|Year : 2013 | Volume
| Issue : 4 | Page : 137-144
An analytical study on peripheral blood smears in anemia and correlation with cell counter generated red cell parameters
Ashutosh Kumar, Rashmi Kushwaha, Chani Gupta, US Singh
Department of Pathology, King George's Medical University, Lucknow, Uttar Pradesh, India
|Date of Web Publication||26-Feb-2014|
Flat No. 504, T.G Hostel Khadra, Lucknow - 226 005, Uttar Pradesh
Source of Support: None, Conflict of Interest: None
Context: Manual examination of peripheral blood smear in diagnosis of anemia has taken a backseat with the advent of automated counters. Though a lot of studies have been done to assess the efficacy and significance of red blood cell parameters in different hematological conditions fewer efforts have been made to standardize the visual examination of peripheral blood smears for diagnosing anemias.
Aims: Standardization and grading of abnormal red cell morphology in peripheral blood smear and counter based red cell indices in cases of anemia of various etiologies.
Settings and Design: Cross-sectional study of one year duration conducted in the Hematology laboratory, in a tertiary care hospital in North India.
Materials and Methods: In 60 anemic patients, automated counts and peripheral blood smear were prepared and evaluated by three observers, according to a red cell morphology grading guide.
Statistical Analysis Used: ANOVA, Tukey post hoc test were used.
Results: Objective grading of peripheral blood smears in cases of anemia have a good inter observer correlation and hence have reduced subjective variation. Manual parameters like microcytosis, macrocytosis and hypochromia expressed as a percentage, have shown significant correlation, with their corresponding automated parameters, and the regression model so generated may provide a novel way for quality control of automated counters, if calculated for different models.
Conclusions: Even in the age of molecular analysis, the blood smear remains an important diagnostic tool and sophisticated modern investigations of hematologic disorders should be interpreted in the light of peripheral blood features as well as the clinical context.
Keywords: Anisocytosis, automated parameters, microcytosis, peripheral blood smear
|How to cite this article:|
Kumar A, Kushwaha R, Gupta C, Singh U S. An analytical study on peripheral blood smears in anemia and correlation with cell counter generated red cell parameters. J Appl Hematol 2013;4:137-44
|How to cite this URL:|
Kumar A, Kushwaha R, Gupta C, Singh U S. An analytical study on peripheral blood smears in anemia and correlation with cell counter generated red cell parameters. J Appl Hematol [serial online] 2013 [cited 2020 Aug 8];4:137-44. Available from: http://www.jahjournal.org/text.asp?2013/4/4/137/127896
| Introduction|| |
Peripheral blood examination has been a window for hematological ongoings since decades. Analyzing blood films routinely has facilitated interpretation of various hematological disorders and has been a major diagnostic tool especially for etiopathological work up of anemias.
Along the years there have been studies from time to time assessing the utility and accuracy of most of the automated generated parameters in general as well as with respect to specific types of anemia, but fewer efforts have been made in the direction of devising methods to objectively assess peripheral blood smear manually, and its practical utility in leading to a diagnosis.
This study is an attempt to standardize further few automated red cell parameters, and also objective grading of RBC morphology on peripheral smear and interpreting its utility in indicating a diagnosis. We hope to get a significant correlation between inter-observer assessments, get a degree of correlation between automated and manual parameters, and hope to assess possible causes of common discrepancies and establish utility of manual grading in making a diagnosis.
| Material and Methods|| |
This was a cross-sectional study of one year duration conducted in a tertiary care hospital in North India. It was carried out with the objectives of standardization and grading of abnormal red cell morphology in peripheral blood smear, counter based red cell indices in cases of anemia of various etiologies and comparative evaluation of peripheral blood smear examination and automated red cell indices including RDW, MCV, MCHC, MCHC, fragmented RBC and others, for anisopoikilocytosis in diagnosing anaemia.
All patients with non neoplastic anemias were screened and selected on the basis of clinical evaluation and hemoglobin values.
Hematological studies included complete blood count by automation and comparing it to the general blood picture.
Sixty patients identified by hemoglobin values lower than 9 g/dl, with chief complaints, along with clinical correlation were identified.
All peripheral blood films were prepared manually and were stained by a single trained person to minimize variation in smear spreading and staining due to interpersonal differences in technique. Two practicing hematologists and one second year medical junior resident examined each peripheral blood film independently. Each peripheral blood film was visually graded for RBC morphology following guidelines given in Blood cell morphology Grading guide written by Gene Gulati and published in 2009.  [Table 1].
The EDTA sample of each patient was run through automated counters.
Continuous data were summarized as Mean ± SD, while discrete (categorical) data was summarized in percent. The continuous groups were compared by one-way analysis of variance (ANOVA) and the significance of mean difference between the groups was done by Tukey post hoc test after ascertaining the normality by Shapiro-Wilk test and the homogeneity of variance by Levene's test. Groups were also compared by one way ANOVA followed by Tukey's post hoc test. The categorical variables were compared by Chi-square (χ2 ) test. Concordance correlation coefficient was used to asses inter observer reliability and reproducibility. A two-sided (α =2) P < 0.05 was considered statistically significant. All analysis were performed on STATISTICA (window version 6.0).
| Observations and Results|| |
The present study evaluates peripheral blood smears in anemia and correlates it with cell counter generated red cell parameters of 60 anemic patients.
The basic characteristics of all anemic patients are summarized in [Table 2]. The age of all anemic patients ranged from -60 yrs with mean (± SD) 20.47 ± 17.12 yrs and median 15 yrs. Most of the anemic patients were below 10 yrs (31.7%). Out of the total patients, 33.3% patients were diagnosed with Iron deficiency Anemia (IDA), 31.7% with Megaloblastic anemia (MA), 21.7% with Hemolytic anemia (HA) and 13.3% belong to other anemic group.
Hemolytic anemias were grouped together under a single category, with majority being comprised of Thalassemia. Thalassemia Intermedia and Thalassemia Major constituted 4 cases each, Thalessemia minor 3 cases, hereditary spheroytosis and thrombotic thrombocytopenic purpura 1 case each.
Manually Graded Red Cell Parameters
Reproducibility- Inter-rater agreement
The reproducibility (inter-rater agreement) of grades given by three independent observers on manual reporting was assessed using concordance correlation coefficient (r value) method and summarized in [Table 3]. Concordance correlation revealed significantly (P < 0.01 or P < 0.001) high inter-rater agreement on all manual parameters with highest being for Tear drop cells and least for Elliptocytes. In other words, manual reporting had a high reproducibility for assessing cell counter generated red cell parameters on peripheral blood smears in anemia.
|Table 3: Summary of inter-rater agreement (n=60) of three observers on manual reporting on different parameters|
Click here to view
| Reliability-Correlation of manual reporting with Automated parameters|| |
The correlation (reliability) of manual reporting (in grades) with automated parameters (MCV, MCH and MCHC) was assessed by Pearson correlation coefficient (r value) method and summarized in [Table 4]. Pearson correlation revealed not a significant correlation between Microcytosis and MCV (r = 0.004, P > 0.05), Hypochromia and MCH (r = 0.026, P > 0.05), and Hypochromia and MCHC (r = 0.010, P > 0.05), Macrocytosis and MCV (r = 0.003, P > 0.05). In other words, manual reporting in terms of grades did not show statistically significant reliability for assessing cell counter generated red cell parameters in peripheral blood smears in anemia.
The correlation (reliability) of manual reporting (final grades in %) with automated parameters (MCV, MCH and MCHC) was assessed by Pearson correlation coefficient (r value) method and summarized in [Table 5]. Pearson correlation revealed a significant and negative (inverse) correlation between Microcytosis and MCV (r = −0.51, P < 0.001), Hypochromia and MCH (r = −0.56, P < 0.001), and Hypochromia and MCHC (r = −0.58, P < 0.001) while significant and positive (direct) correlation between Macrocytosis and MCV (r = 0.54, P < 0.001). In other words, manual reporting had a high reliability for assessing cell counter generated red cell parameters on peripheral blood smears in anemia.
|Table 5: Correlation (n=60) of manual reporting parameters with automated parameters|
Click here to view
| Reliability-Correlation of manual reporting with Diagnosis|| |
The correlation (reliability) of manual reporting (in grades) with Diagnosis (1: IDA, 2: MA, 3: HA, 4: Others) were assessed by Pearson correlation coefficient (r value) method [Table 6]. Pearson correlation revealed there was no significant correlation between Microcytosis, Hypochromia and Anisocytosis. However, Macrocytosis did show borderline significant association with the diagnosis (r = 0.25, P < 0.05).
The correlation (reliability) of manual reporting (final grades in %) with diagnosis (1: IDA, 2: MA, 3: HA, 4: Others) was assessed by Pearson correlation coefficient (r value) method [Table 7]. Pearson correlation revealed a significant and negative correlation of Microcytosis (r = −0.41, P < 0.001) and Hypochromia (r = −0.26, P < 0.05) with the diagnosis while Poikilocytosis showed significant and positive correlation with the diagnosis (r = 0.28, P < 0.05). In other words, manual reporting had a high reliability for assessing diagnosis red cell parameters on peripheral blood smears in anemia.
|Table 7: Correlation (n=60) of manual reporting parameters (%) with diagnosis|
Click here to view
The mean percentage of Target cells (19 cases), Tear drop cells (23 cases), elliptocytes (14 cases) and Ovalocytes (17 cases) in various anemias is depicted in [Figure 1], [Figure 2], [Figure 3], [Figure 4] respectively.
| Discussion|| |
The mean hemoglobin as per the different diagnostic groups, showed minor variations with IDA group having a mean hemoglobin of 5.85 g/dl, MA 5.25 g/dl, HA 6.88 g/dl and others 5.03 g/dl. On applying ANOVA, the difference was found to be statistically insignificant, indicating that severity of anemia is not a function of the type of anemia but instead depends on the severity which can be seen at any grade.
For all the parameters, a mean value was calculated with respect to the different diagnostic groups. MCV showed a mean of 68.83 fl for IDA, 99.4fl for MA, 78.85 fl for HA and 87.5fl for others. On applying ANOVA, the P value was found to be statistically significant hence consolidating the fact that MCV was an expected reliable parameter, in particular for differentiating MA from IDA group, and MA from HA group which in our study were comprised predominantly of β Thalassemias. Although in our study there was a difference in the mean MCV of IDA and HA, it was not found to be statistically significant, probably because our HA group had predominantly β Thalassemias, and also a case of Hereditary Spherocytosis, both of which have a microcytic picture. The finding was in concordance with a study in 2002 by Melo MR et al., on the use of erythrocyte (RBC) indices in the differential diagnosis of microcytic anemias to evaluate prospectively RBC indices as a diagnostic tool. 
A similar trend was seen with MCH. Mean and standard deviation for each group were calculated and it showed a significant test result on applying ANOVA. As expected from previous studies, MCH showed a significant difference between MA group, and IDA group as well as between MA group and HA. The mean MCH for IDA and HA was not found to be statistically significant, probably because reduced hemoglobin is a feature of both IDA as well as Thalassemias.
MCHC though an indicator of reduced hemoglobin, did not give as consistent a result as MCH, though overall it was just significant on ANOVA, was helpful in differentiating mainly MA from IDA. The above pattern indicates that though the MCH was decreased in our cases of HA, the MCHC was not markedly reduced as compared to IDA.
Along the years there have been many studies on RDW still it has not been standardized. In our study also ANOVA gave a statistically insignificant result, it being significant only in differentiating IDA from HA. In 2008 Mauro Buttarello et al., concluded that, a continued effort still needed to be made for some parameters like RDW, IRF, MCVr, and MPV for which results provided were still too different when produced by different analyzers. 
It has been a common dictum that increased RDW is one of the first signs of identifying IDA and in some studies has also been helpful in differentiating Thalassemia minor from IDA. In our study also we had three cases of Thalassemia Minor, too less to form a separate group, so we had them under HA. We calculated the mean RDW for each type of β Thalassemia. The mean RDW for Thalassemia minor was 14.5 as compared to that for IDA which was 16.23 years. Ours was a very small sample of Thalassemia minor, but in other studies RDW has not proven its worth, as in a study in 2010, Ferrara et al., evaluated the reliability of red blood cell indices and formulas to discriminate between β Thalassemia trait and iron deficiency in children. They concluded that none of RBC indices or formulas appeared reliable to discriminate between the two. 
One of our aims was to assess the reliability and applicability of objectively grading the peripheral blood smears. For this we had graded all the smears according to blood cell morphology grading guide by G. Gulati. All the 60 smears were observed by our three observers and grades were given to each parameter. 
One of the statistical aims of our study was to assess interobserver correlation, to assess the reproducibility of the method and to assess the degree of subjective variation if such a system of manual grading to assess RBC morphology objectively was brought into practice; i.e., we wished to see how much the subjective difference in opinion will hamper this method.
A concordance correlation coefficient was applied, comparing the observations of each observer from the other two. Correlation Coefficient for parameters anisocytosis, poikilocytosis microcytosis, macrocytosis, hypochromia, target cells, tear drop calls schistocytes, spherocytes, and ovalocytes showed statistically significant correlation for each observer with the other two [Table 3]. In case of Elliptocytosis, a lesser degree of concordance was seen among observers, still it was also statistically significant. The probable reason could be subjective variation in calling a cell elliptocyte or ovalocyte by different observers.
One of the most important reasons for variability of whatever degree could be the field chosen by the different observers. Our observers counted a minimum of 200 cells; such an error could probably be minimized by counting cells in more number of fields and counting more number of cells, thereby further improving the inter-observer correlation.
Our next step was to assess how far the manual grades could be correlated with the parameters generated on automated cell counters. For doing this we first took a median grade of four manually graded parameters, in each case, for comparison with their respective counterparts on automated generated red cell indices. For those parameters, in a case where a median value could not be reached, the grade given by the most experienced observer was considered as final. When we applied concordant correlation coefficient for anisocytosis with RDW, microcytosis with MCV, macrocytosis with MCV, hypochromia with MCH and hypochromia with MCHC, the results were found to be statistically insignificant [Table 4]. In contrast, in a study by David Simel in 1988 on visual inspection of blood films as against automated analysis of RDW in which they did semi quantitative analysis of anisocytosis using the following ordinal scale: 0, no anisocytosis; 1+, mild anisocytosis; 2+, moderate anisocytosis; 3+, prominent anisocytosis; and 4+, marked anisocytosis it was found that semi quantitative assessment of anisocytosis correlated fairly well with automated counter generated values of RDW. 
The probable reason for our result could be that the automated parameters were a continuous data and had a much wider range than the grades which were discrete and had a range from 1 to 4 only. Moreover the grading system we followed had grades for percent of the cell type and the size, in case of microcytosis and macrocytosis, and central pale area in case of hypochromia. So, when percent denoted a different grade and size denoted a different grade, we chose percent as the dominating factor. Probably this inconsistency in choosing the final grade, and ignoring the other parameter, could have been one of the causes, for statistically insignificant results.
Therefore we thought that instead of comparing the automated parameters with overall grade, maybe we should take the net percentage given in each case for each parameter, according to the final grade and the grade given by the most experienced observer.
On calculating correlation coefficient for each manual parameter in percent and its corresponding automated parameter, Pearson correlation revealed a significant and negative (inverse) correlation between microcytosis and MCV (r = −0.51, P < 0.001), hypochromia and MCH (r = −0.56, P < 0.001), and hypochromia and MCHC (r = −0.58, P < 0.001) while significant and positive (direct) correlation between macrocytosis and MCV (r = 0.54, P < 0.001) [Table 5].
The regression models were as follows:
Microcytosis and MCV: MCV = 109.8 − 0.710 (% Microcytosis); [Figure 5].
|Figure 5: Scatter plot showing correlation of microcytosis % with MCV. The regression model: MCV = −0.710 (% microcytosis) + 109.8.99 Correlation coefficient r = 0.51|
Click here to view
Macrocytosis and MCV: MCV = 71.6 + 0.615 (% Macrocytosis); [Figure 6].
|Figure 6: Scatter plot showing correlation of macrocytosis % with MCV. The regression model: MCV = 0.615 (% macrocytosis) + 71.76 Correlation coefficient r = 0.54|
Click here to view
Hypochromia and MCH: MCH = 30.19 − 0.142 (% Hypochromia); [Figure 7].
|Figure 7: Scatter plot showing correlation of hypochromia % with MCH. The regression model: MCH = −0.142 (% hypochromia) + 30.99 Correlation coefficient r = 0.56|
Click here to view
Hypochromia and MCHC = 32.19 − 0.074 (% Hypochromia); [Figure 8].
|Figure 8: Scatter plot showing correlation of hypochromia % with MCHC. The regression model: MCH = −0.074 (% hypochromia) + 32.19 Correlation coefficient r = 0.58|
Click here to view
In other words, manual reporting in terms of percentage had a high reliability for assessing cell counter generated red cell parameters on peripheral blood smears in anaemia. The regression models so generated could open doors for a novel way of quality control of automated counters through manual assessment.
After seeing that the method had reproducibility as well as good correlation with the automated counts, our next aim was to see if these manual grades correlated with the diagnosis. For this, Pearson correlation coefficient was applied first for final grades with the diagnosis. Though a level of correlation was reached it was found to be statistically insignificant [Table 6].
Again as in the case of correlation with automated parameters, correlation of diagnostic groups with percentage given to each parameter was also done. The correlation (reliability) of manual reporting in percent with Diagnosis (1: IDA, 2: MA, 3: HA, 4: Others) were assessed by Pearson correlation coefficient (r value) [Table 7]. Pearson correlation revealed a significant and negative correlation of Microcytosis (r = −0.41, P < 0.001) and Hypochromia (r = −0.26, P < 0.05) with the Diagnosis while Poikilocytosis showed significant and positive correlation with the Diagnosis (r = 0.28, P < 0.05). In other words, objective manual reporting in percentage had a high reliability for assessing diagnosis on peripheral blood smears in anemia. However, Macrocytosis did not show a significant association with the diagnosis (r = 0.21, P > 0.05) but had a positive and borderline significance. In our study, MCV showed a significant association in differentiating MA from both IDA as well as HA, and also showed a significant concordant correlation coefficient (0.54) with manually graded macrocytosis in percent. However, the manual grading in percent was found insignificant in association with the diagnosis, in concordance with the study in 1978, by RJL Davidson and P Hamilton.  In contrast, in 2006 in an article by Florence Aslina et al., on MA and other causes of macrocytosis it was stated that the peripheral blood smear was more sensitive than RBC indices for identifying early macrocytic changes because the MCV represented the mean of the distribution curve and was insensitive to the presence of small numbers of macrocytes. 
Some parameters which we assessed manually did not have a corresponding parameter on automated counts. For these parameters we calculated the mean percentage, and compared them with our final diagnosis.
Of the 60 samples, target cells were reported in 19 cases including cases of Thalassemia Major, Thalassemia Minor, Thalassemia Intermedia, IDA, MA and Anemia of Chronic Disease, thus covering almost the entire spectrum of anemia diagnosis which we had in our study. Next we compared the mean in each diagnosis and found it highest at 32%, for Thalassemia minor, followed by 19% for Thalassemia intermedia, and 6.16%, for Thalassemia major [Figure 1].
Tear drop cells were reported in around 23 cases and gave the maximum mean of 22.67 in patients with Thalassemia major followed by around 6% in Thalassemia intermedia, and about 4.6 in Thalassemia minor, which was only marginally higher than that in IDA and MA. Though tear drop cells are seen in whole range of diagnosis, the number was significantly higher for Thalassemia Major, and partially higher for Thalassemia Intermedia, indicating that besides presence, percentage of tear drop cells and target cells can go a long way in, if not establishing but making a provisional diagnosis of Thalassemia Major, Minor and Intermedia [Figure 2].
Similarly, ovalocytes showed a clear-cut peak for MA with 15.5%, significantly different from those of other diagnostic groups whose mean values lied in the range of 2 to 6% [Figure 4].
In 1993 Wisconsin medical technologists, Pat Garrity and Jeri Walters developed a two tiered system for grading morphology which recognized the fact that some abnormal forms were significant in low numbers (i.e., spherocytes, schistocytes, sickle cells, Howell jolly bodies, etc.) and others were only significant when present in high numbers (i.e., target cells, ovalocytes, microcytes, macrocytes, etc.). They called them "splitters" and "lumpers".  In our findings also, though target cells, tear drop cells and ovalocytes were present in low numbers across all the diagnostic groups, a higher percentage signified Thalassemia types.
As stated above, numerous studies have found red cell indices, inadequate alone to discriminate between different diagnoses.
Barbara J. Bain in her review on the place of peripheral blood smear examination in the age of automation in 2005 had said that even in the age of molecular analysis, the blood smear remains an important diagnostic tool and sophisticated modern investigations of hematologic disorders should be interpreted in the light of peripheral-blood features as well as the clinical context.  To add further, if we could analyze the peripheral blood smears objectively, it well be instrumental in not only reducing interobserver variation, and in making the diagnosis but may also serve as a quality control for automated counters using the regression models we got for different parameters. For the application of these regression formulas on different types of automated counters and their applicability remains to be seen. Further studies with larger sample size, targeting one parameter and automated counter at a time are needed.
Not many studies have been taken up in the past to assess the validity and applicability of manual objective assessment of the peripheral smears for red cell parameters. Our study therefore is an attempt for the same, and to provide a cost effective method to consolidate, though not replace the automated cell counters, and also to provide a proposed method for quality control of automated counters for red cell parameters.
| References|| |
|1.||Gulati G. Blood cell morphology grading guide. Hongkong: American Society for Clinical Pathology; 2009. |
|2.||Melo MR, Purini MC, Cancado RD, Kooro F, Chiattone CS. The use of erythrocyte (RBC) indices in the differential diagnosis of microcytic anemias: Is it an approach to be adopted? Rev Assoc Med Bras 2002;48:222-4. |
|3.||Buttarello M, Plebani M. Automated blood cell counts: State of the art. Am J Clin Pathol 2008;130:104-16. |
|4.||Ferrara M, Capozzi L, Russo R, Bertocco F, Ferrara D. Reliability of red blood cell indices and formulas to discriminate between β thalassemia trait and iron deficiency in children. Hematology 2010;15:112-5. |
|5.||Simel DL, DeLong ER, Feussner JR, Weinberg JB, Crawford J. Erythrocyte anisocytosis. Visual inspection of blood films vs automated analysis of red blood cell distribution width. Arch Intern Med 1988;148:822-4. |
|6.||Davidson R.J.l. Hamilton P.J. high mean red cell volume: Its incidence and significance in routine haematology. J Clin Pathol 1978;31:493-8. |
|7.||Aslinia F, Mazza JJ, Yale SH. Megaloblastic anemia and other causes of macrocytosis. Clin Med Res 2006;4:236-41. |
|8.||Garrity P, Walters G. Standardization of red cell morphology reporting (video), Clinical Laboratory Education. Milwaukee, WI; 1993. |
|9.||Bain BJ. Diagnosis from the blood smear. N Engl J Med 2005;353:498-507. |
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7], [Figure 8]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7]