Transplantation (2000) 25, 1157–1164  2000 Macmillan Publishers Ltd All rights reserved 0268–3369/00 $15.00 www.nature.com/bmt Enumeration of HPC in mobilized peripheral with the Sysmex SE9500 predicts final CD34؉ cell yield in the apheresis collection

J Yu, W Leisenring, W Fritschle, S Heimfeld, H Shulman, WI Bensinger, LA Holmberg and SD Rowley

Clinical Research Division, Fred Hutchinson Cancer Research Center, Division of Oncology, University of Washington, Seattle, WA, USA

Summary: CD34ϩ cells/kg patient weight is recommended, with even more reliably rapid engraftment observed after infusion of Enumeration of CD34؉ cells in the peripheral blood higher CD34ϩ cell doses.1–4 CD34ϩ cells in the peripheral before apheresis predicts the quantity of those cells col- blood are normally present in low numbers. After mobiliz- lected, although the cytometric techniques used are ation with hematopoietic cytokines or chemotherapy- complex and expensive. We found that a subpopulation containing regimens, the number of CD34ϩ cells in the of lysis-resistant cells in the peripheral blood, identified blood increases greatly. However, multiple collections are by the Sysmex SE9500 and designated as HPC, can often required to reach the target dose of CD34ϩ cells for serve as a surrogate marker predictive of the yield of transplantation, either because of insufficient mobilization, CD34؉ cells. Spearman’s rank statistics were used to inappropriate timing and/or operation of the collection examine the correlation between WBC, MNC, HPC and procedures. CD34؉ cells in the peripheral blood and final CD34؉ The level of CD34ϩ cells in the peripheral blood reaches cell yield for 112 samples of peripheral blood and its peak at different times according to the mobilization matching apheresis collections from 66 patients and regimen used. For donors or patients treated with a hemato- donors. The results indicate that WBC and MNC in the poietic cytokine such as filgrastim only, the peak value of peripheral blood were poor predictors of CD34 content, CD34ϩ cells in the peripheral blood reliably occurs while HPC gave a correlation coefficient of 0.62. The between the fourth and sixth days,3,5,6 greatly simplifying positive predictive values of different cutoff levels of the management of these donors. However, for patients HPC in the peripheral blood ranging from 5 to 50 ؋ treated with a chemotherapy-containing regimen, the time 106/l increased from 0.80 to 0.93 when the target collec- to maximal CD34ϩ cells in the peripheral blood varies tion was 1 ؋ 106 cells/kg. However, for patients with greatly.7–9 The quantity of CD34ϩ cells is much higher after HPC levels below various cutoff levels, the proportion treatment with a chemotherapy-containing regimen, but it of the collections not reaching that target goal ranged is also more difficult to determine when peak levels of between 0.36 and 0.43, indicating that most collections CD34ϩ cells are present in the peripheral circulation. Some will still exceed the target goal of CD34؉ cells. When centers initiate apheresis collections when the white blood ,the target collection was 2.5 ؋ 106 CD34؉ cells/kg, the cell (WBC) count rebounds to a certain level. However positive predictive value was lower and negative predic- there is little correlation between the WBC count and the tive value was higher. Bone Marrow Transplantation number of CD34ϩ cells,7,10,11 resulting in inefficient collec- (2000) 25, 1157–1164. tions and increasing the overall cost of obtaining PBSC for Keywords: apheresis; CD34ϩ cells; hematopoietic stem the transplant. Other centers use the quantity of CD34ϩ cells; peripheral blood stem cells; stem cell transplantation; cells in the peripheral blood to time initiation of apheresis Sysmex hematology analyzer procedures. We and others have demonstrated a very close correlation between the number of CD34ϩ cells in the blood and the number of CD34ϩ cells collected by apheresis.7,10–13 However, the clinical utility of this test for the timing of Successful peripheral blood stem cell (PBSC) transplant- apheresis is limited by logistical factors. For example, cur- ation depends on the infusion of an adequate number of rent flow cytometric techniques to measure CD34ϩ cells hematopoietic stem cells to achieve rapid and durable hem- ϩ are technically difficult and expensive. Also, several hours atological recovery. The number of CD34 cells is widely are usually required to obtain the result, which complicates used as a parameter for qualifying the engraftment potential patient management. of the PBSC component. For rapid hematological recovery Sysmex has developed a hematology analyzer that ident- ϫ 6 and engraftment, infusion of a minimum of 2–5 10 ifies a small population of immature myeloid cells in the peripheral blood according to cell size, cell density and dif- ferential lysis resistance.14–16 The machine parameters for Correspondence: Dr SD Rowley, Fred Hutchinson Cancer Research identifying these cells were determined with the use of pur- Center, 1100 Fairview Ave N, PO Box 19024, D5-390, Seattle, WA ϩ 98109-1024, USA ified CD34 cell populations. Enumeration of these cells, Received 14 December 1999; accepted 14 February 2000 designated by Sysmex as hematopoietic progenitor cells HPC predicts CD34؉ cell yield JYuet al 1158 (HPC), requires use of a modified lysis solution, but is rap- techniques previously reported.3,13 In brief, leukapheresis idly and inexpensively accomplished. We studied the HPC was performed using a continuous flow separator count to determine if the quantity of these cells in the peri- (Spectra version 4.7; COBE BCT, Engelwood, CO, USA). pheral blood of patients and donors who underwent PBSC Venous access was established either by peripheral vein or collections could be correlated with the quantity of CD34ϩ central venous catheter. Anticoagulant solution, consisting cells in the apheresis component and could potentially be of heparin 10 U per ml of ACD-A, was infused at a ratio used to determine the timing of apheresis after admin- of 1 ml anticoagulant to 30 ml whole blood. A set volume istration of chemotherapy-based mobilization regimens. of 12 liters of blood processed per apheresis was used for the adult patients and donors who underwent standard vol- ume apheresis (n ϭ 54 patients and 83 procedures) and up Materials and methods to 6ϫ the patient’s total blood volume (TBV) not to exceed 36 liters for patients who underwent large volume apheresis (n ϭ 14 patients and 24 procedures). The blood volume Patient and donor population processed was 3ϫ TBV not to exceed 12 liters for pediatric patients (р48 kg) who received standard volume apheresis Patients and donors eligible for this study were referred to n ϭ the Fred Hutchinson Cancer Research Center (FHCRC) for ( 4 patients and 5 procedures). Inlet flow rate was main- tained at 50–80 ml per minute in the standard volume aph- autologous or allogeneic PBSC transplantation. Patients eresis procedure, 80–100 ml per minute in the large volume were referred for autologous PBSC transplantation for treat- ment of a variety of malignant diseases. Donors eligible apheresis procedure, and 2 ml/kg per minute not to exceed 80 ml per minute for the pediatric patients. The collection for this study were HLA-matched family members donating rate was maintained at 1–2 ml per minute. The criterion for PBSC for allogeneic transplantation. All patients and ϩ donors gave informed consent for the administration of fil- adequate PBSC collection was a target number of CD34 cells of 5 ϫ 106/kg patient weight. Once initiated, leuka- grastim with or without preceding chemotherapy and for collection of PBSC. All protocols for PBSC mobilization, pheresis was continued daily in an attempt to achieve collection and transplantation were approved by the this goal. Institutional Review Board of the FHCRC. Enumeration of nucleated cells in the peripheral blood Mobilization regimens and leukapheresis components The choice of mobilization techniques was dictated by the WBC counts in the peripheral blood were determined using clinical protocols in effect and by the disease status of the a Sysmex NE 8000 (Sysmex, Corporation of America, patients. For 44 patients, PBSC were mobilized with Long Grove, IL, USA). WBC counts in the apheresis chemotherapy followed by filgrastim (Amgen, Thousand components were obtained using a Sysmex E 2500. Cell Oaks, CA, USA), both to reduce the tumor burden and to differentials were manually performed on Wright Giemsa facilitate PBSC harvesting. Regimens included cyclophos- stained specimens and the numbers of mononuclear cells phamide (Cy) 4 g/m2 i.v. (n ϭ 3); Cy and paclitaxel 250 (MNC, defined as lymphocytes and monocytes) were mg/m2 i.v. (n ϭ 9); and Cy and etoposide 200 mg/m2 daily calculated. i.v. for 3 days (total dose 600 mg/m2) with (n ϭ 6) or without (n ϭ 9) dexamethasone 10 mg four times daily p.o. Enumeration of HPC for 4 days (total dose 160 mg).17 For seven patients PBSC were collected after a variety of different chemotherapy Enumeration of HPC was performed using the Sysmex SE regimens specific to the disease being treated. All patients 9500. HPC were identified in the immature information treated with chemotherapy mobilization regimens also channel (IMI) that uses radio frequency (RF) and direct received filgrastim 10 ␮g/kg s.c. daily until completion of current (DC) methods to measure cell size and cell density. the collection procedures. In general, apheresis procedures From the orginal sample, 2.4 ␮l is mixed with a proprietary for patients treated with chemotherapy-containing regimens lytic reagent (STROMATOLYSTER-IM), which lyses were initiated when the WBC in the peripheral blood mature leukocytes but leaves immature cells intact based ϫ 9 on the difference of lipid content between those cells.14–16 exceeded 1 10 /l after mobilization, although some ϩ patients were managed according to peripheral blood Using purified CD34 cells, the lower RF and DC signal CD34ϩ cell level. For this study, the HPC level in the per- were previously identified for the HPC detection area. In ipheral blood was not used to time apheresis. For 11 the IMI scattergram, HPC are presumed to appear in a dis- patients who did not need cytoreduction and 11 allogeneic tinct area of the blast region. HPC were reported both as donors, PBSC were mobilized with filgrastim alone at doses absolute number and as a percentage of the WBC in the of 10–16 ␮g/kg/day s.c. For these patients and donors, sample. apheresis was initiated on the fourth (autologous patients) or fifth (allogeneic donors) day of filgrastim treatment. Enumeration of CD34ϩ cells The quantities of CD34ϩ cells in the peripheral blood and Collection of peripheral blood stem cells the leukapheresis component were determined by a stan- A total of 112 PBSC components were collected from 55 dard flow cytometric analysis technique as previously patients and 11 normal allogeneic donors using apheresis described.13 A sample of 1 ϫ 106 nucleated cells was

Bone Marrow Transplantation HPC predicts CD34؉ cell yield JYuet al 1159 stained with phycoerythrin-conjugated anti-CD34 (HPCA- tic regression models. Since individual patients may con- 2; Becton Dickinson, San Jose, CA, USA) and fluorescein tribute more than one collection to the analysis, standard isothiocyanate-conjugated anti-CD14 (Leu-M3; Becton error estimates used for obtaining confidence intervals were Dickinson) at the concentrations recommended by the calculated using robust sandwich variance estimates as manufacturer after lysis of red blood cells in 0.83% described previously.20,21 Analyses were also conducted ammonium chloride lysis solution and washing with phos- using the target value of 5 ϫ 106 CD34ϩ cells per liter of phate-buffered saline (PBS) with 0.5% (w/v) human serum blood processed during apheresis as the definition of an albumin. Additional cells were stained with the anti-CD14 adequate collection (not reported). conjugate and phycoerythrin-conjugated IgG1 (Becton Dickinson) as an isotope control. After incubation for 25 min at 4°C in a light protected area, the cells were washed Results twice and resuspended in PBS with 1% paraformaldehyde. Analysis was performed on a FACSCAN (Becton Dickin- Patient and donor population son, Palo Alto, CA, USA). Listmode data were analyzed using Winlist 2.0 software (Verity Software House, Top- Patient and donor characteristics are shown in Table 2. There were 44 patients treated with chemotherapy followed sham, ME, USA). A region was established to include all nucleated cells and exclude and red blood cells by filgrastim and 22 patients and donors (11 each) who using forward and 90° light scatter. A second gate was were treated with filgrastim alone. established to include only CD14-negative cells with low side scatter. The number of bright CD34ϩ cells with low Cell counts in the peripheral blood and PBSC side scatter was then determined. The percentage of CD34ϩ components cells was calculated by subtracting the number of cells ϩ stained with the isotope control from the number of cells Peripheral blood cell counts and the yield of CD34 cells for all collections are shown in Table 3. There were 67 stained with the CD34 antibody and dividing by the number collections after chemotherapy plus filgrastim mobilization of nucleated cells counted in the first region. The number of CD34ϩ cells in the peripheral blood or apheresis compo- regimens and 45 collections after filgrastim only mobiliz- ation. The median number of CD34ϩ cells collected (per nent was determined by multiplying this percentage by the kg patient weight) was 2.3 ϫ 106 after chemotherapy-con- total number of WBC in the peripheral blood or leuka- ϫ 6 pheresis component. taining mobilization and 2.2 10 for collections after fil- grastim alone. The higher median numbers of CD34ϩ cells (ϫ 106/l) and HPC (ϫ 106/l) in the peripheral blood after Statistical methods chemotherapy-containing mobilization reflects the better mobilization achieved using chemotherapy in combination This is a retrospective study of HPC quantity as a predictor ϩ with cytokines. Similarly, the lower peripheral blood WBC of CD34 cells collected by apheresis. Patient and donor characteristics, as well as peripheral blood and apheresis for these patients compared to that for patients and donors treated only with filgrastim reflects the recovery of these components are described using summary statistics such as median values and ranges. Associations between various patients from marrow-hypoplasia producing chemotherapy factors were assessed using Spearman’s rank correlation regimens. coefficient (␳).18 Linear regression models were fitted to the logarithm of the number of CD34ϩ cells in the apheresis Table 2 Patient and donor characteristics component as a function of the logarithm of CD34ϩ cells, HPC, WBC and MNC in the peripheral blood.19 To evalu- Chemotherapy Filgrastim only plus filgrastim ate the ability of HPC to perform as a diagnostic test for ϩ ϫ 6 predicting target yields of CD34 cells of 1 10 and 2.5 No. patients 44 22 6 ϫ 10 cells/kg, we calculated the sensitivity, specificity and Age: median (range) 43.5 (1–66)a 47.00 (26–70) predictive values of HPC, utilizing several different cutoff Weight: median 76.5 (11.0–118) 69.50 (48.0–115.0) points for HPC (80, 50, 20, 5 ϫ 106/l). These quantities (range) are defined in Table 1. Where sufficient data were available, Gender: F:M 24:20 12:10 Diagnosis: confidence intervals were calculated for these measures BCAb 10 0 using probabilities and standard errors predicted from logis- NHL 9 3 MM 7 1 Others 18 7 Table 1 Key measures of accuracy for a diagnostic tool Donor — 11

ϩ у Positive predictive value: Probability of CD34 cell collection aShown are the patient and donor characteristics including age, weight, у target for patients with HPC criterion gender and diagnosis, categorized under chemotherapy-containing regimen ϩ Ͻ Negative predictive Probability of CD34 cell collection or filgrastim only regimen. The regimens used for mobilization are Ͻ value: target for patients with HPC criterion described in Methods. у Sensitivity: Probability of HPC criterion for patients bBreast cancer (BCA), non-Hodgkin’s lymphoma (NHL), multiple mye- ϩ у with CD34 cell collection target loma (MM), and others including acute myeloid leukemia, chronic Ͻ Specificity: Probability of HPC criterion for patients myeloid leukemia, chronic lymphoblastic leukemia, Hodgkin’s disease, ϩ Ͻ with CD34 cell collection target ovarian carcinoma, germ cell tumor, medulloblastoma, primitive neuroec- todermal tumor or osteosarcoma.

Bone Marrow Transplantation HPC predicts CD34؉ cell yield JYuet al 1160 Table 3 Cell counts in the collection components and peripheral blood Table 4 The correlation coefficients between peripheral blood cell counts and the yield of CD34ϩ cells (ϫ 106/kg) Chemotherapy plus Filgrastim only filgrastim (n ϭ 45) Chemotherapy Filgrastim (n ϭ 67) plus filgrastim only

Apheresis component CD34ϩ cells 0.96a 0.94 CD34ϩ cells 2.3 (0.1–48.1)a 2.2 (0.4–10.7) (ϫ 106/l) (ϫ 106/kg) HPC (ϫ 106/l) 0.62 0.59 Peripheral blood WBC (ϫ 109/l) Ϫ0.18 0.43 counts MNC (ϫ 109/l) Ϫ0.03 0.52 CD34ϩ cells 22.3 (0.9–502.0) 16.2 (2.4–113.8) ϫ 6 b ( 10 /l) aShown are the correlation coefficients (␳) between cell counts in the peri- ϫ 6 HPC ( 10 /l) 74.8 (1.0–1240.2) 65.5 (3.1–1353.9) pheral blood including CD34ϩ cells, HPC, WBC and MNC and the yield ϫ 9 WBC ( 10 /l) 26.2 (1.4–68.1) 40.5 (6.3–144.0) of CD34ϩ cells (ϫ 106/kg). The number of collections analyzed was 67 ϫ 9 MNC ( 10 /l) 2.6 (0.4–7.8) 4.9 (0.8–21.6) from chemotherapy-mobilized patients and 45 from filgrastim only mobil- ized patients and donors. The number of collections for which CD34ϩ cell aShown are the medians (range) of the yield of CD34ϩ cells (ϫ 106/kg counts from the peripheral blood were available were 28 (after chemo- patient body weight) in the apheresis component and of the different cell therapy mobilization) and 33 (after filgrastim only mobilization). counts in peripheral blood, categorized by mobilization regimen. bThe quantity of CD34ϩ cells (ϫ 106/l) was available for 28 collections for patients treated with a chemotherapy-containing mobilization regimen and 33 collections for the patients and donors treated with filgrastim-only highly predictive for the yield of CD34ϩ cells in the apher- mobilization regimen. esis collection. The HPC level in the peripheral blood had a stronger correlation with CD34ϩ cell yield than did WBC and MNC counts, but was weaker than the correlation The correlation between cell counts in the peripheral ϩ blood and the yield of CD34ϩ cells between CD34 cell count in the peripheral blood and CD34ϩ cell yield. The correlation coefficients between The relationships between various peripheral blood cell HPC (ϫ 106/l) and CD34ϩ cells (ϫ 106/l) in the peripheral counts at the start of apheresis and CD34ϩ cell yield (per blood was 0.55 (n ϭ 28). The correlation coefficients kg body weight) in the apheresis component are shown in between cell counts in the peripheral blood and CD34ϩ cell Figure 1, and the correlation coefficients (categorized by yield per liter of blood processed during the apheresis pro- mobilization technique) are listed in Table 4. For collec- cedure were found to be very similar to those between cell tions after either a chemotherapy-containing mobilization counts in the peripheral blood and CD34ϩ cell yields per regimen or filgrastim only, the CD34ϩ cell count was kg body weight described above (data not shown).

400 ab400 200 200 100 100 /kg) /kg) 50 50 6 6 25 25 12 12 6 6 3 3 cells (x10 cells (x10 1 1 + + CD34 CD34 0.1 0.1 1 12 25 50 100 200 400 800 1 12 25 50 100 200 400 800 in the apheresis component in the apheresis component CD34+ cells (x106/l) in the peripheral blood HPC (x106/l) in the peripheral blood

400 cd400 200 200 100 100 /kg) /kg) 50 50 6 6 25 25 12 12 6 6 3 3 cells (x10 cells (x10 1 1 + + CD34 CD34 0.1 0.1 1 12 25 50 100 200 400 800 1 12 25 50 100 200 400 800 in the apheresis component in the apheresis component WBC (x109/l) in the peripheral blood MNC (x109/l) in the peripheral blood

Figure 1 The relationship between the yield of CD34ϩ cells and peripheral blood cell counts. Shown are the yield of CD34ϩ cells (ϫ 106/kg) in the apheresis components (Y-axis) vs:(a) the number of CD34ϩ cells ϫ 106/l (n ϭ 61, ␳ ϭ 0.98); (b) HPC ϫ 106/l (n ϭ 112, ␳ ϭ 0.62); (c) WBC ϫ 109/l (n ϭ 112, ␳ ϭϪ0.04); and (d) MNC ϫ 109/l (n ϭ 111, ␳ ϭ 0.14) in the peripheral blood (X-axis) in all subjects. The solid lines depict predicted values from linear regression models.

Bone Marrow Transplantation HPC predicts CD34؉ cell yield JYuet al 1161 The predictive value of HPC level in the pre-apheresis were met to carry out the procedure. Because the groups peripheral blood of patients at the extreme HPC levels were small, the con- fidence intervals are quite wide in these regions and some- The above results suggest that HPC level in the peripheral ϩ times not estimable. Only one patient had a HPC level blood may predict the yield of CD34 cells collected by below 5 ϫ 106/l on the day of collection (this person failed apheresis and may be useful in determining when to initiate ϩ to reach the target goal of CD34 cells in the one collection apheresis procedures after a chemotherapy-based mobiliz- obtained) and the negative predictive value of this cutoff ation regimen. We evaluated the HPC level in the peri- level cannot be determined. pheral blood as a diagnostic tool for predicting a target ϩ In this study, the percentages of the collections reaching yield of CD34 cells of at least 1 ϫ 106/kg and 2.5 ϫ the target yields of 1 ϫ 106 and 2.5 ϫ 106 CD34ϩ cells/kg 106/kg for a defined subset of patients who underwent col- were 80% and 50%, respectively. Therefore, these predic- lection after a chemotherapy-containing regimen and for whom 10–14 liters of blood were processed during the aph- tive values of HPC are only applicable to a similar popu- eresis procedure. We quantified the degree to which higher lation of patients for which those percentages of collections HPC levels in the peripheral blood were associated with a are expected to reach the target, regardless of HPC levels. greater probability of reaching the target yield of CD34ϩ However, the sensitivity and specificity of HPC, which are cells (Table 5). often used to describe the characteristics of a diagnostic For patients with HPC levels у80 ϫ 106/l, all collections test, do not rely on the proportion of the patients who reach reached the target yield of at least 1 ϫ 106 CD34ϩ cells/kg. the target. Table 6 displays the sensitivity and specificity The positive predictive values of different cutoff levels of of HPC as a diagnostic test. The sensitivity decreases and HPC in the peripheral blood ranging from 5 to 50 ϫ 106/l the specificity improves when the cutoff levels of HPC are increased from 0.81 to 0.93 for all collections. Thus, a increased. Figure 2 shows the trade-off in sensitivity and higher HPC level in the peripheral blood predicted that a specificity as the cutoff level of HPC is shifted. Among the ϫ 6 ϩ higher proportion of the collections would reach the target collections reaching the target yield of 1 10 CD34 goal of 1 ϫ 106 CD34ϩ cells/kg. In contrast, for patients cells/kg, 100% had a HPC level in the peripheral blood ϫ 6 with HPC levels below a cutoff level of 80, 50 or 20 ϫ above 5 10 /l, 89% had a HPC level in the peripheral 106/l, the proportion of the collections not reaching the tar- blood above 20 ϫ 106/l and 71% had a HPC level in the get goal of 1 ϫ 106 cells/kg ranged between 0.36 and 0.43, peripheral blood above 50 ϫ 106/l. Among the collections ϩ indicating that most collections will still exceed this target not reaching the target yield of 1 ϫ 106 CD34 cells/kg, goal of CD34ϩ cells. Considering 2.5 ϫ 106 CD34ϩ 100% had a HPC level in the peripheral blood below 80 cells/kg as the target yield resulted in a higher negative ϫ 106/l and 78% had a HPC level in the peripheral blood predictive values than were seen with a target yield of 1 ϫ below 50 ϫ 106/l. Note that because a small number of 106 cells/kg for all HPC cut off levels, but at the cost of patients failed to reach the target collection goal of 1 ϫ lower positive predictive values (Table 5). In other words, 106 CD34ϩ cells/kg, the estimates of specificity are one could have more confidence deciding not to carry out accompanied by fairly wide confidence intervals. If the tar- apheresis based on a criterion of having an HPC Ͻ80 ϫ get yield is considered to be 2.5 ϫ 106 CD34ϩ cells/kg, 106/l when the goal is to achieve a target of 2.5 ϫ 106 then, in comparison to the target yield of 1 ϫ 106 cells/kg, CD34ϩ cells/kg rather than a target of 1 ϫ 106 cells/kg. the sensitivity is increased and the specificity is decreased, However, this would mean that a higher percentage of over the full range of HPC cutoff values. That is, a higher aphereses would likely not meet that goal, if the criteria proportion of the successful apheresis collections were

Table 5 The predictive value (95% confidence intervals) of HPC for specific target collections of CD34ϩ cells among collections in which 10–14 liters of blood were processed

Cutoff value Positive-predictive value Negative-predictive value of HPC (ϫ 106/l) n (95% CI)a n (95% CI)a

Target of 1 ϫ 106 CD34ϩ cells per kg 80 19 1.0 (Ϫb, 1.0) 25 0.36 (0.19, 0.58) 50 27 0.93 (0.74, 0.98) 17 0.41 (0.18, 0.69) 20 37 0.84 (0.67, 0.93) 7 0.43 (0.14, 0.77) 5 43 0.81 (0.66, 0.91) 1 1.0 (Ϫ, 1.0)

Target of 2.5 ϫ 106 CD34ϩ cells per kg 80 19 0.89 (0.67, 0.97) 25 0.80 (0.54, 0.93) 50 27 0.67 (0.47, 0.82) 17 0.76 (0.44, 0.93) 20 37 0.54 (0.34, 0.73) 7 0.71 (0.32, 0.93) 5 43 0.51 (0.33, 0.69) 1 1.0 (Ϫ, 1.0) aShown are the predictive values and 95% confidence intervals (CI) for different HPC cutoff values. 95% CI were based on logistic regression models with robust sandwich variance estimates. bNot estimable.

Bone Marrow Transplantation HPC predicts CD34؉ cell yield JYuet al

1162 ϩ Table 6 The sensitivity and specificity of HPC for specific target col- blood varies greatly.7–9 Several predictors for CD34 cells lections of CD34ϩ cells amongst collections in which 10–14 liters of blood yield have been used by different transplant centers for the were processed timing of apheresis. The quantity of CFU-GM (colony- forming units –macrophage) has a close corre- Cutoff value of Sensitivity (95% CI)a Specificity (95% CI)a ϩ HPC (ϫ 106/l) lation with both CD34 cell quantity and engraftment, but requires 2 weeks to obtain results.9,12,22 The WBC count, n ϭ 35 n ϭ 9 an indicator of bone marrow recovery after chemotherapy Target of 1 × 106 CD34+ cells per kg administration, has been used to determine when to initiate 80 0.54 (0.35, 0.72) 1.0 (Ϫb, 1.0) apheresis after use of a chemotherapy-containing mobiliz- 50 0.71 (0.51, 0.86) 0.78 (0.45, 0.94) ation regimen. We previously published that a WBC count 20 0.89 (0.74, 0.95) 0.33 (0.12, 0.65) ϫ 9 5 1.0 (Ϫ, 1.0) 0.11 (0.02, 0.48) less than 3–5 10 /l was associated with a very poor yield of CD34ϩ cells.13 Furthermore, we and others have demon- Target of 2.5 ϫ 106 CD34ϩ cells per kg strated a poor correlation between WBC and the yield of ϩ ϩ 80 0.77 (0.51, 0.92) 0.91 (0.69, 0.98) CD34 cells.7,10–11,13 The level of CD34 cells in the peri- 50 0.82 (0.55, 0.94) 0.59 (0.41, 0.75) pheral blood correlates closely with CD34ϩ cell yield. Sev- 20 0.91 (0.71, 0.98) 0.23 (0.10, 0.44) eral authors have reported that a level of CD34ϩ cells in 5 1.0 (Ϫ, 1.0) 0.05 (0.01, 0.27) the peripheral blood above 10–20 ϫ 106/l was a good thres- 7,11–13,22 a hold at which to initiate apheresis. However, tech- Shown are the sensitivity/specificity and 95% confidence intervals (CI) ϩ under different HPC cutoff values. 95% CI were based on logistic niques to enumerate CD34 cells are complex and the regression models with robust sandwich variance estimates. reporting of results may take several hours, often resulting bNot estimable. in a 1 day delay before apheresis can be started. Sysmex have developed a technique to measure a popu- ϩ 14–16 1 lation of cells that appears to encompass CD34 cells. This test was developed using purified CD34ϩ cells, but unlike flow cytometry-based techniques, requires only 90 s 0.75 to perform. Detection and enumeration of HPC in the peri- pheral blood could possibly provide a standard and rapid Sensitivity alternative for predicting the yield of stem cells collected Specificity 0.5 by apheresis. Our data show a correlation between the num- bers of HPC and CD34ϩ cells in the peripheral blood, but

Probability at a ratio of about 3 or 4 to 1 indicating that this test also Ϫ 0.25 detects a population of CD34 cells. Since the HPC popu- lation is not equivalent to the population of CD34ϩ cells it is as expected that the correlation between HPC levels in ϩ 0 the peripheral blood and the yield of CD34 cells is not as ϩ 0 20 40 60 80 100 120 140 high as the correlation between CD34 cells in the periph- ϩ HPC cutoff level (x106/m l) eral blood and the yield of CD34 cells. The types of other cells present in the HPC population were not determined Figure 2 The trade-off in sensitivity and specificity. Shown is the trade- from this study. However, despite the inclusion of CD34Ϫ off in sensitivity and specificity (probability, Y-axis) as the cutoff level cells in the HPC population, we found that the correlation of HPC ϫ 106/l (X-axis) is shifted. between the number of peripheral blood HPC and the yield of CD34ϩ cells is better than the correlation between WBC identified with HPC, while a lower proportion of the unsuc- count and the yield of CD34ϩ cells. cessful apheresis collections were identified with HPC The number of HPC predicts the yield of CD34ϩ cells using the higher target CD34ϩ level. collected by apheresis and can be used to time the initiation of collection procedure for the patients treated with chemo- therapy-based mobilization regimens. By analyzing the pre- Discussion dictive values of HPC and sensitivity/specificity of HPC, we found that not only is the correlation of the number of The timing of apheresis is a critical issue for the efficient HPC in the peripheral blood with the yield of CD34ϩ cells and cost effective collection of sufficient peripheral blood better than for WBC levels, but also quantified the degree stem cells for transplantation. A dose of at least 2.5–5.0 to which higher HPC levels in the peripheral blood is asso- ϫ 106/kg CD34ϩ cells is recommended for reliably rapid ciated with a greater probability of reaching the target of hematological recovery, which may take one to more than CD34ϩ cell yield. With an HPC level above 50 ϫ 106/l, eight apheresis collections to achieve.1–4 For donors and 93% of the collections reached the target yield of 1 ϫ 106 patients treated with hematopoietic cytokines only, apher- CD34ϩ cells/kg. However, the negative predictive value of esis can be performed on the fourth to sixth days of cyto- HPC enumeration indicates less clinical utility. If apheresis kine administration on which the peak value of CD34ϩ cells were not performed because of an HPC level below 50 ϫ in the peripheral blood reliably occurs.3,5,6 However, for 106/l, almost 60% of components that would otherwise patients treated with a chemotherapy-containing regimen, meet our definition of a minimally successful collection the time to maximal level of CD34ϩ cells in the peripheral would be missed. If the target collection were 2.5 ϫ 106

Bone Marrow Transplantation HPC predicts CD34؉ cell yield JYuet al 1163 CD34ϩ cells/kg, the trade-off would be in the opposite References direction, with only 67% of collections (above a cutoff level 50 ϫ 106/l HPC) reaching the target and 76% of those below that cutoff not reaching the target. Determining the 1 Bensinger WI, Longin K, Appelbaum FR et al. Peripheral sensitivity and specificity of HPC enumeration, which is blood stem cells (PBSCs) collected after recombinant granulo- different from the predictive value of HPC, does not rely cyte colony stimulating factor (rhG-GSF): an analysis of factors correlating with the tempo of engraftment after on the population of the patients who reach the target. transplantation. Br J Haematol 1994; 87: 825–831. ϩ ϫ Among those collections with CD34 cell yield above 1 2 Perez-Simon JA, Caballero MD, Corral M et al. Minimal 6 6 10 /kg, all had HPC levels above 5 ϫ 10 /l, 89% had HPC number of circulating CD34ϩ cells to ensure successful leu- levels у20 ϫ 106/l and 71% had HPC levels у50 ϫ 106/l. kapheresis and engraftment in autologous peripheral blood Among collections with CD34ϩ cell yields below 1 ϫ progenitor cell transplantation. Transfusion 1998; 38: 385– 106/kg, all had HPC levels below 80 ϫ 106/l and 78% had 391. HPC levels below 50 ϫ 106/l. Among those collections 3 Bensinger WI, Singer J, Appelbaum FR et al. Autologous with CD34ϩ cell yield above 2.5 ϫ 106/kg, all had HPC transplantation with peripheral blood mononuclear cells levels above 5 ϫ 106/l, 91% had HPC levels у20 ϫ 106/l, collected after administration of recombinant granulocyte 82% had HPC levels у50 ϫ 106/l and 77% had HPC levels stimulating factor. Blood 1993; 81: 3158–3163. 4 Bender JG, To LB, Williams S, Schwartzberg LS. Defining a у ϫ 6 ϩ 80 10 /l. Among those collections with CD34 cell therapeutic dose of peripheral blood stem cells. J Hematother ϫ 6 yields below 2.5 10 /kg, 91% had HPC levels below 80 1992; 1: 329–341. 6 6 ϫ 10 /l and 59% had HPC levels below 50 ϫ 10 /l. Taken 5 Grigg AP, Roberts AW, Raunow H et al. Optimizing dose together, these data suggest that the HPC level of у50 ϫ and scheduling of G-CSF (granulocyte colony-stimulating 106/l may be a valid determinant for deciding when to factor) for mobilization and collection of peripheral blood pro- initiate apheresis procedures after treating a patient with a genitor cells in normal volunteers. Blood 1995; 86: 4437– chemotherapy-based mobilization regimen. Individual 4445. ϩ 6 Tjonnfjord GE, Steen R, Evensen SA et al. Characterization collection centers using a higher CD34 cell target for ϩ apheresis may prefer to use a higher HPC cutoff level, of CD34 peripheral blood cells from healthy adults mobil- such as у80 ϫ 106/l. ized by recombinant human granulocyte colony-stimulating factor. Blood 1994; 84: 2795–2801. In conclusion, these data suggest that HPC enumeration 7 Stewart AK, Imrie K, Keating A et al. Optimizing the CD34ϩ using the Sysmex technology can be used to determine and CD34ϩthy-1ϩ stem cell content of peripheral blood col- when to initiate apheresis after chemotherapy-based mobil- lections. Exp Hematol 1995; 23: 1619–1627. ization regimens. Patients with high HPC levels in the peri- 8 Demirer T, Buckner CD, Appelbaum FR et al. Rapid pheral blood are also likely to have higher levels of CD34ϩ engraftment after autologous transplantation utilizing marrow cells and require fewer apheresis procedures to achieve the and recombinant granulocyte colony-stimulating factor-mobil- target goal of CD34ϩ cells for transplantation. HPC enu- ized peripheral blood stem cells in patients with acute myelo- meration appears to be valuable as a positive predictor of genous leukemia. Bone Marrow Transplant 1995; 15: 915– initiating apheresis in that a level above 50 ϫ 106/l is asso- 922. ciated with collections meeting our definition of success. 9 Ho AD, Gluck S, Germond C et al. Optimal timing for collec- tions of blood progenitor cells following induction chemo- However, we cannot be as confident, if the HPC level is therapy and granulocyte–macrophage colony-stimulating fac- below a defined value, that the collection will fail to meet ϩ tor for autologous transplantation in advanced breast cancer. our target goal of CD34 cells. For this study, we set a low Leukemia 1993; 7: 1738–1746. ϩ target value of CD34 cells to be collected in the apheresis 10 Elliott C, Samson DM, Armitage S et al. When to harvest procedure as our definition of success. Therefore, a delay peripheral-blood stem cells after mobilization therapy: predic- in initiating collection when the level of CD34ϩ cells is tion of CD34-positive cell yield by preceding day CD34-posi- presumably increasing may improve subsequent collections tive cell concentration in peripheral blood. J Clin Oncol 1996; and there will be little risk to the patients from a 1 or 2 14: 970–973. 11 Schots R, Riet IV, Damiaens S et al. The absolute number of day delay. The enumeration of HPC may be a valuable tool ϩ for laboratories and apheresis centers that have access to circulating CD34 cells predicts the number of hematopoietic stem cells that can be collected by apheresis. Bone Marrow this hematology analyzer. Transplant 1996; 17: 509–515. 12 Passos-Coelho JL, Braine HG, Davis JM et al. Predictive fac- tors for peripheral-blood progenitor-cell collections using a single large-volume leukapheresis after cyclophosphamide and Acknowledgements granulocyte–macrophage colony-stimulating factor mobiliz- ation. J Clin Oncol 1995; 13: 705–714. 13 Yu J, Leisenring W, Rowley SD et al. The predictive value of We thank the staff of the Apheresis Unit, the Hematology Labora- white cell or CD34ϩ cell count in peripheral blood for timing tory and the Cryobiology Laboratory of the Fred Hutchinson Can- apheresis and maximizing yield. Transfusion 1999; 39: 442– cer Research Center for performing the apheresis procedures, 450. HPC and other blood cell counts and CD34ϩ cell counts. We also 14 Takekawa K, Yamane T, Suzuki K et al. Identification of thank Sysmex Corporation of America for providing access to the hematopoietic stem cells by the SE-9000TM automated hema- analyzer and the supplies needed to conduct this work. Support tology analyzer in peripheral blood stem cell harvest samples. for this project was provided in part by grant numbers CA18029, Acta Hematologica 1997; 98: 54–55. CA15704, and DK56465 from the National Institutes of Health, 15 Takekawa K, Yamane T, Suzuki K et al. Identification of Bethesda, MD, and by the Sysmex Corporation. hematopoietic stem cells by the SE-9000TM automated hema-

Bone Marrow Transplantation HPC predicts CD34؉ cell yield JYuet al 1164 tology analyzer in peripheral blood stem cell harvesting sam- 20 Leisenring W, Pepe MS, Longton G. A marginal regression ple. Blood 1996; 88: 250b. modeling framework for evaluation of medical diagnostic 16 Yamane T, Takekawa K, Tatsumi N. Possibility of identifi- tests. Stat Med 1997; 16: 1263–1281. cation of hematopoietic stem cells using a conventional blood 21 Leisenring W, Alonzo T, Pepe MS. Comparisons of predictive cell counter. Eur J Haematol 1995; 55: 207–208. values of binary medical diagnostic tests for paired designs. 17 Bensinger WI, Appelbaum FR, Rowley S et al. Factors that Biometrics (in press). influence collection and engraftment of autologous peripheral- 22 Haas R, Mohle R, Fruhauf S et al. Patient characteristics asso- blood stem cells. J Clin Oncol 1995; 13: 2547–2555. ciated with successful mobilizing and autografting of peri- 18 Conover WJ. Practical Non-parametric Statistics. Wiley: New pheral blood progenitor cells in malignant lymphoma. Blood York, 1980. 1994; 83: 3787–3794. 19 Weisberg S. Applied Linear Regression. Wiley: New York, 1985.

Bone Marrow Transplantation