Professional Documents
Culture Documents
AC Vigorito, WM Azevedo, JFC Marques, AM Azevedo, KAB Eid, FJP Aranha, I Lorand-Metze,
GB Oliveira, MEP Correa, ARC Reis, ECM Miranda and CA de Souza
Bone Marrow Transplantation Unit, State University of Campinas, SP, Brazil
Eligibility criteria
Bone marrow (BM) has been the standard source of cells This study was a prospective randomised, phase III trial
to rescue lympho-haematopoiesis after myeloablative ther- performed in a single institution. It began in February 1995
apy but recently allogeneic peripheral blood progenitor and closed in October 1997. All patients had at least 100
cells (PBPC) have been replacing bone marrow for autolog- days of follow-up after transplantation. Eligibility criteria
ous transplants.1–4 Moreover, pilot and retrospective studies for entry into the study were age between 10 and 60 years,
suggest that HLA-identical PBPC can also restore haemo- haematological malignancy as primary disease and HLA-
poiesis without an increase in the incidence or severity of identical sibling donors. A detailed explanation about the
acute graft-versus-host-disease (a-GVHD).5–9 PBPC mobil- procedure and its potential complications were given to
patients and donors. Informed consent was obtained using
Correspondence: CA de Souza, Hemocentro/Unicamp, PO Box 6198, protocols and forms according to our institution regulations.
Barão Geraldo, 13081–970, Campinas, SP, Brasil Patients, donors and treatment characteristics are shown
Received 22 May 1998; accepted 30 July 1998 in Table 1.
ABMT and PBPCT in haematological malignancies
AC Vigorito et al
1146
Table 1 Patient, donor and treatment characteristics Evaluations and definitions
PBPC BM PBPC and BM were analysed for CD34+ cells and T cell
subsets by flow cytometry as previously described.18 Mono-
Patients (No.) 18 19 clonal antibodies were obtained from Becton Dickinson
Age in years, median (range) Immunocytometry Systems (San Jose, CA, USA).
Patients 29.5 (9–51) 35 (17–56) CD34-positive cells were quantified using a modification
Donors 30 (14–60) 32 (12–60) of the method described by Sutherland et al.18 In this modi-
Patient gender M = 12 M = 15
F=6 F=4 fication, the antibody CD14/FITC was used instead of
Donor gender M→M=4 M→M=7 CD45 to exclude myeloid/monocytic cells contaminating
M→F=1 M→F=2 the CD34/PE-positive population defined as CD14 negative
F→M=8 F→M=8 and presenting a low side scatter.
F→F=5 F→F=2
Early disease T cell subsets were analysed by the whole blood
(CML, 1st CP; AML, 1st CR; AML, 13 13 lyse/wash method using the monoclonal antibodies CD3-
1st rel; ALL, 1st CR; MDS-RA) FITC/CD4-PE, CD3-FITC/CD8-PE and CD4-FITC/CD8-
Advanced disease PE. All data were collected and analysed in the FACscali-
(CML, AP/BC; AML ⬎ 1st rel; 5 6 bur flow cytometry equipment (Becton Dickinson), using
refractory AML, MM, NHL;
ALL ⬎ 2nd; MDS-RAEB) the software Cell Quest.
Alive 10 10 Neutrophil engraftment was defined as the first of 2 con-
Dead 8 9 secutive days with an absolute neutrophil count (ANC)
Follow-up (days) 335 631 ⬎0.5 × 109/l following a post-transplant nadir. Platelet
(120–1079) (281–1044)
Myeloablative regimens engraftment was defined as the first of 7 consecutive days
Bu (16)/Cy (120) 17 16 with a platelet count ⬎20 × 109/l without platelet trans-
Bu (16)/Cy (120)/VP-16 (40) — 3 fusion. Engraftment was also documented by marrow aspir-
CY (120)/TBI (13.2 Gy) 1 — ation, biopsy and cytogenetics.
GVHD prophylaxis Assessment and grading of a-GVHD and c-GVHD were
CSP/MTX 18 16
CSP/Pred — 3 made using the Glucksberg19 and Shulman criteria,20
respectively. Supportive care was provided in private rooms
M = male; F = female; CML = chronic myeloid leukaemia; CP = chronic equipped with HEPA filters. Irradiated red blood cell and
phase; AML = acute myeloid leukaemia; rel = relapse; ALL = acute lym- platelet transfusions, as well as broad-spectrum antibiotics
phoblastic leukaemia; CR = complete remission; MDS =myelodysplastic together with anti-fungal agents, were given whenever indi-
syndrome; RA = refractory anaemia; AP/BC = accelerated phase/blastic cated. Ceftazidime was started as soon as the neutrophil
crisis; MM = multiple myeloma; NHL = non-Hodgkin’s disease;
RAEB = refractory anaemia with excess of blast; Bu (16) = busulfan (16 count dropped below 0.5 × 109/l and changes were made as
mg/kg); Cy (120) = cyclophosphamide (120 mg/kg); VP-16 = etoposide indicated by cultures and clinical course, keeping the
(40 mg/kg); TBI = total body irradiation; CSP = cyclosporine; patient on antibiotics until the granulocyte count reached
MTX = methotrexate; Pred = prednisone. 1.0 × 109/l. All patients received ganciclovir (5 mg/kg/day
3 times a week) after marrow recovery as cytomegalovirus
(CMV) infection prophylaxis until day 75 post-transplant.
Nutritional support, including total intravenous nutrition,
was provided to assure an adequate nutritional balance.
Donors
All PBPC donors received rhG-CSF (Granulokine; Roche, Statistical methods
SP, Brazil) by subcutaneous injection (10 g/kg daily for
5 consecutive days). Apheresis was performed on day 5 of Analysis was based on status of the patients on 1 February
G-CSF administration. Day 5 rhG-CSF donor adminis- 1998. Our main objective was to compare overall survival
tration was timed to coincide with day 0 of the allograft and then compare objective response rates, disease-free sur-
recipient. Leukaphereses were performed by a Dideco Viv- vival and toxicity. Overall survival (OS) in both groups was
acell (Mod. BT-798 CE/A; Mirandola, Italy) continuous measured from the date of randomisation to the date of
flow blood cell separator using a standard leukocyte apher- death or last follow-up evaluation. Disease-free survival
esis program. The collection usually lasted 3–5 h and a only applies to patients who achieved a complete remission.
vein-to-vein technique was used to avoid use of an indwell- Duration was calculated from the time of CR assessment to
ing catheter. After collection, the apheresis product was the date of relapse or last follow-up evaluation confirming
infused into recipients immediately. A goal of this study freedom from disease. All data were analysed with descrip-
was to collect and infuse a minimum of 2 × 108 of total tive statistical methods. Probability of events (neutrophil
mononuclear cells (TMNC).16 and platelet recovery, a-GVHD, c-GVHD) and survival
Bone marrow was obtained from donors by standard curves were analysed using Kaplan–Meier21 product limit
methods.17 Day 0 for the recipient was defined as the day estimates, and groups were compared using the log-rank
of BM or PBPC infusion. test or Breslow’s test with SPSS software version 7.5 for
A summary of pre-treatment preparative regimens and Windows 95. Proportions of patients within each group of
GVHD prophylaxis is shown in Table 1. No growth factors characteristics and outcome for patients receiving PBPC or
were used after transplantation. BM were compared by the Fisher’s test or 2 test, when
ABMT and PBPCT in haematological malignancies
AC Vigorito et al
1147
appropriate. Comparisons of continuous variables were also Table 2 Graft characteristics
performed with the Mann–Whitney test, considering all P
values ⬍0.05 significant. PBPC (M) BM (M) P
All donors received the proposed dose of rhG-CSF. Leu- Day ANC ⬎0.5 × 109/l
kapheresis procedures and rhG-CSF were generally well Median (range) +16 (11–25) +18 (13–30) 0.15a
tolerated. The median total blood processed was 12.0 l Day platelets ⬍20 × 109/l
(9.4–28.5). The median white blood cell count on day 5 of Median (range) +12 (9–36) +17 (10–40) 0.01a
rhG-CSF was 31.7 × 109/l (12.3–62.5). No side-effects with No. of platelets units transfused
Median (range) 12.5 (4–44) 17.5 (1–341) 0.32
rhG-CSF were observed in seven of 18 donors (39%). Mild No. of RBC units transfused
bone pain occurred in eight out of 18 donors (44%) whereas Median (range) 4 (2–12) 5 (1–75) 0.20
two of 18 (11%) experienced headache. All donors showed Days to discharge
a decrease in platelet counts after apheresis. The median Median (range) +21 (18–42) +26 (18–48) 0.08
pre-apheresis platelet count was 239.5 × 109/l (149.0– a
369.0) and post-apheresis it was 119.0 × 109/l (58.0–217.0) Breslow’s test.
ANC = absolute neutrophil count; RBC = red blood cells.
(P ⬍ 0.0001). Seven of 18 donors (39%) had platelet
reductions below 100 × 109/l. However, no bleeding epi-
sodes occurred. There were no relevant changes in hema- reach an ANC of 0.5 × 109/l, the median day to achieve
tocrit. No significant problems were seen during apheresis such a count was 16 (11–25) for recipients of PBPC and
despite the large volume of blood that was processed. 18 (13–30) for recipients of BM. The median period to
achieve a count of ⬎0.5 × 109/l neutrophils was not sig-
Comparability of PBPC and BM groups nificantly shorter for patients receiving PBPC (P = 0.15)
(Figure 1).
As shown in Table 1, PBPC and BM groups were well One patient receiving PBPC developed progression of
matched for diagnosis, age, phase of disease and treatment. his primary disease before reaching platelet transfusion
Recipients of allogeneic PBPC received a median of independence. Two patients in each group died before
5.5 × 108/kg (2.2–21.0) total nucleated cells (TNC); reaching platelet transfusion independence. One patient in
3.2 × 108/kg (2.0–5.4) total mononuclear cells (TMNC); the BM group was not evaluable due to primary
4.7 × 106/kg (1.26–71.6) CD34+ cells; 317.0 × 106/kg engraftment failure. The median day to achieve platelet
(160.6–644.7) CD3+ cells; 154.5 × 106/kg (23.6–12.2) independence was day 12 (9–36) for the PBPC group and
CD3+/CD4+ cells; and 97.3 × 106/kg (45.7–217.5)
CD3+/CD8+ cells. All patients, except two, received a
1.0
single apheresis product.
Recipients of BM received a median of 2.6 × 108/kg
(1.5–3.4) total nucleated cells (TNC); 1.1 × 108/kg (0.6– 0.8
1.9) total mononuclear cells (TMNC); 5.3 × 106/kg (1.2–
BM
17.5) of CD34+ cells; 75.7 × 106/kg (26.4–157.3) CD3+ PBPC
Probability
0.2
Engraftment and transfusion requirements
Parameters of engraftment and transfusion requirements are
0
shown in Table 3. One patient transplanted with PBPC died 0 5 10 15 20 25 30
and one developed disease progression before reaching neu-
Days
trophils ⬎0.5 × 109/l. One patient transplanted with BM
had primary engraftment failure. Among those who did Figure 1 Probability of 0.5 × 109/l granulocytes.
ABMT and PBPCT in haematological malignancies
AC Vigorito et al
1148
1.0 1.0
PBPC
0.8 0.8
BM
Probability
0.6 0.6
Probability
0.4 0.4
PBPC
0.2 0.2 BM
0 0
0 10 20 30 40 0 20 40 60 80 100
Days Days
0.6
PBPC BM P BM
0.4
Acute GVHD evaluable 15 16
Grade 2–4 4/15 (26/6%) 3/16 (18.7%) 0.40a
Grade 3–4 2/4 2/3
0.2
Chronic GVHD evaluable 14 15
Total affected 10/14 (71.4%) 8/15 (53.3%) 0.53b
Extensive 10/10 (100%) 4/8 (50%) 0.02a
Limited 0 4/8 (50%) 0.02a 0
0 1 2 3
Years
a
Fisher’s test.
test.
b 2
Figure 4 Probability of total affected c-GVHD.
ABMT and PBPCT in haematological malignancies
AC Vigorito et al
1149
developed extensive chronic GVHD while four out of eight 1.0
(50%) developed extensive chronic GVHD in the BM
group (P = 0.02). 0.8
Causes of death
Probability
0.6 BM = 51%
As shown in Table 6, five out of eight (62.5%) of the cases
receiving PBPC died of transplant-related complications as PBPC = 47%
0.4
compared to seven out of nine (77.8%) who received BM.
There were two deaths from bacterial infection in the PBPC
group and one in the BM group. One death occurred from 0.2
gastric-intestinal CMV and one from hepatic veno-occlus-
ive disease (VOD). There was one case of haemorrhagic
0
cystitis in the BM group. No deaths due to CMV, VOD or
0 1 2 3
haemorrhage were seen in the PBPC group. One patient in Years
each group died of a-GVHD and two patients in each group
died of c-GVHD. In both groups, all relapsed or pro- Figure 5 Overall survival.
gressing patients died, three out of eight patients (37.5%)
in the group receiving PBPC and two out of nine patients 1.0
(22.2%) in the BM group.
0.8
Relapse
Among patients transplanted with PBPC, two out of 18 PBPC = 58%
Probability
0.6
(11%) patients have relapsed and one out of 18 (5.5%) had
early disease progression. In patients transplanted with BM, BM = 52%
two out of 19 (10.5%) have relapsed. 0.4
Survival 0.2
1150
Collection of PBPC was well tolerated by donors. The counts are higher in PBPC than in marrow recipients during
dose of rhG-CSF was 10 g/kg for 5 days followed by a the first 2 months, thereafter the monocyte counts are simi-
single leukapheresis with a median blood volume processed lar, whereas the T cells counts were higher in the PBPC
of 12 l on day 5. Using this standard procedure we achieved recipients for at least 6 months.30 This mechanism may be
a median of 4.7 × 106/kg CD34+ cells of recipient body associated with the higher incidence of c-GVHD. Chronic
weight, which was lower than BM yield (5.2 × 106/kg of GVHD has been associated with a low incidence of leu-
recipient). Data from the literature suggest the necessity for kaemia relapse.31,32 Therefore, allogeneic PBPC transplan-
at least two leukaphereses to obtain more than tation could also be associated with a lower incidence of
5.0 × 106/kg,6,7,10,22,23 which has been associated with a fas- relapse.14,33 In the current study we could not confirm this
ter engraftment.24 Data using only one leukapheresis and observation. Relapses were few and similar in number in
10 g/kg/day by 5 days of rhG-CSF, have shown a similar both groups, which may reflect the small number of
median number of CD34+ cells (4.6 × 106/kg) compared to patients analysed.
our results.25 These data, however, suggest that our mobilis- Mortality was similar in both groups and the majority of
ation protocol was adequate, as engraftment in the PBPC deaths were procedure-related. Also, the higher frequency
group was faster with these PBPC numbers. In addition, of extensive chronic GVHD in the PBPC group correlated
long-term follow-up showed stable engraftment in all neither with death nor with worsening of life quality. It is
patients. However, the number of aphereses can be still unclear whether the benefits of fast engraftment or a
increased if necessary to obtain a higher number of CD34+ possible graft-versus-leukaemia reaction outweigh any
cells. Despite the similarity in CD34+ cell numbers infused adverse effects such as c-GVHD.6,12 In patients at high risk
in the BM and PBPC groups, the different kinetics can be of relapse, developing of c-GVHD could have beneficial
explained by the biological properties of PBPC. In autolog- effects.6
ous transplantation with PBPC where patients were mobil- We believe that allogeneic PBPC transplants should not
ised with cyclophosphamide and rhG-CSF, the PBPC were be used routinely, but only in prospective randomised trials,
insensitive to cell-cycle restraint imposed by contact with until issues regarding c-GVHD and relapse are clarified.
marrow-derived stromal cells, which therefore did not
reduce the proportion of progenitors participating in neutro-
phil production.26 Moreover, filgrastim increases numbers Acknowledgements
of circulating megakaryocyte progenitors in PBPC donors,
and this has also been associated with accelerated platelet We greatly appreciate the help of Francislaine Queiroz Costa for
recovery in recipients of autologous filgrastim-stimulated outstanding secretarial assistance and Eduardo Gasparotto Roveri
PBPC.7 for performing the apheresis procedures.
One of the greatest concerns with allogeneic PBPC trans-
plantation is the possibility of increasing incidence and
severity of GVHD due to the high number of T cells present References
in the graft.8 However, neither experimental nor clinical
data exist to show a correlation between the occurrence of 1 Sheridan WP, Begley CG, Juttner CA et al. Effect of periph-
acute GVHD and T cell number, in either BM8 or PBPC7 eral blood progenitor cells mobilised by filgrastim on platelet
transplants. The low incidence of acute GVHD in PBPC recovery after high-dose chemotherapy. Lancet 1992; 339:
transplant recipients may be attributed to G-CSF-induced 640–644.
preferential differentiation of T helper cells into Th2 rather 2 Gianni AM, Siena S, Bregni M et al. Granulocyte–macro-
Th1 cells, resulting in PBPC grafts with relatively few Th1, phage colony-stimulating factor to harvest circulating haema-
which have been implicated in the pathogenesis of acute topoietic stem cells for autotransplantation. Lancet 1989; 2:
GVHD,27 or the large number of monocytes in PBPC grafts 580–585.
that suppress T cells.28 In our study we did not find any 3 To LB, Roberts MM, Haylock DN et al. Comparison of
difference in the incidence and severity of acute GVHD in haematological recovery times and supportive care require-
ments of autologous phase peripheral blood stem cell trans-
the PBPC group compared to the BM transplant group. The
plants, autologous bone marrow transplants and allogeneic
incidence of chronic GVHD was similar but we observed bone marrow transplants. Bone Marrow Transplant 1992; 9:
a trend to developing total c-GVHD in the PBPC group 255–284.
(P = 0.08). Indeed, comparing the severity of c-GVHD in 4 Bensinger W, Singer J, Appelbaum F et al. Autologous trans-
both groups, all PBPC patients developed extensive chronic plantation with peripheral blood mononuclear cells collected
GVHD compared to 50% of patients in the BM group after administration of recombinant granulocyte colony-stimu-
(P = 0.02). A high incidence and severity of chronic GVHD lating factor. Blood 1993; 81: 3158–3162.
after PBPC transplantation has already been reported by 5 Azevedo WM, Aranha FJ, Gouvea JV et al. Allogeneic trans-
others.12–15 In our study, involvement of various tissues was plantation with blood stem cells mobilised by rhG-CSF for
similar between groups, except for a higher incidence of hematological malignancies. Bone Marrow Transplant 1995;
16: 647–653.
oral mucosal lesions in the PBPC group (P = 0.03). Pre- 6 Bensinger WI, Weaver CH, Appelbaum F et al. Transplan-
vious data from our group suggest that patients with more tation of allogeneic peripheral blood stem cells mobilised by
severe lesions of the labial salivary glands will have shorter recombinant human granulocyte colony-stimulating factor.
survivals;29 data suggest that the polarisation of Th1→Th2 Blood 1995; 85: 1655–1658.
may wane or that the monocytes may die within the first 7 Korbling M, Przepiorka D, Huh Y et al. Allogeneic blood
months after grafting.12 Although monocytes and T cell stem cell transplantation for refractory leukemia and lym-
ABMT and PBPCT in haematological malignancies
AC Vigorito et al
1151
phoma: potential advantages of blood over marrow allografts. 22 Miflin G, Charley C, Stainer C et al. Stem cell mobilization
Blood 1995; 85: 1659–1665. in normal donors for allogeneic transplantation: analysis of
8 Schmitz N, Dreger P, Suttorp M et al. Primary transplantation safety and factors affecting efficacy. Br J Haematol 1996; 95:
of allogeneic peripheral blood progenitor cells mobilised by 345–348.
filgrastim (granulocyte colony-stimulating factor). Blood 23 Miflin G, Russel NH, Hutchinson RM et al. Allogeneic per-
1995; 85: 1666–1672. ipheral blood stem cell transplantation for haematological
9 Bensinger WI, Clift R, Martin P et al. Allogeneic peripheral malignancies – an analysis of kinetics of engraftment and
blood stem cell transplantation in patients with advanced hem- GVHD risk. Bone Marrow Transplant 1997; 19: 9–13.
atologic malignancies: a retrospective comparison with mar- 24 Brown RA, Adkins D, Goodnough LT et al. Factors that
row transplantation. Blood 1996; 88: 2794–2800. influence the collection and engraftment of allogeneic periph-
10 Dreger P, Haferlach T, Eckstein V et al. G-CSF mobilized eral-blood stem cells in patients with hematologic malig-
peripheral blood progenitor cells for allogeneic transplan- nancies. J Clin Oncol 1997; 15: 3067–3074.
tation: safety, kinetics of mobilization and composition of the 25 Grigg AP, Roberts AW, Raunow H et al. Optimizing dose
graft. Br J Haematol 1994; 87: 609–613. and scheduling of filgrastim (granulocyte colony-stimulating
11 Martinez C, Urbano-Ispizua, Rozman C et al. Effects of G- factor) for mobilization and collection of peripheral blood pro-
CSF administration and peripheral blood progenitor cell col- genitor cells in normal volunteers. Blood 1995; 86: 4437–
lection in 20 healthy donors. Ann Hematol 1996; 72: 269–272. 4445.
12 Storek J, Gooley T, Siadak M et al. Allogeneic peripheral 26 Scott MA, Apperley JF, Bloxham DM et al. Biological proper-
blood stem cell transplantation may be associated with a high ties of peripheral blood progenitor cells mobilized by cyclo-
risk of chronic graft-versus-host disease. Blood 1997; 90: phosphamide and granulocyte colony-stimulating factor. Br J
4705–4709. Haematol 1997; 97: 474–480.
13 Urbano-Ispizua A, Garcia-Conde S, Brunet S et al. High inci- 27 Pan L, Delmonte J, Jalonen CK, Ferrara JLM. Pretreatment
dence of chronic GVHD after allogeneic peripheral blood pro- of donor mice with granulocyte colony-stimulating factor pol-
genitor cell transplantation from matched related donors.
arizes donor T lymphocytes toward type-2 cytokine pro-
Blood 1996; 88 (Suppl. 1): 617a (Abstr.).
duction and reduces severity of experimental graft-versus-
14 Korbling M, Mirza N, Thall P et al. 100 HLA-identical allo-
host-disease. Blood 1995; 86: 4422–4429.
geneic blood stem transplantations: The MD Anderson Cancer
28 Ino K, Singh RK, Talmadge JE. Monocytes from mobilized
experience. Bone Marrow Transplant 1997; 19 (Suppl. 1):
stem cells inhibit T cell function. J Leuk Biol 1997; 61:
S72 (Abstr.).
15 Majolino I, Saglio G, Scimè R et al. High incidence of chronic 583–591.
GVHD after primary allogeneic peripheral blood stem cell 29 Correa MEP, Metze K, Azevedo AM et al. Histology of the
transplantation in patients with hematologic malignancies. labial salivary glands as a prognostic factor after allogeneic
Bone Marrow Transplant 1996; 17: 555–560. bone marrow transplantation. Pathol Res Pract 1997; 193:
16 Bender J, To TB, Schwartzberg LS. Defining a therapeutic 414 (Abstr.).
dose of peripheral blood stem cells. J Hematother 1992; 1: 30 Ottinger HD, Beelen DW, Scheulen B et al. Improved immune
329–342. reconstitution after allotransplantation of peripheral blood
17 Thomas ED. Allogeneic marrow transplantation using stem cells instead of bone marrow. Blood 1996; 88: 2775–
matched sibling donors. Transplant Proc 1974; 6: 115–119. 2779.
18 Sutherland DR, Keating A, Nayar R et al. Sensitive detection 31 Sullivan KM, Weiden PL, Storb R et al. Influence of acute
and enumeration of CD34+ cells in peripheral and cord blood and chronic graft-versus-host disease on relapse and survival
by flow cytometry. Exp Hematol 1994; 22: 1003–1110. after bone marrow transplantation from HLA-identical sib-
19 Glucksberg H, Storb R, Fefer A. Clinical manifestation of lings as treatment of acute and chronic leukemia. Blood 1989;
graft-versus-host-disease in human recipients of marrow from 73: 1720–1728.
HLA matched sibling donors. Transplantation 1974; 18: 32 Slavin S, Akerstein A, Naparstek E et al. The graft-versus-
295–304. leukemia phenomenon: is GVL separable from GVHD? Bone
20 Shulman HM, Sullivan KM, Weiden PL. Chronic graft-ver- Marrow Transplant 1990; 6: 155–161.
sus-host syndrome in man. A long-term clinicopathologic 33 Glass B, Uharek L, Zeis M et al. Allogeneic peripheral blood
study of 20 Seattle patients. Am J Med 1980; 69: 204–217. progenitor cell transplantation in a murine model: evidence
21 Kaplan ED, Meier P. Non parametric estimation from incom- for an improved graft-versus-leukemia effect. Blood 1997; 90:
plete observations. J Am Stat Assoc 1958; 53: 457–481. 1694–1700.