Variability in erythrocyte size, anisocytosis, is useful for the describing, classifying, or prognosing of pathophysiological phenomena in dogs (1–4), such as anemia (5–9), microcytosis (10, 11), quatrefoil red blood cells (12), cardiopulmonary diseases (13–18), pancreatitis (19), inflammatory bowel disease (20), or diabetes mellitus (21), and of physiological phenomena such as sex and aging (22, 23).

The degree of anisocytosis is typically expressed either as variation in erythrocyte diameters in stained blood films (24) and/or by increased red cell distribution width (RDW) in automated analyzer counts (25). Evaluating a blood smear by conventional microscopy requires experience (26) and is prone to subjective assessment, which may lead to variability within and between observers. Furthermore, different grading schemes exist, some with very simple categorizations such as “absent or present” (9, 22). Others have applied subjective grading on a scale from +1 to +4, with +1 being the smallest amount (27), or grading +1 to +4 based on the number of variable-sized erythrocytes in a monolayer at 1,000x microscopic field (28) or high-power field (HPF) (21). Anisocytosis has even been graded by the number of microcytes visible in the standard field of view at × 100 magnification (11).

Red blood cell distribution width (RDW) is generated by most modern automated hematological analyzers from the distribution curve of erythrocyte volume derived by impedance changes, originally developed by Coulter (29, 30), and as seen in Idexx Procyte Dx (31), Sysmex pocH-100iV Diff (32), and in Abbot Sapphire, by flow cytometry and 2-angle laser light scatter, as present in instruments from Siemens such ad Advia2120 (33). In the SysmexXE5000, the size distribution of the erythrocytes is measured using the Sheath Flow Direct Current method where the signal amplitude changes when a blood cell passes through an aperture of the detector (34). The International Council for Standardization in Hematology (ICSH) (35) recommended standardization of the statistical method for analysis of erythrocyte volume distribution and for the algorithms used for the calculation of RDW in 1990. Seemingly, different manufacturers use different statistical methods and algorithms. A method to harmonize RDW by recalculating values according to instrument-specific polynomial curves has been suggested (36). Further, in some analyzers, RDW is calculated from the distribution histogram at the 50% height level above the baseline (Abbott); others use 20% height level above baseline (Sysmex XE500 and Mindray) or determine RDW from the erythrocyte volume histogram in the window between 60 and 120 fL (Siemens), with RDW being calculated as the ratio expressed in percentage between the standard deviation (SD) of erythrocyte volumes and the mean cell volume (MCV) (36, 37). In some analyzers, such as the Idexx Procyte Dx, RDW is reported as RDW-CV and RDW-SD, the difference being that RDW can be reported either as percentage (RDW-CV) or in femtoliters representing a standard deviation from the mean (RDW-SD). Irrespective of the method, RDW results below the reference interval should not be considered clinically meaningful, whereas an increased RDW value reflects a greater difference in the size of erythrocytes, which can be due to the presence of smaller and/or larger erythrocytes (36, 38).

In medical hematology, a study in 1975 did not show any strong relationship between anisocytosis graded on blood smears and analyzer-detected erythrocyte size variability, as only nine out of 40 patients with increased erythrocyte volume variability were identified as slightly or moderately anisocytic on blood smears (39). However, Simel et al. (40) found that semiquantitative grading of anisocytosis correlated with RDW (Kendal tau-b 0.22-0.40) and, when using intraclass correlation coefficient analysis, inter- and intra-observer agreement ranged between 0.336–0.553, and 0.400–0.642, respectively. Kumar et al. (41) also found high inter-observer agreement for semiquantitative grading of anisocytosis on blood smears (concordance correlation coefficient 0.73–0.76) but non-significant correlation between semi-quantitative grading of anisocytosis and RDW (concordance correlation coefficient 0.011). Finally, Jen et al. (42) found no intra-observer agreement (kappa coefficient 0.19) and no inter-observer agreement (kappa coefficient ranging between 0.16–0.19) in anemic patients.

In veterinary medicine, studies aiming to assess observer agreement and correlation between RDW and visual assessments of anisocytosis are rare. One study by Kumiega et al. (16) found that RDW results were within the reference range for dogs with degenerative mitral valvular disease while manual microscopic analysis of the blood smears revealed the presence of anisocytosis, although no direct correlation was reported. de Souza et al. (9) communicated that, in cases where anisocytosis was observed in blood smears in anemic dogs, these dogs also had higher RDW values and concluded that RDW values were more accurate than microscopic observations to detect anisocytosis.

As few veterinary studies have compared visual assessments of anisocytosis and RDW, the first research aim of this study was to evaluate whether anisocytosis assessed by conventional microscopy correlated to RDW. The second research aim was to evaluate observer variability in assessing the degree of anisocytosis using conventional microscopy. The research objectives were to assess the correlation of the semi-quantitative grading of anisocytosis by means of conventional microscopy of canine blood smears to RDW, and to assess intra- and inter-observer variation in assessing the degree of anisocytosis using conventional microscopy. The underlying rationale was that if semi-quantitative grading of anisocytosis correlated with RDW, the more precise of the two measures should be used for clinical decision-making.

Of the 100 blood samples included in the study, none were classified as unacceptable by any of the observers. All observers noted that some (fewer than 15) blood smears were difficult to assess because of concomitant moderate to marked poikilocytosis or rouleaux formation. Because samples with poikilocytosis or rouleaux formation are likely to occur in daily practice, it was decided not to exclude these from the analysis to obtain a realistic as possible estimation of the observer variability in everyday practice. The underlying raw data set is available from the corresponding author.

The 100 RDW values ranged from 10.6 to 22.1% with 16 values (16%) being above the upper reference limit for RDW of 14.3%. Overall, the three observers classified 51–61% as grade 1, 20–40% as grade 2, 5–9% as grade 3, and 0–8% as grade 4 (Figure 1). Of the 16 samples with RDW above 14.3%, the three observers classified between 0 and 2 cases as Grade 1, 4–7 cases as Grade 2, 4–8 cases as Grade 3, and 0–8 cases as Grade 4.

The upper reference limit for RDW is 14.3% at our laboratory. Using this limit, most observations by the three observers were anisocytosis grade 1 and 2 for RDW below 14.3%, and anisocytosis grade 2, 3, and 4 for RDW above 14.3% (Figures 2, 3).

Anisocytosis graded on blood smears correlated moderately but significantly in this study to RDW with Kendahl-tau-b ranging between 0.37 and 0.51 (p < 0.0001) (Figure 4).

Intra-observer agreement as assessed by quadratically weighted kappa was weak to moderate, with resulting kappa-coefficients being 0.58, 0.68, and 0.75.

Inter-observer agreement as assessed by free-marginal kappa was weak. On the first occasion, the three observers agreed on classification on 39 occasions. Inter-observer variability expressed as kappa-value was 0.44 (0.36–0.53). On the second occasion, the three observers agreed on 39 occasions and inter-observer variability was 0.44 (0.35–0.53).

RDW values are available in many veterinary laboratories, as is semiquantitative visual grading of anisocytosis. The two methods assess the same phenomenon, i.e., variability in erythrocyte size. In line with findings in medical hematology (40), the two methods also correlated in this study (Figure 2).

Visual microscopy to assess the degree of anisocytosis is associated with observer variability. In medical hematology, intra- and inter-observer agreement are moderate at best (39–42). This was also the case in the present study, where weak to moderate intra-observer agreement and weak inter-observer agreement were found. Several reasons for the weak to moderate observer agreement exist. For the semiquantitative assessment of anisocytosis, Gulati's method was applied in the present study to reduce additional errors from defining and assessing e.g., microcytes. However, in Gulati's method there are no solid borders between each category. For example, grade 2 includes RBC with a difference 2–3× and grade 3 includes a difference at 3–4×, thus if the degree of anisocytosis is borderline between the two grades, a variation is possible when assigning to one grade only.

Another reason for the weak to moderate observer variability is difficulties in assessing the degree of anisocytosis when poikilocytosis is present concomitantly. In this case, the Advia 2120i analyzer performs isovolumetric sphering of erythrocytes prior to analysis (33, 62, 63) and thus reduces poikilocytosis from RDW measurement, whereas the visual observer may experience difficulties in assessing the degree of anisocytosis. Also, marked rouleaux formation on blood smears may force visual assessments to be conducted nearer the feathered edge of the smear, potentially outside the monolayer. Although none of the observers excluded any samples as unacceptable, some samples that had either moderate to marked poikilocytosis or rouleaux formation were included to reflect everyday practice; this potentially could have interfered with observer variability, resulting in observer variability to be weak to moderate. Also, having each observer define the monolayer on each slide on each occasion made it very unlikely that the same field on each slide was evaluated by each observer on each occasion. This in turn adds to the magnitude of variability within and between observers. In contrast to the weak to moderate observer agreement found in this study, imprecision of RDW measurements on the Advia 2120i is below 2.5%. Seemingly, RDW measurement is more precise than visual assessment and thus, for clinical decision-making RDW values should be used. However, visual assessment of a blood smear remains a vital part of the CBC e.g., to validate the RDW value and other RBC features such as dimorphic RBC populations (64).

In the present study, only samples from dogs were included, and weak to moderate intra-observer agreement and weak inter-observer agreement were found. As intra- and inter-observer agreement in medical hematology are moderate at best (37–40), this could indicate that observer agreement is also weak to moderate in other animal species as well, thereby potentially also indicating that RDW values should be preferred over visual assessments of anisocytosis for clinical decision-making in other domestic animal species.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal studies were approved by the Ethical Committee at the Department of Veterinary Clinical Sciences, SUND, University of Copenhagen approved the project (Approval No. 2023-07), and all samples were anonymized. The studies were conducted in accordance with the local legislation and institutional requirements. Written informed consent was not obtained from the owners for the participation of their animals in this study because the study was approved by the Ethical Committee and no study on animals were conducted. Samples were already submitted and collected for diagnostic and/or therapeutic reasons so no additional samples were required. The samples were also anonymized and the study did not focus on the dogs included.

AJ: Conceptualization, Investigation, Writing—original draft, Writing—review and editing. AK: Investigation, Writing—original draft, Writing—review, and editing. LN: Investigation, Writing—original draft, Writing—review, and editing.