANES-1 [
      • Shisana O.
      • Labadarios D.
      • Rehle T.
      • Simbayi L.
      • Zuma K.
      • Dhansay A.
      • et al.
      South African National Health and Nutrition Examination Survey (SANHANES-1).
      ], 9.5% of South Africans have an HbA1c >6.5%, indicating diabetes, and the actual magnitude of what has been called a “burgeoning diabetes epidemic” in ZA is unknown due to significant underdiagnosis [
      • Pheiffer C.
      • Pillay-van Wyk V.
      • Joubert J.D.
      • Levitt N.
      • Nglazi M.D.
      • Bradshaw D.
      The prevalence of type 2 diabetes in South Africa: a systematic review protocol.
      ]. The high percentage of low HCT in the working population (up to 1 out of 6 patients) thus poses a substantial risk to ensuring adequate treatment: individuals may be equipped with an SMBG system not suitable for their individual needs. The common low HCT during child-bearing age could also pose a problem in the context of potential gestational diabetes, which has been observed in up to 25.8% of pregnancies in ZA [
      • Adam S.
      • Rheeder P.
      Screening for gestational diabetes mellitus in a South African population: Prevalence, comparison of diagnostic criteria and the role of risk factors.
      ], depending on the criteria applied.
      In other sub-Saharan African countries, the frequency of low HCT is likely to be similar or worse, as anemia is even more prevalent: especially in the tropical belt, chronic malaria, thalassemia, sickle cell anemia, intestinal parasites, chronic infections as well as malnutrition (especially iron deficiency) are even more common [

      Kassebaum NJ, GBD 2013 Anemia Collaborators. The Global Burden of Anemia. Hematol Oncol Clin North Am 2016;30:247–308. https://doi.org/10.1016/j.hoc.2015.11.002.

      ,
      • Kassebaum N.J.
      • Jasrasaria R.
      • Naghavi M.
      • Wulf S.K.
      • Johns N.
      • Lozano R.
      • et al.
      A systematic analysis of global anemia burden from 1990 to 2010.
      ]. The fact that Africa has the highest diabetes growth rate in the world exacerbates the problem [

      IDF Diabetes Atlas. 8th ed. Brussels: International Diabetes Federation; 2017.

      ].
      That high HCT is more frequently observed in ZA than in Europe can be explained by a variety of conditions leading to hypoxic conditions: In a study on variances in the prevalence of chronic obstructive pulmonary disease (COPD) across the globe, Cape Town showed a high prevalence of Global Initiative for Chronic Obstructive Lung Disease (GOLD) stage II and stage III-IV COPD. It also showed one of the lowest prevalences of healthy, unobstructed airflow [
      • Buist A.S.
      • McBurnie M.A.
      • Vollmer W.M.
      • Gillespie S.
      • Burney P.
      • Mannino D.M.
      • et al.
      International variation in the prevalence of COPD (the BOLD Study): a population-based prevalence study.
      ]. COPD, especially in these advanced stages, can lead to secondary erythrocytosis [
      • El-Korashy R.I.
      • Amin Y.M.
      • Moussa H.A.
      • Badawy I.
      • Bakr S.M.
      Study the relationship of erythropoietin and chronic obstructive pulmonary disease.
      ]. The above-mentioned high prevalence of TB in Cape Town is also likely to contribute to this observation.
      In addition, low HCT may have been observed in all centers due to injury, surgery, or illness. Although these cases typically do not represent the normal population, this could be problematic for people with diabetes as well, when blood sugar levels are measured with an SMBG system not specified for a low HCT range.
      Our study has several limitations: from the data available to us, we cannot exactly reconstruct who of the measured patients really had diabetes. Obtaining the diagnosis for all patients in such a study to clearly identify diabetes patients only is virtually impossible for this kind of real-world data studies. As such, the actual probability of people with diabetes having out-of-range HCT is not directly derivable. We see, however, no obvious reason as to why people with diabetes should have a higher HCT than the average outpatient. On the contrary, advanced diabetes often precipitates renal complications with a possible consequence of anemia [
      • Mehdi U.
      • Toto R.D.
      Anemia, diabetes, and chronic kidney disease.
      ].
      A further limitation of our study is the definition of the outpatient status. This status generally refers to patients who were not hospitalized overnight but can also, in some cases refer to patients who were hospitalized later. We excluded these cases where we could identify them with the approach described in the methods section. Nonetheless, the real-world data analyzed in this study is representative for a genuine cohort of patients – and many hospitals are confronted with comparable patients in daily routine.
      Some of our samples are repeated measurements of the same patient. Since our data is allocable to individual patients, we avoided inflating the results by analyzing multiple measurements of already included patients with potentially unchanged medical conditions (presumably additionally biased by a higher frequency of visits during acute disease). We did so by only counting the measurement when patients first presented themselves at the hospital and refraining from analyzing any subsequent values within the following four weeks. We then analyzed the frequency of such medical episodes and of individual patients presenting with low or high HCT, rather than the total number of low or high measurements.
      Lastly, we acknowledge that having included only data from a large academic hospital in ZA, due to the structure of the healthcare system, we may have captured an above-average ratio of patients whose case was so complex as to require escalation and referral to this level (e.g. requiring close cooperation of several distinct specialties). The actual frequency of extreme values in the general population may be somewhat lower.
      There may be other confounders we have not considered or controlled for, but we do not think their effect size to be large enough to have a relevant impact on our general conclusion of underestimation of the importance of the issue of out-of-range HCT.
      To our knowledge, this is the largest study using real-world data to assess the potential impact of a low HCT as a confounding factor in SMBG. The data presented clearly shows that reference intervals based on healthy subjects are not a suitable basis for choosing adequate HCT range specifications for SMBG systems. We show that great care must be taken to match the SMBG system chosen to the specific health status of the user. Information collected in several Asian countries and presented recently at the 2019 conference of the International Diabetes Federation (IDF) in Busan, South Korea, showed that a significant number of the SMBG systems offered only have a specified HCT range of 30–50% or 30–55% [

      Hinzmann R. Accuracy of blood glucose monitoring is essential for correct therapy decisions 2019. Congress of the International Diabetes Federation (IDF) 2019, Roche Satellite Symposium; 2019. Busan, South Korea.

      ]. While analyzing SMBG systems available in the observed countries was out of scope of this study, the information from Asia suggests that the mere availability of an SMBG system on the market is not enough to ensure its appropriateness for a given patient. The problem is, of course, that few people with diabetes actually know their current HCT or if/when it has changed. In the absence of an up to date HCT, SMBG systems with a specified range of 30–55% HCT are evidently inadequate as a catch-all solution. Based on our results, we would recommend a specified range of, at a minimum, 25–55% HCT, better 20–60% HCT, with 10–65% HCT being optimal.

      Acknowledgements

      The authors would like to thank Dr. Sami Wardat (Roche Diabetes Care) for his help with project coordination and Dr. Tony Huschto (Roche Diabetes Care) for his help with data processing. Z.C. would like to thank Mr. Wessel Kleinhans for his help with data collection.

      Declaration of Competing Interest

      R.H. and D.M. are employees of Roche Diabetes Care GmbH. As proprietor of Dr. Hauss Training & Consulting, O.H. is providing consulting, training, and medical writing services. The other authors have no duality of interest to disclose. No other potential conflicts of interest relevant to this article could be identified.

      Author Contributions

      R.H. is the guarantor of this work and wrote the manuscript with contributions by O.H. and D.M.; T.Z., M.S., H.S. and Z.C. provided the data sets analyzed in this study and reviewed and edited the manuscript.
      Prior Presentation: Parts of this study were presented on a poster and in abstract form at the 78th Scientific Sessions of the American Diabetes Association, Orlando, FL, 22-26 June 2018.

      Funding

      No third-party funding was received for this study.

      References

        • Kilpatrick E.S.
        • Rumley A.G.
        • Myint H.
        • Dominiczak M.H.
        • Small M.
        The effect of variations in haematocrit, mean cell volume and red blood cell count on reagent strip tests for glucose.
        Ann Clin Biochem. 1993; 30: 485-487https://doi.org/10.1177/000456329303000513
        • Tang Z.
        • Lee J.H.
        • Louie R.F.
        • Kost G.J.
        Effects of different hematocrit levels on glucose measurements with handheld meters for point-of-care testing.
        Arch Pathol Lab Med. 2000; 124: 1135-1140https://doi.org/10.1043/0003-9985(2000)124<1135:EODHLO>2.0.CO;2
        • Ginsberg B.H.
        Factors affecting blood glucose monitoring: sources of errors in measurement.
        J Diabetes Sci Technol Online. 2009; 3 (https://dx.doi.org/10.1177%2F193229680900300438): 903-913
        • Breton M.D.
        • Kovatchev B.P.
        Impact of blood glucose self-monitoring errors on glucose variability, risk for hypoglycemia, and average glucose control in type 1 diabetes: an in silico study.
        J Diabetes Sci Technol. 2010; 4: 562-570https://doi.org/10.1177/193229681000400309
        • Lyon M.E.
        • Lyon A.W.
        Patient acuity exacerbates discrepancy between whole blood and plasma methods through error in molality to molarity conversion: “Mind the gap!”.
        Clin Biochem. 2011; 44: 412-417https://doi.org/10.1016/j.clinbiochem.2011.01.005
        • Pfützner A.
        • Schipper C.
        • Ramljak S.
        • Flacke F.
        • Sieber J.
        • Forst T.
        • et al.
        Determination of hematocrit interference in blood samples derived from patients with different blood glucose concentrations.
        J Diabetes Sci Technol. 2013; 7: 170-178https://doi.org/10.1177/193229681300700122
      1. Baltaro RJ, Dangott BJ, Tanenberg RJ, Thombare A, Lin SF. Evaluation of Hematocrit and Hypoglycemia in a University Hospital: Blood Glucose Monitoring Experience Using Retrospective Data Analysis. Am. Diabetes Assoc. ADA 75th Sci. Sess., Boston, Massachusetts; 2015. https://doi.org/10.1177/193229681300700122.

        • Hattemer A.
        • Wardat S.
        Evaluation of Hematocrit Influence on Self-Monitoring of Blood Glucose Based on ISO 15197:2013: Comparison of a Novel System With Five Systems With Different Hematocrit Ranges.
        J Diabetes Sci Technol. 2018; 12: 333-340https://doi.org/10.1177/1932296818757550
      2. Kassebaum NJ, GBD 2013 Anemia Collaborators. The Global Burden of Anemia. Hematol Oncol Clin North Am 2016;30:247–308. https://doi.org/10.1016/j.hoc.2015.11.002.

        • World Health Organization
        The global prevalence of anaemia in 2011.
        World Health Organization, Geneva2015
        • von Elm E.
        • Altman D.G.
        • Egger M.
        • Pocock S.J.
        • Gøtzsche P.C.
        • Vandenbroucke J.P.
        The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies.
        J Clin Epidemiol. 2008; 61: 344-349https://doi.org/10.1016/j.jclinepi.2007.11.008
      3. Eurostat: European Health Interview Survey (EHIS) hlth_ehis_cd1e 2014. Available from http://appsso.eurostat.ec.europa.eu/nui/show.do?dataset=hlth_ehis_cd1e&lang=en [accessed 20 Nov 2020].

        • Shisana O.
        • Labadarios D.
        • Rehle T.
        • Simbayi L.
        • Zuma K.
        • Dhansay A.
        • et al.
        South African National Health and Nutrition Examination Survey (SANHANES-1).
        HSRC Press, Cape Town2013
      4. Heil W, Ehrhardt V. Reference Ranges for Adults and Children - Pre-Analytical Considerations. 9th ed. Mannheim; 2008.

        • Kassebaum N.J.
        • Jasrasaria R.
        • Naghavi M.
        • Wulf S.K.
        • Johns N.
        • Lozano R.
        • et al.
        A systematic analysis of global anemia burden from 1990 to 2010.
        Blood. 2014; 123: 615-624https://doi.org/10.1182/blood-2013-06-508325
      5. Moonasar D (coordinator), National Department of Health Republic of South Africa Malaria Elimination Strategic Plan 2019 – 2023 2019. Available from: http://www.health.gov.za/index.php/2014-08-15-12-54-26/category/532-2019-strategicdocuments?download=3570:malaria-elimination-strategic-plan-for-south-africa-2019-2023 [accessed 13 Nov 2020].

      6. South Africa National AIDS Council (SANAC). South Africa’s national strategic plan for HIV, TB and STIs 2017-2022. Pretoria: South Africa National AIDS Council; 2017. Available from: https://sanac.org.za/thenational-strategic-plan/ [accessed 10 Aug 2020].

        • Abay F.
        • Yalew A.
        • Shibabaw A.
        • Enawgaw B.
        Hematological Abnormalities of Pulmonary Tuberculosis Patients with and without HIV at the University of Gondar Hospital, Northwest Ethiopia: A Comparative Cross-Sectional Study.
        Tuberc Res Treat. 2018; 2018: 5740951https://doi.org/10.1155/2018/5740951
        • Kahase D.
        • Solomon A.
        • Alemayehu M.
        Evaluation of peripheral blood parameters of pulmonary tuberculosis patients at St. Paul’s Hospital Millennium Medical College, Addis Ababa, Ethiopia: comparative study.
        J Blood Med. 2020; 11: 115-121https://doi.org/10.2147/JBM.S237317
        • Pheiffer C.
        • Pillay-van Wyk V.
        • Joubert J.D.
        • Levitt N.
        • Nglazi M.D.
        • Bradshaw D.
        The prevalence of type 2 diabetes in South Africa: a systematic review protocol.
        BMJ Open. 2018; 8https://doi.org/10.1136/bmjopen-2017-021029
        • Adam S.
        • Rheeder P.
        Screening for gestational diabetes mellitus in a South African population: Prevalence, comparison of diagnostic criteria and the role of risk factors.
        S Afr Med J. 2017; 107: 523-527https://doi.org/10.7196/SAMJ.2017.v107i6.12043
      7. IDF Diabetes Atlas. 8th ed. Brussels: International Diabetes Federation; 2017.

        • Buist A.S.
        • McBurnie M.A.
        • Vollmer W.M.
        • Gillespie S.
        • Burney P.
        • Mannino D.M.
        • et al.
        International variation in the prevalence of COPD (the BOLD Study): a population-based prevalence study.
        Lancet Lond Engl. 2007; 370: 741-750https://doi.org/10.1016/S0140-6736(07)61377-4
        • El-Korashy R.I.
        • Amin Y.M.
        • Moussa H.A.
        • Badawy I.
        • Bakr S.M.
        Study the relationship of erythropoietin and chronic obstructive pulmonary disease.
        Egypt J Chest Dis Tuberc. 2012; 61: 53-57https://doi.org/10.1016/j.ejcdt.2012.10.026
        • Mehdi U.
        • Toto R.D.
        Anemia, diabetes, and chronic kidney disease.
        Diabetes Care. 2009; 32: 1320-1326https://doi.org/10.2337/dc08-0779
      8. Hinzmann R. Accuracy of blood glucose monitoring is essential for correct therapy decisions 2019. Congress of the International Diabetes Federation (IDF) 2019, Roche Satellite Symposium; 2019. Busan, South Korea.