EDTA-associated pseudothrombocytopenia: definition and real-world occurrence

Introduction

Pseudothrombocytopenia (PTCP) is the finding of spuriously low platelet counts. First described in 1969 [1], this is a well-known phenomenon that is observed regularly in the hematological laboratory while reports on its prevalence range from 0.01–1 % [2], [3], [4], [5]. Most commonly, the presence of PTCP is associated with the use of ethylenediamine tetra-acetic acid (EDTA) as anticoagulant in the tube used for collecting the blood sample, although it has also been observed in tubes using different anticoagulants, such as citrate [5]. The most commonly assumed pathomechanism is the exposure of normally “hidden” epitopes of platelets associated with an EDTA-mediated alteration of the platelet membrane. These epitopes, usually associated with the platelet GPIIb/IIIa receptor, then bind to non-pathogenic autoantibodies, leading to in vitro platelet aggregation [2, 6]. In some cases, instead of aggregation, the spuriously low platelet counts are caused by platelet satellitism, a related phenomenon in which a rosetting of platelets around neutrophils can be observed in vitro. Similarly, this phenomenon is associated with EDTA and likely caused by autoantibodies against the platelet GP IIb/IIIa receptor that react with the neutrophil Fcγ receptor III (FcγRIII) [7]. Furthermore, the artefactual aggregation of platelets in EDTA whole blood (WB) is time dependent with a significant decrease in platelet numbers being observed within the first two hours after sample collection already [2].

Although the underlying causes and mechanisms have been defined relatively well in decades of research, the very definition of PTCP remains surprisingly vague. As stated at the start, it is usually understood to be “finding of spuriously low platelet counts”. The spuriosity of the phenomenon is usually confirmed by the detection of either of the two described in vitro phenomena (aggregation and satellitism) in a peripheral blood smear and/or through the detection of a normal (or at least significantly higher) platelet count using an anticoagulant other than EDTA (mainly citrate or magnesium sulfate (MgSO₄) for the collection of the blood sample. But several important questions remain: How low is low? The name PTCP suggests that it can only be diagnosed, when the spurious result is that of a thrombocytopenia (and therefore below 150/nL by the most common definition). But what about cases where aggregates and/or a significant difference in the platelet count between EDTA WB and citrate WB can be detected but the platelet count in EDTA is within the reference range? Furthermore, what is a “significant difference” between platelet counts in EDTA and citrate WB? There is no common definition of how far apart these two measurements have to be, either in absolute or in relative terms, in order for PTCP to be suspected. This is further complicated by the fact that the duplicated measurement of platelets in both EDTA and citrate WB regularly returns slightly higher results for the former anticoagulant [8], [9], [10], partly due to matrix effects caused by the dilution [10], partly due to temporal instability of platelets in citrate WB [11], [12], [13], [14]. Therefore, when contemplating the difference between platelet counts in EDTA and citrate WB, one has to bear in mind that the norm for this difference is not zero, or, in other terms: even a difference of zero between two platelet counts could be compatible with a very mild form of PTCP. Finally, to date there is no international consensus on what is to be considered a platelet aggregate in the peripheral blood smear (i.e. how many platelets have to be aggregated) and how many of these aggregates have to be found in how many visual fields for aggregates in the blood smear to be reported. There have been some efforts towards a more unified approach in these two latter questions [12, 15, 16], but no generally agreed consensus.

Therefore, in order to better characterize PTCP as it occurs in real-world, we set out to analyze a large data set of concomitant platelet counts from both EDTA and citrate WB. Our aims were to examine the differences between platelet counts in both materials: Where do they occur and how pronounced are they? Which of these measurements were automatically flagged for potential platelet aggregates by the analyzer and which samples revealed aggregates in the peripheral blood smear (when performed)? Is PTCP really a phenomenon that can be readily detected in a given sample, as the vast amount of literature, mostly comprising of case reports of impressively pronounced cases suggests (some examples: [17], [18], [19], [20], and if yes, what are relevant parameters and cutoffs that might reliably distinguish between samples in which PTCP occurs and samples in which it doesn’t)?

Materials and methods

Results

Correlation between platelet counts in EDTA and citrate whole blood

Generally, there was a highly significant correlation of large effect size between platelets in EDTA and platelets in citrate WB (ρ=0.86; p<0.0001; see Figure 1A). However, there are a number of measurements that deviate markedly from the general trend as indicated by the regression line (which itself deviates markedly from the function y=x, along which the measurements would ideally be scattered were it not for the abovementioned trend of lower platelet counts in EDTA WB compared to citrate WB), especially in the lower left-hand corner of the scatter plot: These measurements, with low platelets in EDTA WB and significantly higher platelets in citrate WB are those that would traditionally be classified as possible evidence of EDTA-associated PTCP. However, the scatter plot also shows that this is no distinct set of measurements, but rather within in continuum with other measurements that are more closely dispersed around the regression line. When the information on automated flags is incorporated, it becomes evident that many, but by no means all, of these discordant platelet counts were flagged for the possible presence of platelet aggregates. On the other hand, many measurements very close to the line y=x were also flagged (Figure 1B). Further incorporating the information of the visual examination of the peripheral blood smear, it becomes evident that these latter measurements tend to be those, for which no platelet aggregates were detected visually (Figure 1C). Combining the information on flags and visual examination, one can further see that by no means all measurements with highly discordant platelet counts between both materials for which aggregates were visually found were flagged (Figure 1D). In detail, 30 of 220 samples (13.6 %) in which platelet aggregates were found via light microscopy were not flagged by the analyzer.

Figure 1:

Scatter plots showing the correlation between platelet counts in EDTA whole blood (x-axis) and citrate whole blood (y-axis). Panel (A) shows the general distribution of measurements, including a density plot, panel (B) adds information in automated flags, panel (C) the results of the visual examination for platelet aggregates of peripheral blood smears and panel (D) combines both information.

Absolute difference in platelet counts between materials in relation to platelet counts

Plotting the absolute difference between platelet counts in both materials against the platelet count in citrate WB, a striking subset of measurements can be seen that deviate almost linearly towards a greater negative absolute difference (indicating higher counts in citrate WB) with increasing platelet counts in citrate WB (Figure 2A). These are also those measurements, that would be traditionally classified as potential EDTA-associated PTCP, again in a continuum with other, more “normal” measurements.

Figure 2:

Scatter plots, plotting the absolute difference in platelet counts between EDTA and citrate whole blood (y-axis) against the platelet count in citrate whole blood (x-axis). Panel (A) shows the general distribution of measurements, including a density plot, panel (B) adds information in automated flags, panel (C) the results of the visual examination for platelet aggregates of peripheral blood smears and panel (D) combines both information.

Automated flags and visual examination for platelet aggregates

Again incorporating the information on flags and visual detection of aggregates, it becomes evident that this subset of measurements is indeed where the flagged measurements are primarily clustered (Figure 2B). However, the visual examination reveals that in most cases, which were flagged, but in which the difference between EDTA and citrate WB was small, no platelet aggregates were found visually, which is quite in contrast to cases with a greater difference (Figure 2C, D). Consequently, pairwise Mann-Whitney-U-tests revealed that the (negative) difference is significantly greater in cases in which aggregates were found visually (and even significantly increasing depending on the amount of aggregates detected visually) compared to cases in which no aggregates were found and cases in which no visual examination was performed (with no significant difference between these latter two; Figure 3A).

Figure 3:

(A) Association between the absolute difference in platelet counts between EDTA and citrate whole blood and the semiquantitative results of the visual examination of peripheral blood smears for platelet aggregates (levels of significance: ****=p<0.0001). (B) Decision tree for the correct classification of samples with respect to the occurrence of platelet aggregates. Each node contains the most common classification in the respective (sub-)population of measurements (no=no aggregates; yes=aggregates found), the distribution of measurements with respect to the occurrence of platelet aggregates (the first number in the second line indicating the fractions of measurements without aggregates) and the share of the whole population represented by the individual node. (C, D) scatter plots of the correlation between platelet counts in EDTA whole blood (x-axis) and citrate whole blood (y-axis; panel (C)), and of the absolute difference in platelet counts between EDTA and citrate whole blood (y-axis) against the platelet count in citrate whole blood (x-axis, panel (D)), using only measurements that fulfill the criterion diff<7.5/nL and incorporating information on flags and the visual examination via microscopy.

Seeing as many cases that were flagged could not be confirmed via visual examination and that many cases that cluster with confirmed cases of PTCP in the scatter plots were not flagged (in addition to the small number cases that were not flagged but examined visually anyway and found to contain platelet aggregates), we calculated sensitivity and specificity of the automatic flag (compared to the visual examination as gold standard) and found these to be 86.4 % and 5.8 %, respectively. However, these numbers have to be interpreted with utmost caution: Due to the fact that mostly flagged samples were examined visually, the sensitivity is probably greatly overrated as witnessed by the many cases that were not flagged (and therefore mostly not examined visually) but that cluster with confirmed cases of PTCP, at least a subset of which likely would have been classified as PTCP by visual examination had it been performed. Also, even the poor specificity of 5.8 % is likely artificially increased by the rule that only EDTA WB samples containing fewer than 110/nL platelets were allowed to be flagged.

Discussion

Our results clearly show that EDTA -associated PTCP is a regular occurrence within the hematological laboratory. However, our results point to an important methodological problem in the detection of PTCP: The automated flags, which can be used either for guiding which samples to examine visually for platelet aggregates via light microscopy or for generating a comment for the clinician warning of the possibility of PTCP, appear to be extremely unspecific and at the same time not very sensitive (at least for the equipment used in this study). This is in notable contrast to recent results by another group, which found markedly better values for sensitivity and specificity of the automated flags generated by Sysmex equipment [24]. This contrast is possibly explained by differences in the evaluation of the peripheral blood smears, with the method used by Lunde et al. possibly being more sensitive in the detection of platelet aggregates. The fact that the absolute difference in platelet counts between EDTA and citrate WB did not differ significantly between cases in which no aggregates were found and cases for which no smear was performed combined with the fact that both of these groups of cases differed significantly in this respect to all cases in which platelet aggregates were found int the peripheral blood smear by us suggests that our methods do not overlook a significant number of cases of artefactual platelet aggregation. While ideally every EDTA WB would be examined via light microscopy for platelet aggregates to sensitively and specifically identify samples in which PTCP occurs, this would present an unrealistic work load in the times of automated whole blood counts. Therefore, some other predictor for the occurrence of PTCP is needed. Our data suggest that the difference in platelet counts between EDTA and citrate WB might be a predictor of the occurrence of artifactual in vitro aggregation of platelets, with the threshold of −7.5/nL easily outperforming the automated flags in both sensitivity and specificity in our data set. One obvious, but important limitation of this approach is that the concomitant measurement of platelet counts in both EDTA and citrate WB is needed. However, our findings indicate that this approach would lead to a significant increase in the detection of cases of PTCP while it would also significantly reduce the number of peripheral blood smears performed “in vain” for the detection of platelet aggregates, the large majority of which currently reveal no evidence of PTCP due to the poor specificity of the automated flag.

Still, this theoretical approach in the detection raises at least two important questions which lead us back to the initial issue of the definition of PTCP itself:

First: Is it sensible to define cases as PTCP in which the difference between EDTA and citrate WB is just beyond −7.5/nL and for which platelet aggregates were found via light microscopy? Our data suggest that it may be correct, supporting the assumption that PTCP is a phenomenon that occurs in a wide range of pronunciation that is within a continuum with seemingly normal samples. On the one hand, it is dubitable that from a clinical point of view such cases would be considered significant enough to receive the label of PTCP, on the other hand, it is known that the pronunciation of PTCP in EDTA whole blood is time dependent [2]. Therefore, the presence of platelet aggregates, even when associated with a small absolute difference in platelet counts between EDTA and citrate WB, may be indicative of a predisposition towards PTCP, which may be aggravated when the interval between sample collection and measurement is prolonged for whatever reason, making it all the more important that samples from these patients be examined as timely as possible. It has to mentioned, however, that the threshold of −7.5/nL, which was found by the analysis of a huge data set, might be impractical on a single-specimen basis for another reason: Due to the fact that platelet counts in both EDTA and citrate WB each are subject to both biological and analytical variations [25, 26], which necessarily compound when calculating their difference, a difference of −7.5/nL will almost certainly fall within the range of the combined measurement uncertainty of both determinations. Should a meaningful threshold for the absolute difference in platelet counts between both materials ever be proposed, it should in any case also incorporate these uncertainties.

The second question pertains to the cases that the presented approach overlooks: Cases in which platelet aggregates could be found via light microscopy but no meaningful difference between EDTA and citrate WB. There are two possible explanations for these cases: Either, there are cases in which EDTA-associated platelet aggregation can be found as an in vitro phenomenon that doesn’t influence the platelet count in EDTA WB significantly, or these are cases in which the platelet count in both EDTA and citrate WB is subject to PTCP and therefore similarly spuriously decreased. The latter possibility has certainly been described in the literature [27], but to prove it, one has to perform a platelet count from a MgSO₄ anticoagulated tube, which in our laboratory is unfortunately very rarely ordered (only 134 times from 95 patients, none of which in a pertinent case within the time frame of the data analysis; an obvious limitation of the study). If the first possibility were found to be true, the question of the definition of PTCP would become even more complicated (as well as, to a degree, academic): Can a case be defined as a pseudothrombocytopenia even if the associated thrombocytopenia can be confirmed in its severity (therefore not being “pseudo”) solely on the basis of platelet aggregates in the visual examination of the peripheral blood smear?

Our proposal would be to define PTCP based on two criteria, which would have to be met: 1) Platelet aggregates in the peripheral blood smear and a defined difference between platelet counts in at least two different materials (EDTA WB and citrate WB or EDTA WB and MgSO₄ WB if the first comparison reveals no meaningful difference). However, our data is not suited to propose a meaningful difference in platelet counts between EDTA WB and citrate WB, among other limitations not least because of the bias introduced into the data by the rules guiding the examination of peripheral blood smears that are based on automated flags which, even if the flags perform poorly in terms of specificity and sensitivity, still likely increases the pre-test probability of the detection of aggregates in the peripheral blood smear.

Therefore, in order to define a meaningful difference in platelet counts between EDTA and citrate WB, a study would have to be performed that overcomes the several limitations of the current study: During it, peripheral blood smears should be performed indiscriminately for all concomitant platelet counts in EDTA and citrate WB (and ideally in magnesium-sulfate WB, also), with the evaluation of the peripheral blood smears being performed in a standardized manner, possibly following the protocol proposed by the Groupe Francophone de’Hematologie Cellulaire (GFHC) [12] or that designed by Lunde et al. [24]. A lack of standardization in this regard must certainly be mentioned among the limitations of the current study, as does the aforementioned bias introduced by the process that guided the selection of samples from which peripheral blood smears were performed in our study. Further, the platelet counts would have to be performed in a much more representative cohort than the one examined by us: The low median platelet counts in both materials suggest that our data is subject to a considerable selection bias. Also, to increase the validity of the results obtained, at least a subset of samples with a standardized temporal interval between sample collection and determination of the platelet counts should be included. The lack of standardization of this interval in the current study, owed to its retrospective character, must also be counted among its limitations. As explained above, longer intervals between sample collection and measurement are associated with more pronounced EDTA-associated in vitro aggregation of platelets [2] while the platelet count in citrate WB might also decrease spuriously over time, due to a certain temporal instability of platelets in this matrix [11], [12], [13], [14], further advocating, by the way, the use of MgSO₄ as anticoagulant for the investigation of possible PTCP. At least the fact that platelet aggregates were very rarely found in cases in which platelet counts in citrate WB were lower than in EDTA WB suggests that the temporal delay in our study was sufficiently short in order for PTCP not to be “masked” in the absolute difference in platelet counts by the spurious decline of platelet numbers in citrate WB in a significant number of cases.

During a study performed as outlined above, not only the sensitivity and specificity of the automated flags might be appraised more realistically, but also a more valid predictor of the occurrence of artefactual in vitro aggregation of platelets might be defined.

Until such a study is performed, the question of the exact definition of PTCP remains difficult to answer. However, it can be concluded from the current study that automated flags, generated during the platelet count in one material alone, may be insufficient as a predictor of the occurrence of platelet aggregates and, by extension, of the occurrence of PTCP, which may in turn only be detected validly via the performance of platelet counts in at least two different materials in synopsis with the visual examination of peripheral blood smears of at least one of these materials, that returns the lower platelet count.

In the meantime, a change in nomenclature is in our opinion called for: The term pseudothrombocytopenia has two implications which are not universally correct: Firstly, it implies (via the prefix “pseudo”) that the finding of platelet counts below the reference range is always spurious, while our results show that there are cases in which platelet aggregation leads to decrease platelet counts in EDTA WB, but the platelet count in citrate WB also shows a thrombocytopenia, albeit with a higher platelet count. Secondly, the term implies, as mentioned above, that the phenomenon described by it always leads to platelet counts below the reference range, which is not necessarily true, as platelet aggregation could also be shown in our data for cases of completely normal platelet counts in EDTA WB. Therefore, the term pseudothrombocytopenia should be regarded as a misnomer and be replaced e.g. by “anticoagulant associated in vitro aggregation of platelets”, as an umbrella term, with the word anticoagulant optionally being replaced by any specific anticoagulant, such as EDTA or citrate, depending on the specific phenomenon described.