This paper examines the potential roles of graft platelet count in relation to postCD34 recovery and issues in apheresis stem cell collection and processing. Instead of finding a replication of the reported negative correlation of platelet count with postCD34 viability, we discovered that postCD34 recovery was clearly reduced when platelet count went either very high (i.e. the G6 group) or very low (i.e. the G1 group). It was further discovered that the G1 group was enriched with collections from NHL and Hodgkin patients often submitted to more intensive mobilization strategies and the G6 group was enriched with collections from Myeloma patients, and both groups had elevated male preponderance.

These results indicated that postCD34 recovery reduction could be associated with cancer types and gender, which was confirmed in the multivariate regression analysis. Considering WCC varied within a relatively narrow range (Fig. 1d and Supplementary Fig. S2), graft platelet counts probably indicated an underlying CD34 cell state that sustained when platelet counts were within a niche range but went out of kilter when platelet counts went extremes. Unfortunately, we don’t have information available about the cancer, such as the staging and previous chemotherapies, to better dissect this variable.

The observation of reduced postCD34 recovery in G1 reflected the fact that NHL and Hodgkin patients tended to have much lower platelet counts in the graft than Myeloma patients as showed in Table 1. NHL and Hodgkin’s patients were often mobilized with chemotherapy before apheresis collection, which could result in low platelet counts in the blood stream. The strong correlation between the blood stream platelet count and the graft platelet count (Supplementary Fig. S1) supports this explanation. Furthermore, of a total of 30 “Chemo + G-CSF”, 13 were non-Myeloma: 12 in Group1, 1 NHL-B with platelet count of 720 ×109/L in Group 2. It can be argued that the expected myelotoxicity of the chemotherapy used for lymphomas is higher than the drugs used for myeloma, explaining the fact from 17 Myeloma, only 3 were in Group1. Hence, the observation coincided with the delayed engraftment platelet recoveries in NHL patients in our previous study using the same protocol [5]. While it is of interest to see how well the delayed platelet recovery might be replicated in NHL patients of the current study cohort, it is intriguing why most but not all NHL patients had rather low platelet concentrations. A similar question can be asked why rather high platelet concentrations presented only in some Myeloma patients. Much of the answer would probably come from oncology and pathophysiology studies but the clues could be highly informative to apheresis collection and cell processing for these patients.

It is unclear how the platelet content of the graft could be associated with stem cell survival. One possibility is that increased platelet content and, therefore, degranulation during the cryopreservation process would lead to an inflammatory response and cell loss, as seen in G6. On the other hand, in an early review of PBSC transplant, delayed platelet recovery in heavily pre-treated patients undergone intensive chemotherapy was discussed along with CD34 cell state in terms of the disrupted cell cycle by mobilization and endogenous factors such as endogenous cytokines and/or the haematopoietic microenvironment [11]. Further studies confirmed the impact of cell cycle and micro-environment [12, 13]. Furthermore, the composition of CD34 cell subpopulations appeared to be significantly different across sources of stem cell collections and thus could be used to predict engraftment kinetics and immune reconstitution in recipients [14, 15]. It is plausible to hypothesize that the same causes promoting different graft platelet content could also be associated with different inflammatory response patterns, stem cell repertoire and functional capabilities. Hence, if the platelets have a direct role determining cell viability or act as a surrogate marker of the underlying “cell state”, it is still undetermined, and further investigation must take place.

This study highlighted the need for comprehensive analysis accompanied by careful dissection of clinical observations when studying complicated clinical problems. As reported at the beginning of the Results section, using only Pearson’s correlation analysis of all data would lead to a wrong conclusion that platelet count was significantly negatively correlated with postCD34 recovery, which in fact was observed only in the G4 group (Fig. 1c). The wrong conclusion could even be considered as a replication of Valentini et al. [6] if not careful. Indeed, taking a close look at the means (variation range) and distribution of raw platelet count reported in Table 1 and Figure 3 of Valentini et al. [6] respectively, one could easily tell that the higher mean of platelet counts was driven largely by two outliers with extremely high platelet counts. Another example is the characterisation of the G1 group where median and mean (Table 2) and recovery (Fig. 1b) together made a nearly complete view and thus enabled in-depth analysis to derive the clinical insights into this odd group. It is important to bear in mind that the timing of the platelet assessment in the graft can be different among studies. In this study, we assessed the platelet content before the processing and addition of the cryopreservation agents and hence, dilution of the cell concentration. If the assessment was done immediately before cryopreservation, the figures would be different.

The current report also showed that modelling statistical interactions in complex clinical problems was beneficial in revealing valuable new insights into the clinical complex that otherwise would have gone missing. The identified statistical interactions not only explained the extra variance of the complex problem but also helped with the interpretation of the observed data patterns (Fig. 1d) in a clinically meaningful way (Supplementary Fig. S3). For example, considering interactions between gender, cancer type and platelet group together could facilitate deriving clinical hypothesis to be tested to better understanding of the G1 and G6 groups. Nonetheless, cares should be taken when modelling and interpreting statistical interactions in small data that are vulnerable to risk of over-parametrization. In any case, replication in independent data is highly recommended.

Our study has some limitations. Firstly, the sample size is relatively small because it is not the current practice to monitor the platelet concentration of the PBSC units. Secondly, the retrospective nature is associated with inevitable biases. Thirdly, engraftment outcomes are yet to become available to assess the impact of reduced postCD34 recovery. Moreover, our study defined the viable post-thaw CD34 recovery as the end-point and Valentini, as the post-thaw CD34 viability. The postCD34 recovery is based on the comparison between the viable CD34 cell enumeration in the pre-processing and the post-cryopreservation samples. On the other hand, CD34 viability is assessed independently in each sample as a percentage of the total CD34 cells (viable and non-viable). Therefore, although recovery and viability are informing the underlying quality of the graft, and they tend to be generally concordant, they cannot be interpreted as synonyms.

In conclusion, extreme graft platelet counts indicate a risk of reduced postCD34 recovery. To validate these findings, further studies enroling a larger sample and with comprehensive data about the patients should be performed.