Diagnostic accuracy of reticulocyte hemoglobin content in Thai patients with microcytic red cells as a test for iron deficiency anemia

This was a prospective study of diagnostic accuracy. The study protocol was reviewed and approved by the Institutional Review Board of the Faculty of Medicine, Chulalongkorn University, Bangkok, Thailand (approval No. 222/2011). Every participant included signed an informed consent form. We included consecutive adult inpatients and outpatients (age ≥18 years), who had a MCV <80 fL and were treated at the King Chulalongkorn Memorial Hospital from August 1st, 2010 to January 31st, 2012. Exclusion criteria were pregnancy, creatinine clearance <30 mL/min (calculated using the Crockcoft–Gault equation), a history of iron supplement use within the previous 3 months, or an unclassifiable type of diagnosis.

Venous blood samples were collected by venipuncture and immediately processed (within 6 h) at the King Chulalongkorn Memorial Hospital laboratory for CBC and reticulocyte count with CHr, using an XE-AlphaN automated blood cell analyzer (Sysmex Corporation, Kobe, Japan). The serum ferritin and serum iron, total iron binding capacity (TIBC) were determined using an Elecsys 2010 immunoassay analyzer (Roche Diagnostics, Mannheim, Germany). Hemoglobin analysis was conducted using isoelectric focusing electrophoresis (IEF) or high performance liquid chromatography (HPLC), or both.

Three diagnostic groups were defined. An IDA group with a ferritin reference level <50 μg/L and hemoglobin level that rose ≥1 g/dL within one month after oral iron therapy. Patients in the thalassemia group had hemoglobin analysis showing thalassemia and/or hemoglobinopathy and/or confirmed α-thalassemia traits using gap-polymerase chain reaction (gap-PCR). We measured serum ferritin in all thalassemia samples to exclude concomitant IDA. The last group of patients had anemia of inflammation (AI), showed a ferritin level ≥200 g/L and a history of inflammatory diseases. Patients who could not be classified into one of these groups were excluded. The control participants were nonanemic healthy volunteers with normal MCV and normal red cell distribution width (RDW).

Data were analyzed using SPSS statistical software for Windows, version 15.0 (SSPS Inc., Chicago, IL, USA) to identify the cutoff value for optimal sensitivity and specificity using receiver operating characteristic (ROC) curves. Multiple comparisons were made using a one-way analysis of variance (ANOVA) at 95% confidence interval (CI) and the differences were considered significant at the 0.05 level. A Pearson correlation coefficient was used to define relationship between CHr and other variables.

The sample size was calculated at 95% confidence with a 6% acceptable error. At least 70 patients with microcytic anemia, 51 of whom were iron deficient, and at least 70 normocytic control samples were required to provide adequate power for this study.

We recruited 99 healthy volunteers into the control group and screened 181 patients with microcytic red blood cells. Sixteen patients were excluded because 2 were <18 years old, 2 had creatinine clearance <30 mL/min, 5 patients were in unclassifiable groups of diagnosis, and 7 patients had suspected -thalassemia trait (microcytic anemia with normal ferritin level), but a gap-PCR had not been performed to confirm α-thalassemia. Ultimately, 168 patients were enrolled in this study; 53 with IDA, 50 with AI, and 65 with thalassemia.

In the thalassemia group, there were 49 patients with thalassemia traits; 15 with hemoglobin E trait, 14 with homozygous hemoglobin E, 9 with β-thalassemia trait, 6 with hemoglobin E with α-thalassemia trait, 4 with α-thalassemia trait (confirmed by gap-PCR), and 1 with a β-hemoglobin variant. Sixteen patients with thalassemia disease were 4 with β-thalassemia with hemoglobin E, 7 with hemoglobin H diseases, 3 with hemoglobin H with hemoglobin CS, 1 with hemoglobin AE Bart’s with hemoglobin CS, and 1 homozygous hemoglobin CS.

The laboratory data are shown in Table 1. Patients with AI were significantly older than those in the other groups (Table 1). The mean CHr in this study was 27.3 pg. The laboratory normal range is 25–35 pg. The CHr values of patients in the AI and thalassemia trait groups were significantly less than those in the control group, but significantly higher than those in the groups with thalassemia disease or IDA. Patients in the group with thalassemia disease had significantly higher reticulocyte counts than those in all the other groups including IDA. Because of the non-normal distributions, reticulocyte and ferritin data were transformed logarithmically. Comparisons of the transformed data had similar outcomes to those shown in Table 1 (data not shown).

Table 1

The means (standard deviations) for age, sex, and red cell indices of the 3 groups of microcytic patients and healthy controls

Table 2

The correlation of age and other variables (95% CI)

A Pearson correlation analysis showed that CHr was positively correlated with hemoglobin (Hb) or degree of anemia (r = 0.77, P < 0.001), MCV (r = 0.86, P < 0.001), mean corpuscular hemoglobin (MCH, r = 0.89, P < 0.001), and mean corpuscular hemoglobin concentration (MCHC, r = 0.57, P < 0.001), but it was not correlated with age or RDW at 95% confidence interval (Table 2 and Figure 1).

Figure 1

Receiver operating characteristic (ROC) curve analysis was performed to determine appropriate cutoff values of CHr as shown in Figure 2. Using a CHr cutoff point of 24.6pg, the sensitivity was 73.6%, specificity was 74.8%, positive predictive value (PPV) was 42.4%, and negative predictive value (NPV) was 92.0% for diagnosis IDA for the whole population (Figure 2A). Many false positive samples were from thalassemia patients. Normal volunteers were not included in these calculations.

Figure 2

When patients in thalassemia groups (traits and diseases) was excluded, the cutoff point CHr of 27.2 pg showed a sensitivity of 84.9%, specificity of 81.2%, PPV of 60.8%, and NPV of 93.8% (Figure 2B). As there was a statistical difference in CHr between the thalassemia trait and thalassemia disease (P = 0.001), only thalassemia disease was excluded in a reanalysis. The result was similar to that obtained by excluding all thalassemia patients (Figure 2C).

In addition, MCHC of the IDA group was similar to thalassemia disease, but significantly lower than those of the other groups (Table 1). On ROC analysis using a MCHC of 31.6 g/dL as the cutoff point, the sensitivity was 81.1%, specificity was 83.7%, PPV was 81.5%, and NPV was 81.3% for differentiating IDA from thalassemia trait (Figure 2D).

Our study demonstrates a caveat in using CHr in diagnosis of IDA in Thais because of the high prevalence of thalassemia and hemoglobinopathy. Consistent with a previous study [6], the cutoff value of CHr for the diagnosis IDA in Thai patients (24.6 pg) is <28 pg as specified elsewhere [2] because of the higher prevalence of microcytic red cells that are correlated with low CHr. As shown in Table 1, the CHr is useful to differentiate IDA from AI and thalassemia trait, but cannot accurately differentiate IDA from thalassemia disease in patients with microcytic anemia. However, thalassemia disease has a distinctive feature of high reticulocyte counts that can differentiate it from all the other causes of microcytic anemia (Table 1). Furthermore, the present study shows that a low MCHC is another helpful parameter in differentiating IDA from thalassemia traits.

The differential diagnoses of microcytic anemia include IDA, AI, thalassemia trait, and thalassemia disease. As initial tests, CBC and reticulocyte count should be routinely investigated simultaneously. If absolute reticulocyte count is high (>100 × 10⁹/L), thalassemia disease should be considered (Figure 3) correlating with clinical history of long-standing anemia since childhood and physical examination of chronic hemolysis, such as splenomegaly and jaundice. If the CHr value is >24.6 pg, a diagnosis of IDA is unlikely (NPV 92.4%). The presence of chronic infection, inflammation or malignancy suggests AI and a positive family history suggests thalassemia trait.

Figure 3

Considering the group with low CHr (≤24.6pg) and a normal reticulocyte count, a MCHC <31.6 g/dL suggests the probability of IDA over thalassemia trait. The laboratory normal range of MCHC is 33–37 g/dL. By contrast, a MCHC >31.6 g/dL suggests a diagnosis of thalassemia trait. The combination of CHr and MCHC can differentiate IDA from thalassemia better than using CHr alone (sensitivity 69.8%, specificity 88.1%, PPV 59.7%, and NPV 92.0%). In patients with microcytic anemia with normal reticulocytes, CHr ≥24.6 pg and MCHC ≥31.6 g/dL, IDA is very likely (specificity over 90%) as shown in Figure 3. In this group, therapeutic diagnosis with oral iron supplement, which should be considered especially for premenopausal women. If hemoglobin rises >1 g/dL in a month, IDA is suggested. If hemoglobin does not improve, iron study and hemoglobin analysis should be performed for a definite diagnosis.

In conclusion, a CBC and reticulocyte count with CHr are useful investigations for Thai patients with microcytic red blood cells. However, CHr cannot clearly differentiate IDA from thalassemia, unless combined with information about reticulocyte count.