Platelet clump assessment using the Cellavision peripherical blood application – do we need manual microscopy?

Introduction

Platelet count is one of the key parameters in a complete blood count, and its rapid reporting can be crucial for the urgent management of patients with thrombocytopenia. One of the most common analytical pitfalls in platelet counting is the presence of platelet clumps in the blood sample [1], which can lead to a falsely low platelet count as measured by the hematology analyzer and may result in significant medical consequence [2].

Platelet clumping is caused by the exposure of a cryptic antigen revealed when calcium is chelated by ethylenediaminetetraacetate (EDTA) in the blood collection tube [3]. Importantly, this phenomenon is purely in vitro and reflects the action of non-pathological autoantibodies [4]. Although this condition is not indicative of an underlying pathological autoimmune disorder, it can lead to a significant diagnostic error – a falsely decreased platelet count. In the general population, the frequency of platelet clumping is estimated to range from 0.1 to 0.2 %, but this can rise to as high as 2 % among hospitalized patients, where it is more frequently encountered due to complex underlying conditions [5], [6], [7].

To address this, the French expert group, Groupe Francophone d’Hématologie Cellulaire (GFHC), has established guidelines for detecting platelet clumps under specific clinical conditions. These include cases of unexpected thrombocytopenia below 100 × 10⁹/L in adults or 150 × 10⁹/L in children, a decrease in platelet count greater than 50 % in adults or 30 % in children when compared with the previous result, the presence of specific automated analyzer alarms that suggest platelet clumps, or a known history of platelet clumping in a previous sample from the same patient. By convention, a platelet clump is defined as the presence of at least five aggregated platelets on the blood smear [8].

The manual search for platelet clumps on a blood smear is labor-intensive and requires significant technician time to ensure accurate control. However, in recent years, automated digital microscopes with image acquisition capabilities have been increasingly deployed in clinical laboratories. These systems are routinely used for white blood cell differentials and morphological analysis of various cell lines, offering significant time savings, improved traceability, and standardized practices [9], [10], [11]. Despite these advantages, concerns remain regarding their ability to reliably detect platelet clumps on blood smears, which may still necessitate manual review.

The objectives of this study were to evaluate the performance of automated platelet clump detection using the Cellavision system and to identify specific situations in which manual verification of the blood smear may not be required.

Materials and methods

Results

Frequency of platelet clumps in the general population

We first evaluated the frequency of platelet clumps in nearly one million hemograms at our laboratory, stratified by age. The overall frequency of platelet clumps was 0.29 %. A significant frequency was observed in children under 10 years old, with a notable increase in frequency with advancing age, reaching over 0.5 % in individuals aged 80 and older (Figure 1).

Figure 1:

Frequency of platelet clumps stratified by age.

Detection and characteristics of platelet clumps

We observed that the alarms for suspected platelet clumps provided by the SYSMEX analyzers individually had low performance (Table 1). Some alarms showed better sensitivity depending on the initial platelet count measured by the hematology analyzer.

Table 1:

Frequency of platelet clumps by platelet count and alarm/slide review rules (Presence of platelet clumps/number of smears with alarm). The data in this Table correspond to the samples for which the presence of an alarm or slide review rules made it possible to detect the presence of platelet clumps on the blood smear.

Alarm or slide review rules	Platelet count
	<50 × 109/L	50–100 × 109/L	100–150 × 109/L	>150 × 109/L
Alarm: Plt clumps?	6/15 (40 %)	6/56 (11 %)	13/85 (15 %)	56/90 (62 %)
Alarm: Plt interference/Abn distribution	8/15 (53 %)	8/56 (14 %)	13/85 (15 %)	15/90 (17 %)
Slide review rule: Plt low unknown	2/15 (13 %)	30/56 (53 %)	34/85 (40 %)	0/90 (0 %)a
Slide review rule: Delta checkb	2/15 (13 %)	26/56 (46 %)	11/85 (13 %)	2/90 (2 %)
Slide review rule: Prior platelet clumps	4/15 (27 %)	23/56 (41 %)	26/85 (31 %)	21/90 (23 %)
Number of {Alarm+Slide reviews} for each sample:
1	11	27	72	86
2	2	22	13	4
3	2	4	0	0
4	0	3	0	0
5	0	0	0	0

1. ^aBy definition, this rule does not apply to cases with an initial platelet count >150 × 10⁹/L. ^bDecrease in platelet count by more than 50 % for an adult and >30 % for a child compared with previous result. Abn, abnormal; Plt, platelet.

Across all smears, platelet clumps were found in 69 % of cases in the readable area, where the leukocyte count is performed (Figure 2). Clumps were located on the lateral edges in half of the cases and at the feather edge in nearly 90 % of the cases (since clumps can be found in different areas, the total exceeds 100 %). Only 40 % of the cases had clumps in all areas (feather edge, lateral edges, and readable area). We also noted variability in the frequency of platelet clumps across smear areas based on the initial platelet count from the analyzer. When the initial count was very low, clumps were dispersed across the smear (likely due to their large size and number), whereas at higher platelet counts, clumps were primarily found at the feather edge (Figure 3, p=0.001 for 100–150 × 10⁹/L and p=0.012 for >150 × 10⁹/L).

Figure 2:

Relative distribution of platelet clumps on different parts of the blood smears. The presence of platelet clumps was detected by manual microscope observation on 246 samples, and the localization of the clumps was recorded as follows: presence on one edge (lateral edge 1), the other edge of the smear (lateral edge 2), the reading area (corresponding to the one chosen for the white blood cell differential or morphological observation of red blood cells), and the feather edge. Each of these areas was observed in at least five representative fields on each smear. The total adds up to more than 100 % because clumps can be found in different areas of the smear.

Figure 3:

Detection of platelet clumps based on platelet count when the smear is observed at the feather edge, the readable area of the lateral edges.*p=0.001; **p=0.012.

To further analyze the distribution and characteristics of these clumps throughout the smear, we quantified the number of clumps and the number of platelets within these clumps per 1 mm section along the length of the smear (Figure 4). A nadir in the number of clumps was observed halfway through the smear, with smaller clump sizes in the same area.

Figure 4:

Location and size of platelet clumps. We quantified the number of clumps and the number of platelets comprising these clumps in 1 mm sections. The data presented represent the average of at least five fields analyzed at each distance from the blood drop deposition, with these measurements being repeated and averaged over five different blood smears exhibiting platelet clumps.

Detection of platelet clumps on Cellavision

We reviewed the 246 blood smears using the Cellavision DC-1 or DI-60 automated microscope, checking for the presence of platelet clumps in both the leukocyte and red blood cell/platelet image displays. Clumps were found in the leukocyte tab in just over 60 % of cases and in the red blood cell tab in over 80 % of cases. Overall, platelet clumps were detected in 93 % of cases across these two tabs (Figure 5).

Figure 5:

Sensitivity of detection of platelet clumps detection on platelet count and the available fields in the Cellavision software. RBC, red blood cell; Plt, platelets.

Interestingly, the performance of platelet clump detection in the leukocyte tab decreased as the initial platelet count from the analyzer dropped, while it increased for the red blood cell observation field. Combining these results, we achieved excellent sensitivity for initial counts below 100 × 10⁹/L or above 150 × 10⁹/L. Additionally, we found that clumps were less frequently detected on Cellavision in cases with prior history of platelet clumps in the same patient (Figure 6).

Figure 6:

Sensitivity for the detection of platelet clumps with Cellavision based on the reasons for investigating platelet clumps. Plt, platelets.

Among the 246 cases in our study, performing a manual review of blood smears only in cases where the initial platelet count on the analyzer was below 50 × 10⁹/L and/or when there was a history of platelet clumps in the same patient led to the detection of 97 % of platelet clumps using Cellavision images. This approach reduced the need for manual smear reviews by three-quarters, significantly decreasing the required technician time and turnaround time for reporting results to the physician.

Discussion

The present study provides insights into the frequency, detection, and characteristics of platelet clumps in a large cohort of nearly one million hemograms. The overall incidence of platelet clumps was 0.29 %, with a marked increase in frequency in both pediatric and elderly populations. This rate is slightly higher than that observed in other studies [12], 13], but the increase in incidence with age is consistent with a recent report [14]. This bimodal distribution suggests that platelet clumping may be influenced by age-related physiological and pathological factors. Platelet satellitism, a related phenomenon characterized by platelets adhering to neutrophils [15], is even rarer (less than 0.01 %) and was not observed in our study, though it would have been easily detected through digital imaging.

Our findings highlight that alarms from automated hematology analyzers, while helpful, may not provide sufficient sensitivity on their own to reliably detect platelet clumps, as described before [16].

Furthermore, our detailed examination of clump distribution within the blood smear reveals important patterns. Platelet clumps were more frequently found at the feather edge and lateral edges of the smear, particularly when the initial platelet count was low. This observation suggests that larger and more numerous clumps tend to be pushed to the periphery of the smear during preparation. We also quantified for the first time the variations in the number and size of platelet clumps along the blood smear. We demonstrated that these clumps are the rarest and smallest in the Cellavision reading area, which may explain the device’s limited performance. These characteristics explain the limitations of image acquisition by this version of Cellavision that focus primarily on the central zone of the smear [17].

Nevertheless, the use of automated digital microscopy proved to be a valuable tool for detecting platelet clumps, with high sensitivity (93 %) when combining the leukocyte and red blood cell observation fields. However, we noted that detection performance varied according to the initial platelet count and the rationale for the investigation of clumps. Specifically, sensitivity decreased for lower counts in the leukocyte tab but increased in the red blood cell observation field, indicating that clumps are more likely to be visualized in the latter at low platelet counts. This suggests that observing platelet clumps using digital microscope must be carried out using all the fields available in the Cellavision system.

The platelet clump detection performance observed in our study appears to be better than those reported elsewhere [13], 17], 18]. This could be explained by our systematic reclassification of platelet clump images in the leukocyte field, and the variety of platelet counts (with a much larger number of cases in our study) for which clumps were investigated.

Compared to conventional microscopy platelet aggregates cannot be detected by 100 % diagnostic sensitivity using Cellavision digital microscopy, because aggregates might be preferentially smeared to the border arears where digital microscopy does not check. The detection performance of platelet clumps with automated microscopes could be further improved by expanding image acquisition to more fields and targeting the feather edge of the smear.

Based on the current data, we have demonstrated that reviewing only cases with a prior history of platelet clumps or an automated platelet count below 50 × 10⁹/L (with no similar history) using a traditional microscope allows for the detection of 97 % of platelet clumps, while reducing the number of manually reviewed slides by fourfold. This strategy appears to be highly relevant for maximizing the turnaround time for reporting to physicians, while ensuring the expected quality of results.

In conclusion, while automated detection systems are a valuable asset, they require supplementation with manual review to ensure accurate detection of platelet clumps in a limited number of situations. This study underscore the need for a combined approach, especially in cases of extreme platelet counts (due to the immediate medical impact of a severely reduced platelet count) or when prior clumps have been detected. Further studies are warranted to optimize the integration of digital microscopy with integration of the feather edge to enhance diagnostic accuracy.