Article Text
PDF
Abstract
Aims Limited information is available on number and type of cells present in the pericardial fluid (PF). Current evidence and has been garnered with inaccurate application of guidelines for analysis of body fluids. This study was aimed at investigating the performance of automate cytometric analysis of PF in adult subjects.
Methods Seventy-four consecutive PF samples were analysed with Sysmex XN with a module for body fluid analysis (XN-BF) and optical microscopy (OM). The study also encompassed the assessment of limit of blank, limit of detection and limit of quantitation (LoQ), imprecision, carryover and linearity of XN-BF module.
Results XN-BF parameters were compared with OM for the following cell classes: total cells (TC), leucocytes (white blood cell [WBC]), polymorphonuclear (PMN) and mononuclear (MN) cells. The relative bias were −4.5%, 71.2%, 108.2% and −47.7%, respectively. Passing and Bablok regression yielded slope comprised between 0.06 for MN and 5.8 for PMN, and intercept between 0.7 for PMN and 220.3 for MN. LoQ was comprised between 3.8×106 and 6.0×106 cells/L for WBC and PMN. Linearity was acceptable and carryover negligible.
Conclusions PF has a specific cellular composition. Overall, automated cell counting can only be suggested for total number of cells, whereas OM seems still the most reliable option for cell differentiation.
- cell counting
- automation
- differentiation
Statistics from Altmetric.com
Introduction
The pericardium is a fibro-serous, double-walled, conical sac surrounding the heart and large vessels roots. The serous pericardium is composed by parietal and visceral sheets, contains an inner space conventionally defined pericardial cavity, and contains approximately 15–35 mL of pericardial fluid (PF) which has the main function of lubricating and protecting the heart.1 2 The accumulation of liquid within the pericardial space (ie, pericardial effusion), is most frequently due to infections malignancies, metabolic disease or other idiopathic causes.1–9
Limited information is currently available on number and type of cells present in human PF,7 10–18 since the vast majority of data have been generated in animal models.6,14 Overall, the average number of cells may be comprised between ~10×106 and~2200×106 cells/L, while the value may increase between 3600×106 and 14 000×106 cells/L in patients with pericardial effusions of possible neoplastic origin (table 1).7 10–18
Summary of the literature data
Meyers et al 7 originally showed that the leucocyte count in PF tends to be higher in patients with rheumatological diseases (ie, up to 12 100×106 cells/L) and with inflammatory disorders, especially in those with bacterial infections (ie, up 41 410×106 cells/L). Unlike these findings, lower leucocyte counts were observed in patients with hypothyroidism (ie, ~625×106 cells/L). Pericardial effusions associated with bacterial or rheumatic diseases are usually characterised by large presence of neutrophils (NE), while those in patients with hypothyroidism and/or cancer are prevalently characterised by overwhelming presence of monocytes (MO) and other mononuclear cells. The incidence of neoplastic effusions is rather heterogeneous across different epidemiological studies, thus posing important diagnostic dilemmas for distinguishing activated mesothelial cells (ME) from neoplastic elements.7 10–18 All published studies aimed at exploring the cellular composition of PF have some limitations. These basically include the limited sample size, the poor clinical characterisation of the patients, methodological drawbacks (both analytical and statistical), but are also plagued by inaccurate application of current guidelines for analysis of body fluids, such as those issued by the Clinical and Laboratory Standard Institute (CLSI document H56-A)18 or by the International Council for Standardization in Haematology (ICSH).19
The availability of analytically robust and validated methods for cell enumeration is essential for assessing the cellular composition of PF, thus allowing to define the precise aetiology of the effusion and establishing the most appropriate therapeutic management, especially in patients with malignancies or suspected bacterial infection.1–18
According to the CLSI document H56-A18, cell enumeration using counting chambers, and their further differentiation by optical microscopy (OM) on citospin stained with May-Grunwald-Giemsa, remain the reference approaches for cytometric assessment of PF. Nevertheless, not only microscopic enumeration has some well-known limitations (ie, high inaccuracy, poor standardisation and insufficient reproducibility, requires skilled technical staff).20 To evade some challenges of OM, the analysis of body fluids is now performed in many laboratories by using haematological analysers with dedicated modules21–26 or urinary sediment analysers.27–31
The XN-9000 (Sysmex, Kobe, Japan) is a fully automated haematological analyser equipped with a specific module designed for body fluids analysis (XN-BF), which displays excellent performance for cell count and differentiation in cerebrospinal, sinovial, pleural and ascitic effusions.21–24 32 Therefore, we planned an original study to compare OM and automated cytometric analysis of PFs in a cohort of adult subjects undergoing cardiac surgery.
Materials and methods
PF samples
PF samples were collected from 74 consecutive patients (37 women and 37 men), undergoing cardiac surgery, without pericardial diseases (eg, pericarditis), who signed an informed consent for being included in this study. The PF samples were conveyed to the local laboratory for routine analysis. All samples, collected in blood tubes containing K3EDTA (Becton Dickinson, Franklin Lakes, New Jersey, USA), were simultaneously assessed with OM and XN-BF within 2 hours from sampling. Both collection and analysis of samples were carried out according to the CLSI and ICSH guidelines.18 19 The study was approved by the ethical committee of the Papa Giovanni XXIII Hospital and was performed in accordance with the Declaration of Helsinki, under the terms of all relevant local legislation.
XN-BF mode
Total and differential cell counts in body fluid samples are generated by XN-BF by means of hydrodynamic focusing and flow cytometry. The XN-BF module generates total cell (TC-BF), leucocytes (white blood cell [WBC-BF]), polymorphonuclear cell (PMN-BF) and mononuclear cell (MN-BF) counts. Additional ‘research’ parameters include neutrophil (NE-BF), eosinophil (EO-BF), lymphocyte (LY-BF), monocyte (MO-BF) and high fluorescence cells (HF-BF) classification.21–24
A total volume of 88 µL is needed for analysis of BF.21–24 After analysis has been completed, XN-BF performs an automatic rinse cycle, followed by a background check for preventing carryover or cross-contamination with blood or other BF samples. The XN-BF has been calibrated and used according to manufacturer’s specifications. Internal quality control was performed on daily analysis (in duplicate), using two different levels of control material (XN-CHECK; Streck Laboratories, Omaha, Nebraska, USA).
Optical microscopy
OM cell count was performed using Nageotte counting chamber. The samples were initially diluted 1:20 or 1:200 with Turk reagent (Carlo Erba, Italy), and nuclear elements were then enumerated into 12 squares, corresponding to 7.5 µL of PF sample, with an optical microscope at ×400 magnification by two skilled laboratory professionals. A third experienced laboratory professional performed an additional analysis when disagreement was ≥5%. For differential cell counts, PF samples were centrifuged at 100×g for 5 min (Cytospin 2, Thermo Scientific, Massachusetts, USA) and stained with May-Grunwald-Giemsa reagent (Carlo Erba Reagents, Italy). The slide review was carried out at ×400 magnification with ×40 oil-immersion objective (Objective Plan-Apochromat 40×/1.3 Oil DIC M27, D=0.17 mm Carl Zeiss s.p.a, Italy) on 200 cells, by two skilled operators. A third experienced laboratory professional performed an additional analysis when disagreement was ≥5%.18 19 22–24 A more accurate cytomorphological evaluation with a second microscopic analysis was performed at ×1000 magnification with ×100 oil-immersion objective (Objective Plan-Apochromat), when necessary.
Methods comparison
Methods comparison was carried out on all 74 PF samples, which were analysed with both XN-BF and OM for TC and differential counts on cytospin stained with May-Grunwald-Giemsa-, according to the indications of CLSI document H56-A18 and ICSH guidelines.19 The morphological differentiation with OM allowed cell classification in one of the following groups: NE, lymphocytes (LY), MO, eosinophils (EO), basophils (BA), macrophages (MA), ME and other cells (OTH; also including ‘neoplastic cells’). Due to differential classification and designation of cells between XN-BF mode and OM, cells were clustered in a groups, thus enabling a more consistent comparison, as follows:
TC-BF versus TC-OM in absolute value.
WBC-BF versus WBC-OM (=TC-OM–[ME+OTH]) in absolute value.
MN-BF versus MN1-OM (=LY+MO) in absolute and relative value.
MN-BF versus MN2-OM (=LY+MO+MA) in absolute and relative value;
MN-BF versus MN3-OM (=LY+MO+MA+ME+OTH) in absolute and relative value.
MN1-BF(=MN-BF+HF-BF) versus MN3-OM (=LY+MO+MA+ME+OTH) in absolute value (×106 cells/L).
PMN-BF versus PMN-OM (=NE+EO+BA) in absolute and relative value.
The agreement between XN-BF and OM was assessed with Passing and Bablok regression analysis and Bland-Altman plots. Slope and intercept of Passing and Bablok regression were calculated with their 95% Confidence Interval (95% CI). In Bland-Altman plots, absolute and relative differences were plotted against results of OM. The bias was considered significant when the limits of 95% CIs were both above or below zero.
The differences between methods were expressed as percentages of mean values on the y axis (ie, proportionally to magnitude of measurements: [(method A–method B)/mean %)]). Plotting the differences between methods as a percentage of mean values is preferred when the amplitude of differences tends to grow in parallel with measured values. The bias was considered significant when the limits of 95% CIs were both above or below zero.
The normal distribution of data was tested with Shapiro-Wilk test, with significance level set at 5%. The comparison between the median or the mean value was carried out with Kruskal-Wallis or Student’s t-test, with 5% level of significance.
Evaluation of carryover
The carryover of XN-BF for PF assessment was investigated using two PF samples with high cellularity (7186×106−35 782×106 cells/L). Both samples were analysed in triplicate (A1, A2, A3), followed by triplicate analysis of saline solution (B1, B2, B3). The carryover was expressed as percentage (%) and calculated according to the equation: [(B1−B3)/(A3−B3)]×100.18 19
Limit of blank and limit of detection
Limit of blank (LoB) and limit of detection (LoD) for TC-BF and WBC-BF were investigated according to CLSI document EP17-A2.32 LoB was calculated using a non-parametric analysis, as the 95th percentile value obtained from analysis of 60 replicates of PF samples displaying undetectable cells at OM. LoD was calculated using six PF samples, diluted with saline solution, so achieving very low cellularity. Ten replicates of each sample were measured, totalling 60 measurements. The mean values calculated for both TC-BF and WBC-BF parameters ranged between 1 and 6×106 cells/L, respectively. The LoD was calculated as the lowest TC-BF or WBC-BF value which could be measured (with 95% probability) over their respective LoBs. LoD was finally calculated using the equation: (LoD) = (LoB +1.645×SD) (where SD is the pooled SD of results obtained on low value samples).
Functional sensitivity (limit of quantitation)
Functional sensitivity (limit of quantitation [LoQ]) was assessed with 10 replicates of six native samples displaying different cell values, as follows: 2×106−5779×106 cells/L for TC-BF, 2×106−4653×106 cells/L for WBC-BF, 1×106−365×106 cells/L for PMN and 2×106−4366×106 cells/L for MN. The mean TC-BF, WBC-BF, PMN and MN count obtained in each sample was plotted against the coefficient of variation. Functional sensitivity was then mathematically calculated from power regression equation at a concentration where the imprecision corresponded to 20%.32
Linearity
Linearity of XN-BF was also assessed using native samples. A PF sample with high cellularity was serially diluted with Phosphate Buffered Saline, to generate eight different values ranging between 68×106−1999×106 cells/L for TC-BF, 35×106−1233×106 cells/L for WBC-BF, 6×106−76×106 cells/L for PMN-BF and between 26×106and 1200×106 cells/L for MN-BF, respectively. Each PF sample was measured consecutively for five times and the mean value was finally calculated. Results were plotted against the expected cell counts, and linearity was assessed according to the CLSI document EP06-A.33
Imprecision
The within-run imprecision of XN-BF was calculated using 10 replicates of six fresh PF samples, according to the CLSI document EP05-A3.34 The mean TC-BF values in PF samples were comprised between 74.1×106 and 5427.7×106 cells/L.
Results
Comparison between XN-BF and OM
The final study population consisted of 74 patients undergoing cardiac surgery for the following reasons: 19 (25.6%) for mitral valvulopathy, 25 (33.7%) for aortic valvular stenosis, 11 (14.9%) for coronary artery disease, 5 (6.8%) for dilated cardiomyopathy, 3 (4.1%) for unstable angina, 11 (14.9%) for other cardiac diseases. The total number of cells in the PF samples was comprised between 37×106 and 8453×106 cells/L, with a median value of 1846×106 cells/L. Table 2 shows the median value and range of the different cell populations obtained with OM and XN-BF. The Passing and Bablok regression analysis of XN-BF and OM values for total cell and leucocytes yielded slopes of 0.9 and 1.6 and intercepts of 26.1 and 133.8, respectively. The bias was −97.7×106 cells/L for total cells and 594.2×106 cells/L for leucocytes, respectively (figure 1).
Passing and Bablok regression total cell and leucocytes counts obtained with XN-BF versus optical microscopy. (A) Passing and Bablok regression TC-BF versus TC-OM y=0.9x+26.1 (95% CI slope: 0.8 to 1.00; 95% CI intercept: −24.1 to 132.2). (B) Passing and Bablok regression WBC-BF versus WBC-OM y=1.6x+133.8 (95% CI slope:1.3 to 2.3; 95% CI intercept: −17.6 to 383.7). TC-OM, total cell counts by optical microscopy; TC-BF, total cell counts by XN module; WBC-OM, white blood cell counts by optical microscopy; WBC-BF, white blood cell counts by XN module.
Cellular range and median value of all different cell counts
The Passing and Bablok regression analysis of XN-BF and OM values for the different cell populations are shown in table 3. Overall, the slopes were comprised between 0.06 (for MN-BF% vs MN1-OM% and MN-BF% vs MN2-OM) and 2.4 (for PMN-BF vs PMN-OM in absolute value), and the intercepts between 0.7 (for PMN-BF vs PMN-OM in percent value) and 220.3 (for MN-BF vs MN1-OM in absolute value). The MN-BF parameter (absolute values) displayed a high correlation with MN3-OM (ie, the sum of MO, MA, ME and other cells) compared with MN2-OM and MN1-OM (y=0.6x+10.7; y=1.6x+103.3 and y=1.9x+220.3, respectively). The bias ranged between −47.7% for MN expressed in absolute values (MN-BF vs MN3-OM) and 127.1% for PMN expressed in relative value (PMN-BF vs PMN-OM) (table 3 and figure 2).
Comparison of absolute cell counts obtained with XN-BF and optical microscopy (OM). (A) Bland Altman bias plot MN-BF versus MN1-OM with relative bias of 81.1% (95% CI 68.9 to 93.2), and 95% lower and upper limit of agreement (LoA) −21.92% (95% CI −42.82 to −1.02) and 184.07% (95% CI 163.16 to 204.96) (B) Bland Altman bias plot MN-BF versus MN2-OM with relative bias of 70.9% (95% CI 59.1 to 82.9) and 95% lower and upper LoA −29.83% (95% CI −50.28 to −9.37) and 171.75% (95% CI 151.30 to 192.20) (C) Bland Altman bias plot MN-BF versus MN3-OM with relative bias of −47.7% (95% CI −54.8 to −40.5) and 95% lower and upper LoA −108.04% (95% CI −120.28 to −95.79) and 12.65% (95% CI 0.41 to 24.89) (D) Bland Altman bias plot MN1-BF versus MN3-OM with relative bias of −6.0% (95% CI −9.9 to −2.1) and 95% lower and upper LoA −39.22% (95% CI −45.96 to −32.48) and 27.18% (95% CI 20.44 to 33.91). MN-BF, mononuclear cell by XN-BF; MN-OM, monocyte by optical microscopy.
Comparison between XN-module in body fluid mode and optical microscopy for the following parameters: total count of nucleated elements (TC-BF), leucocytes (WBC-BF), polymorphonuclear cells (PMN) and mononuclear cells (MN)
The differential counting of total mononuclear cells in OM (ie, MN3-OM) was better correlated with automated assessment on XN-BF compared with MN1-BF (ie, cell count obtained as a sum of MN-BF mononuclear cell count and high fluorescence cells HF-BF parameter). The resulting Passing and Bablok regression was y=0.9x+53.8, while the relative bias was −6%.
Carryover
The carryover was found to be negligible for all parameters investigated since it has never exceeded 0.3%.
LoB, LoD and functional sensitivity (LoQ)
The LoB was 1.7×106 cells/L for TC-BF, 1.0×106 cells/L for both WBC-BF and PMN and 0.0×106 cells/L for MN. The LoD was 2.9×106 cells/L for TC-BF, 2.0×106 cells/L for both WBC-BF and PMN and 0.9×106 cells/L for MN. The estimated LoQ for TC-BF and WBC-BF was 4.2×106 cells/L and 3.8×106 cells/L, while that for PMN and MN was 6.0×106 cells/L and 4.3×106 cells/L (table 4).
Limit of blank (LoB), limit of detection (LOD), functional sensitivity (limit of quantitation [LoQ]), linearity and within run imprecision of the XN-BF parameters: total count of the nucleated elements (TC-BF), leucocytes (WBC-BF), polymorphonuclear (PMN) and mononuclear cells (MN)
Linearity
The best fitting model was a linear regression for TC-BF (r=1.00), WBC-BF (r=0.99), MN (r=0.99) and PMN (r=0.96) (table 4). The bias between the mean values of WBC-BF or TC-BF and their expected values were always comprised within ±10% (range 64×106−2016×106 and 32×106−1268×106 cells/L, respectively).
Imprecision
The within-run imprecision of XN-BF for PF analysis (table 4) was comprised between 5.0% (mean value, 74.1±3.7×106 cells/L) and 1.8% (mean value, 1524±26.7×106 cells/L) for TC-BF, and between 4.7% (mean value, 63.0±3.0×106 cells/L) and 1.6% (mean value, 308.9±5.0×106 cells/L) for WBC-BF, respectively. The imprecision data for PMN-BF and MN-BF are also shown in table 4.
Discussion
The cytometric analysis of PFs plays an essential role for rapid and accurate diagnosis of pericardial diseases, as well as for establishing their more appropriate pharmacological treatment.1–6 The results of this study show that, in adult subjects without pericardial effusion and evident signs of pericardial pathology (eg, pericarditis or neoplastic infiltration), the median cell value in PFs (ie, 1846×106 cells/L) is aligned with earlier published data (table 1). Unlike these findings, however, the leucocyte count (ie, 428×106 cells/L) was found to be much lower than previously reported7 10–18 (table 1), and can be likely justified with results of cell differentiation. In our study ME (ie, 71.2%) were the most represented cellular elements in PF, while leucocyte count was obtained by subtracting ME and non-classifiable cells (ie, OTH) from the total cell count. A direct comparison between data obtained in this study and those earlier described in the scientific literature may hence be misleading. The poor reproducibility of literature data shall be seen as a major limitation of this comparison. In all previous articles, summarised in table 1, the methods used for performing cell count and differentiation, as well as the statistical approach employed for calculating median or mean values, have not been clearly described.
The performance of XN-BF mode for carryover, LoQ, linearity and imprecision seem overall acceptable according to the predictable cellularity of PF, as shown in table 3. The results obtained by comparing results obtained with XN-BF mode and OM show an acceptable agreement. The overall bias was −4.5% for the total cell count, similar to that already reported for ascitic, pleural and synovial fluids,35 and −3.2% for MN-BF% compared with MN3-OM% (ie, the sum of LY, MO, MA and ME) (table 3). Figure 2 shows a particular trend, highlighting that the bias seems proportional to the number of cells, while concordance is lower for all the other cell populations. The Passing and Bablok regression analysis yielded slopes >1.5 or <0.7, with bias ranging from −47.7% for MN-BF compared with MN3-OM, up to 127.1% by comparing PMN-BF with PMN-OM. These results are somehow unexpected considering the good analytical performance and previous correlation studies performed on other biological fluids (ie, % bias always <21% and slopes generally comprised between 0.9 and 1.0).35 This evident discrepancy can be attributed to some peculiarities of the PF, especially to the different types of cells that can be found in this biological material compared with ascitic and pleural fluids (table 2). The median value of ME was 71.2% in our study, and these cells tend to aggregate, as shown in figure 3A. Finally, the overall morphological characteristic of cells found in PFs are different from those normally present in ascitic and pleural fluids, although the composition was found to be generally similar (figure 3B,C). The high number of ME and the propensity of these cells to aggregate interferes with automated cell counting and, probably, also with counting chamber enumeration. ME and neoplastic cells are identified by XN-BF as high fluorescence cells in ascitic and pleural fluids22 and have a median value of 23×106 and 39×106 cells/L, respectively,22 35 while their median value is typically 20-fold higher in PFs (ie, 557×106 cells/L).
Optical microscopy (OM), white blood cell differential (WDF) Scattergram and extended WDF scattergram of pericardial, ascitic and pleural fluids without neoplastic infiltration or infection. All cells are clustered in according to their internal complexity (side scatter (SSC) axis) and nucleic acid content (side fluorescent, SFL axis). In the scattergram area, green clusters are mononucleated cell (MN-BF parameters), azure clusters are polymorphonuclear cell (PMN-BF parameter), blue clusters high fluorescent cell (HF-BF parameter) and the blue clusters near the axis are debris cells Pericardial fluid sample: (A) morphological characteristic of cells showed by OM (×400 magnification) on cytospin stained May-Grunwald-Giemsa. Differential count shows neutrophils (NE) 4.5%; lymphocytes (LY) 45.5%; monocyte (MO) 7%; macrophages (MA) 1%; mesothelial cells (ME) 42% (B) and (C) WDF-Scattergram and extended WDF scattergram show the prevalence of mononuclear cells (two green cluster). The HF-BF shows an abnormal cluster cell, indicated by the black arrow, the graphic representation of cell populations does not maintain the colour Code, since half of the cell cluster in blue, while the other half appears in green (figure 3D,G). (TC-BF: 3943×106 cells/L; PMN: 3.5%; MN: 96.5%; HF-BF: 1750×106 cells/L by XN-module body fluid mode.) Ascitic fluid sample: (D) morphological characteristic of cells showed by OM (×400 magnification) on cytospin stained May-Grunwald-Giemsa. Differential count shows NE 4%; LY 76%; MO 11%; MA 4%; ME 5% (E) and (F) WDF-Scattergram and extended WDF scattergram show the prevalence of mononuclear cells (two green cluster) and normal blue HF-BF cluster cell, indicated by the black arrow (TC-BF: 485×106 cells/L; PMN: 5%; MN: 95% HF-BF: 35×106 cells/L by XN-module body fluid mode).(Pleural fluid sample: (G) morphological characteristic of cells showed by OM (×400 magnification) on cytospin stained May-Grunwald-Giemsa. Differential count shows NE 52.5%; LY 15%; MO 7.5%; MA 18%; ME 7% (H) and (I) WDF-Scattergram and extended WDF scattergram show equal distribution between polymorphonuclear (azur cluster cell) and mononuclear cells (two green cluster) and normal blue HF-BF cluster cell, indicated by the black arrow (TC-BF: 1200×106 cells/L; PMN: 51.1%; MN: 48.9%; HF-BF: 58×106 cells/L by XN-module body fluid mode).
The comparative analysis of both the XN-BF white blood cell differential (WDF) and WDF-extended scattergrams generated using a PF with those produced on ascitic and pleural fluid samples (figure 3D–I) shows an abnormal cellular cluster of HF-BF (figure 3D,G). The graphic representation of cell populations does not maintain the colour code, since half of the cell cluster (corresponding to the events counted as high fluorescence cells) appears in blue, while the other half (corresponding to the events classified as MN-BF), appears in green (figure 3D,G). This finding is then confirmed by comparing the sum of MN-BF and HF-BF (MN1-BF) with mononuclear cells count in OM, which includes LY, MO, MA, ME and OTH (MN3-OM). In this specific case, Passing and Bablok regression analysis yielded a slope of 0.9, while the bias was −6.0% (table 3).
In conclusion, the results of this study based on analytical protocols recommended by the CLSI document H56-A18 and ICSH guidelines,19 show that automated cell counting in PFs shall only be suitable for estimating the total number of cells, whereas OM is still advisable for cell differentiation, although no definitive evidence is currently available to support the clinical validity of this information. We also demonstrate that the cell clusters present in the PF may differ from those of other BFs (eg, ascitic and pleural fluid). This preliminary data should then be confirmed in larger series, with thorough assessment of PF.
Take home messages
The performance of Sysmex XN body fluid mode seem overall acceptable according to the predictable cellularity of pericardial fluid (PF).
In adult subjects without pericardial effusion and evident signs of pericardial pathology the pericardial fluid has a specific cellular composition the mesothelial cells were the most represented cellular elements.
The results of this study, based on analytical protocols recommended by the Clinical and Laboratory Standard Institute document H56-A and International Council for Standardization in Haematology guidelines, show that automated cell counting in PFs shall only be suitable for estimating the total number of cells, whereas optical microscopy is advisable for cell differentiation.
References
Footnotes
Handling editor Professor Mary Frances McMullin.
Contributors All authors confirmed they have contributed to the intellectual content of this paper and have met the following three requirements: (1) significant contributions to the conception and design, acquisition of data, or analysis and interpretation of data; (2) drafting or revising the article for intellectual content; (3) final approval of the published article.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval This study was approved by the Ethics Committee of ASST Papa Giovanni XXIII.
Provenance and peer review Not commissioned; externally peer reviewed.