Non-myeloablative allogeneic haematopoietic cell transplantation for relapsed diffuse large B-cell lymphoma: a multicentre experience


  • Presented in part at the 49th annual meeting of the American Society of Hematology, December 8–11, 2007, Atlanta, GA.

David G. Maloney MD, PhD, Fred Hutchinson Cancer Research Center, 1100 Fairview Ave N, MS D1-100, Seattle, WA 98109, USA. E-mail:


Patients with relapsed diffuse large B-cell lymphoma (DLBCL) who have failed or are ineligible for autologous haematopoietic cell transplantation (HCT) have a poor prognosis. We examined the outcomes of non-myeloablative allogeneic HCT in this setting. Thirty-one patients with DLBCL and one patient with Burkitt lymphoma received allogeneic HCT following 2 Gy total body irradiation with or without fludarabine. Median age was 52 years. Twenty-four patients (75%) had undergone prior autologous HCT. Disease status at HCT was complete response (14/32, 44%), partial response (9/32, 28%), or refractory (9/32, 28%). Cumulative incidences of acute graft-versus-host disease (GVHD) grades II–IV, grades III–IV, and chronic GVHD were 53%, 19%, and 47% respectively. With a median follow-up of 45 months, 3-year estimated overall (OS) and progression-free survival (PFS) was 45% and 35% respectively. Three-year cumulative incidences of relapse and non-relapse mortality were 41% and 25% respectively. In multivariate models, chemosensitive disease and receipt of ≥4 lines of treatment before HCT were associated with better OS. Patients with chemosensitive disease had 3-year OS and PFS of 56% and 43% respectively. Non-myeloablative allogeneic HCT can produce long-term disease-free survival in patients with chemosensitive relapsed DLBCL who have failed or are ineligible for autologous HCT.

Diffuse large B-cell lymphoma (DLBCL) is the most common subtype of non-Hodgkin lymphoma (NHL). Combination chemotherapy can produce long-term remissions in 20–80% of patients with DLBCL (The International Non-Hodgkin’s Lymphoma Prognostic Factors Project, 1993), and high-dose therapy with autologous haematopoietic cell transplantation (HCT) can salvage 30–40% of patients with DLBCL who relapse after initial chemotherapy (Philip et al, 1987, 1995; Gribben et al, 1989; Mills et al, 1995; Haioun et al, 2000). However, patients who relapse after, or are ineligible for, autologous HCT have a poor prognosis with few effective treatment options and a median survival of 3 months (Petersen et al, 1990; Vose et al, 1992).

Myeloablative allogeneic HCT can provide better disease control than autologous HCT due to immunological graft-versus-lymphoma (GVL) effects and the absence of tumour contamination in the graft (Chopra et al, 1992). However, myeloablative allografting for DLBCL is associated with high treatment-related mortality, particularly in patients who have failed autologous HCT (Chopra et al, 1992; Ratanatharathorn et al, 1994; de Lima et al, 1997; Tsai et al, 1997; Dhedin et al, 1999; Radich et al, 2000; Peniket et al, 2003; Doocey et al, 2005; Law et al, 2006). Additionally, myeloablative HCT is generally restricted to younger and healthier patients, while the average age at diagnosis with DLBCL is 64 years (The Non-Hodgkin’s Lymphoma Classification Project, 1997). Thus, many patients who might benefit from allogeneic HCT are ineligible for myeloablative conditioning.

Non-myeloablative conditioning regimens have permitted expansion of allogeneic HCT to patients who are ineligible for intensive conditioning. Several such regimens have been studied in small cohorts of patients with DLBCL (Branson et al, 2002; Robinson et al, 2002; Escalon et al, 2004; Faulkner et al, 2004; Morris et al, 2004; Armand et al, 2008). Here, we report a multicentre experience with non-myeloablative allogeneic HCT in patients with relapsed DLBCL.

Patients and methods

Eligibility criteria

This analysis included all patients with de novo (untransformed) aggressive or highly aggressive B-cell NHL who underwent allogeneic HCT after non-myeloablative conditioning on Fred Hutchinson Cancer Research Center (FHCRC) multi-institutional protocols between 6 December 1999 and 16 August 2006. Patients were treated at 11 centres, with the FHCRC acting as the coordinating centre. Protocols were approved by the institutional review boards of the FHCRC and collaborating centres. All patients signed informed consent forms approved by the local institutional review boards.

Patients referred to participating centres for consideration of allogeneic HCT for aggressive B-cell NHL were screened using the following criteria; final decisions regarding patient eligibility were made by the treating physicians. Included were patients with aggressive or highly aggressive B-cell NHL whose disease had relapsed after one or more first-line treatments and who were ineligible for high-dose therapy with autologous HCT due to prior autologous HCT or comorbidities. Exclusion criteria were: pregnancy, cardiac ejection fraction <30%, pulmonary diffusion capacity <35% of predicted, decompensated liver disease, Karnofsky performance status <50%, human immunodeficiency virus infection or rapidly progressive bulky lymphoma unresponsive to cytoreductive therapy. Patients with T-cell lymphoma or histological transformation from indolent NHL were excluded from this analysis, as these patients have been analysed and reported elsewhere (Rezvani et al, 2008).

Pretransplant characteristics

Chemotherapy-sensitive disease was defined by partial response (PR) or complete response (CR), according to standard criteria (Cheson et al, 1999), to the last treatment regimen given before allogeneic HCT. Prognosis and comorbidities were assessed retrospectively using the International Prognostic Index (IPI) (The International Non-Hodgkin’s Lymphoma Prognostic Factors Project, 1993) and HCT-Comorbidity Index (HCT-CI) (Sorror et al, 2005). Patients and their donors were matched for human leucocyte antigen (HLA)-A, -B, and -C by at least intermediate-resolution DNA typing, and -DRB1 and -DQB1 by high-resolution techniques (Petersdorf et al, 1998).

Conditioning regimen and postgrafting immunosuppression

Patients were conditioned with 2 Gy total body irradiation (TBI) on day 0, with or without the addition of fludarabine 30 mg/m2/d on days −4, −3, and −2. Postgrafting immunosuppression consisted of cyclosporine or tacrolimus combined with mycophenolate mofetil, as described previously (Maloney et al, 2003; Maris et al, 2003; Baron et al, 2005). Supportive care, including antimicrobial and cytomegalovirus prophylaxis, was administered as described previously (Maris et al, 2003). Haematopoietic growth factors were administered to the recipient only for persistent neutropenia after day +21.

Monitoring after HCT

Patients underwent bone marrow aspiration and peripheral blood cell sorting on days +28, +56, and +84 after HCT to assess chimerism. Unilateral bone marrow biopsy was obtained on day +84 to assess for lymphoma, and yearly thereafter. Patients underwent computed tomography (CT) scans of the chest, abdomen, and pelvis on day +56 after HCT only if abnormal pretransplant, while all patients underwent CT imaging at 3, 6, 12, 18 and 24 months after HCT and annually thereafter through 5 years after HCT. Responses were assessed according to standard criteria (Cheson et al, 1999). Toxicities occurring within the first 100 d after HCT were scored according to the National Cancer Institute Common Toxicity Criteria. Graft-versus-host disease (GVHD) was scored and treated as previously described (Sullivan et al, 1991; Przepiorka et al, 1995).

Long-term follow-up

Performance status was assessed prospectively, using the Karnofsky scale, by patient questionnaires or by patients’ primary oncologists. Dates of discontinuation of immunosuppression by surviving patients were obtained retrospectively through chart review and contact with participating centres.

Statistical analysis

Probabilities of overall and progression-free survival (PFS) were estimated by the Kaplan–Meier method. Any death occurring in the absence of documented disease progression was considered non-relapse mortality (NRM). Death without relapse was considered a competing risk for relapse, and relapse was treated as a competing risk for NRM. The association of various factors with the hazards of failure for the time-to-event endpoints overall and PFS, NRM, relapse, and chronic GVHD were estimated using Cox regression, while logistic regression was used for acute GVHD. Due to the limited number of events, we allowed only two parameters to be estimated in any given multivariable regression model. Factors included in such models were chosen from those statistically significantly at the P ≤ 0·05 level or suggestively associated with outcome in univariate models. All two-sided P-values from regression models were obtained from the likelihood ratio test, where the model with the factor of interest was compared to the model without it. No adjustments were made for multiple comparisons.

Patient characteristics

Twenty-one patients underwent HCT from HLA-identical related donors and 11 from unrelated donors; of the latter, eight were HLA-matched and three were mismatched at one HLA class I allele (Table I). Thirty-one patients had DLBCL while one patient had Burkitt lymphoma. The median age at HCT was 52 (range 18–67) years. The median time from diagnosis to HCT was 3·4 (range 0·6–7·9) years. Patients had received a median of four lines of treatment before allogeneic HCT. Twenty-four of 32 patients (75%) had undergone previous high-dose therapy with autologous HCT; of these, 18 had failed autologous HCT while six were treated on tandem autologous/non-myeloablative allogeneic transplant protocols designed for patients at high risk of disease relapse. Twenty-three patients (72%) had chemosensitive disease at HCT, as demonstrated by CR (n = 14) or PR (n = 9).

Table I.   Patient characteristics (n = 32).
CharacteristicNo. of patients
  1. HCT, haematopoietic cell transplantation; IPI, International Prognostic Index; HCT-CI, haematopoietic cell transplantation comorbidity index; HLA, human leucocyte antigen; TBI, total body irradiation; G-PBMC, granulocyte colony-stimulating factor-mobilized peripheral blood mononuclear cells; CR, complete response; PR, partial response.

 Male21 (66%)
 Female11 (34%)
Age at HCT
 Median, years (range)52 (18–67)
 Age <50 years15 (47%)
 Age ≥50 years17 (53%)
Time from diagnosis to HCT
 Median, years (range)3·4 (0·6–7·9)
Prior treatment
 Median lines of treatment (range)4 (2–7)
 Autologous HCT24 (75%)
  Failed18 (56%)
  Tandem6 (19%)
Prognostic factors
 IPI at HCT, median (range)1 (0–3)
 HCT-CI 0–218 (56%)
 HCT-CI ≥313 (41%)
 HCT-CI unavailable1
 Related21 (66%)
 Unrelated11 (34%)
  1-allele HLA-mismatched3
Conditioning regimen
 2 Gy TBI
  Alone3 (9%)
  Plus fludarabine29 (91%)
Stem cell source
 G-PBMC31 (97%)
 Marrow1 (3%)
Median cell doses
 CD3+3·11 × 108 cells/kg
 CD34+6·65 × 106 cells/kg
Marrow involvement at HCT
 Yes2 (6%)
 No30 (94%)
Disease status at HCT
 CR14 (44%)
 PR9 (28%)
 Relapsed/refractory9 (28%)
Response to last regimen
 Chemosensitive23 (72%)
 Chemorefractory9 (28%)

Transplant details

Three patients with HLA-identical sibling donors were conditioned with 2 Gy TBI alone, while the remaining 29 patients (91%) received 2 Gy TBI and fludarabine as described above. Thirty-one patients received granulocyte colony-stimulating factor-mobilized peripheral blood mononuclear cell allografts, while one received a marrow graft. Median doses of 6·65 × 106 CD34+ cells/kg and 3·11 × 108 CD3+ cells/kg were administered.



All patients had rapid and sustained engraftment. Most patients experienced transient neutropenia (defined as an absolute neutrophil count of <0·5 × 109 cells/l); the median duration of neutropenia was 6 (range 0–27) days, and the median neutrophil nadir was 0·195 × 109 cells/l. Data on platelet and packed red blood cell (RBC) transfusion were available for 22 of 32 patients. Sixteen of these 22 patients (73%) required RBC transfusions; in these patients, the median number of units transfused was 5 (range 1–16). Nine of 22 patients (41%) developed thrombocytopenia (defined as platelet count <20 × 109/l); in these nine patients, the median duration of thrombocytopenia was 2 (range 0–8) days.

GVHD and toxicity

All patients were assessable for acute GVHD. The probabilities of acute GVHD grades II–IV and III–IV were 53% and 19% respectively. Rates of acute GVHD grades II–IV were similar in patients with related versus unrelated donors [52% vs. 55%, Odds Ratio 1·09 (0·25–4·71), P = 0·91]. Extensive chronic GVHD developed in 44% of patients. No statistically significant risk factors for acute GVHD grades II–IV or for chronic GVHD were identified in this patient cohort. Grade IV toxicities were uncommon and consisted mainly of haematological and pulmonary toxicities.

Disease response

The median follow-up of surviving patients was 45 (range 6–70) months. Of the 14 patients in CR at the time of HCT, 10 (71%) remained in CR while four (29%) relapsed. In the 18 patients with measurable disease at the time of HCT, six (33%) achieved CR and one had a PR, for an overall response rate of 39%. Of the six patients undergoing tandem autologous/allogeneic HCT, three died of progressive disease while the remaining three are alive in CR at 6, 18, and 47 months after HCT. The single patient with Burkitt lymphoma died of progressive disease. Overall, the cumulative incidence of disease progression at 3 years was 41% (Fig 1).

Figure 1.

 Overall and progression-free survivals and relapse in all patients (n = 32).

Timing of relapse, GVHD and GVL effects

In patients who relapsed, the median time from HCT to disease progression was 77 (range 21–378) days. For patients with measurable disease who entered CR after HCT, the median time to achieve CR was 61 (range 28–192) days. The median times to development of acute and chronic GVHD were 41 and 152 d respectively.

Survival and non-relapse mortality

For all treated patients, the 3-year estimated overall (OS) and progression-free (PFS) survivals were 45% and 35% respectively (Fig 1). The 3-year cumulative incidence of NRM was 25%. Causes of death are listed in Table II; of the eight non-relapse deaths, three (38%) were due to non-transplant related causes (cardiovascular disease or metastatic colorectal cancer).

Table II.   Causes of death (n = 17).
CauseNo. of patients
  1. GVHD, graft-versus-host disease.

Progressive disease9
GVHD and infection2
Idiopathic pneumonitis1
Second malignancy1 (colorectal cancer)

Univariate analysis identified several risk factors for relapse, NRM, and overall survival (Table III). Patients with chemosensitive disease were at lower risk of mortality [Hazard Ratio (HR), 0·17 (0·06–0·47); P = 0·0005], as were patients who had received four or more lines of treatment before allogeneic HCT [HR, 0·29 (0·11–0·80); P = 0·02] (Fig 2). The dichotomization at four lines of treatment was based partly on the fact that four was the median and partly on the data, so this categorization was partially data-driven. Increase in secondary IPI at HCT led to an increased risk of mortality [HR, 1·83 for each increase of one in IPI (1·15–2·93); = 0·01]. When two of these three factors were examined in multivariable models, the magnitude of the association for IPI diminished after adjustment for chemosensitivity, but IPI remained significantly associated with increased mortality after adjustment for number of lines of prior treatment (categorized as four or more versus fewer). Both chemosensitivity and number of lines of prior treatment remained statistically significant even after adjustment for the other (Table IV).

Table III.   Risk factors for mortality, relapse, and non-relapse mortality in univariate analysis.*
OutcomeVariableHR (95% CI)P
  1. *Also tested were donor type (related versus unrelated) and prior autologous haematopoietic cell transplantation.

  2. †Modelled as a continuous linear variable. HR represents increase in hazard with each increase of one line of treatment, one point in IPI, or 10 years of age.

  3. HR, hazard ratio; CI, confidence interval; HCT, haematopoietic cell transplantation; IPI, International Prognostic Index; NRM, non-relapse mortality.

MortalityLines of treatment before HCT†0·71 (0·47–1·07)0·10
≥4 lines of treatment before HCT0·29 (0·11–0·80)0·02
Chemosensitive disease0·17 (0·06–0·47)0·001
IPI†1·83 (1·15–2·93)0·01
RelapseAge†0·68 (0·45–1·07)0·08
Chemosensitive disease0·20 (0·07–0·63)0·009
NRMLines of treatment before HCT†0·61 (0·31–1·22)0·14
≥4 lines of treatment before HCT0·05 (0·005–0·46)0·002
Chemosensitive disease0·23 (0·05–1·02)0·07
Figure 2.

 Overall and progression-free survivals stratified by chemosensitivity at transplant (Panels A and B) and number of lines of chemotherapy received before transplant (Panels C and D).

Table IV.   Risk factors for mortality and relapse in multivariate analysis.
OutcomeFactors testedHR (95% CI)P
  1. *Modelled as a continuous variable. HR represents increase in hazard with each increase of one point in IPI or 10 years of age.

  2. HR, hazard ratio; CI, confidence interval; IPI, International Prognostic Index.

MortalityIPI*1·37 (0·73–2·57)0·34
Chemosensitive0·25 (0·08–0·84)0·02
Mortality1–3 lines prior treatment1
≥4 lines prior treatment0·18 (0·06–0·56)0·004
Chemosensitive0·12 (0·04–0·35)0·0002
MortalityIPI*1·98 (1·21–3·22)0·007
1–3 lines prior treatment1
≥4 lines prior treatment0·25 (0·09–0·71)0·01
RelapseAge*0·74 (0·49–1·12)0·16
Chemosensitive0·23 (0·07–0·72)0·02

Chemosensitive disease was also associated with a suggestively lower risk of NRM [HR, 0·23 (0·05–1·02); P = 0·07] in univariate analysis, and four or more lines of prior treatment was associated with a lower risk of NRM [HR = 0·05 (0·005–0·46); P = 0·002]. Given the small number of NRM events, it was unclear whether each of these factors was associated with NRM after adjusting for the other. Chemosensitive disease was also associated with a reduced risk of relapse [HR = 0·20 (0·07–0·63); P = 0·009], and increased age was suggestively associated with decreased relapse [HR = 0·68 for each increase of 10 years in age (0·45–1·07); P = 0·08] in univariate models. Inclusion of each of these factors in a single model for relapse led to similar associations as those seen in univariate models (Table IV).

Factors that did not have a statistically significant effect on relapse, NRM, or overall mortality in univariate analysis in this cohort included: prior autologous HCT (none versus failed versus tandem), HCT-CI score, and related versus unrelated donor.

Donor lymphocyte infusion and postrelapse treatment

One patient with relapsed disease underwent three donor lymphocyte infusions (DLI) without effect, and died of progressive disease. Another patient with disease progression at 119 d after HCT received a second allogeneic HCT with myeloablative conditioning and died of multi-organ failure shortly after the second HCT.

Four patients with disease progression after HCT were alive at last follow-up. One of these patients had disease progression 58 d after HCT. Her immunosuppression was rapidly tapered and she was treated with rituximab and local irradiation. Her disease stabilized clinically, although a positron emission tomography (PET) scan at 1 year after HCT showed residual disease activity. The patient declined biopsy or DLI and was followed clinically without further therapy; a repeat PET scan at 3 years after HCT showed disappearance of residual disease, and the patient was alive in CR at last follow-up, 54 months after HCT. Another patient with progressive disease received a second non-myeloablative allogeneic HCT from the same HLA-identical sibling donor; this patient achieved a CR after the second HCT and remains alive and disease-free at last follow-up, 34 months after the second HCT. The third patient progressed 121 d after HCT from an HLA-identical sibling donor. Her immunosuppression was rapidly tapered with a transient response; she subsequently received DLI at 7 months after HCT, which was the time of last follow-up. The final patient developed disease progression 29 d after HCT; her immunosuppression was rapidly tapered, but she continued to have progressive disease at last follow-up, 6 months after HCT.

Long-term outcomes

Of the 15 patients surviving at the time of last follow-up, seven (47%) remained on some form of systemic immunosuppressive medication (IS) while the remaining eight patients (53%) discontinued all systemic immunosuppression. The median time to discontinuation of all systemic IS in disease-free survivors was 16 months (range 6–57 months). Data on performance status at last follow-up were available for 12 of 15 surviving patients at a median of 3·8 years after HCT. In these patients, the median Karnofsky performance status at last follow-up was 95 (range 50–100), and only three of 12 had Karnofsky scores below 90.


Patients with DLBCL who relapse after, or are ineligible for, autologous HCT have a poor prognosis with conventional therapy. This study provides evidence that non-myeloablative allogeneic HCT is an effective salvage therapy which can produce long-term disease-free survival in a subset of these patients. Given that virtually none of these patients would be expected to achieve durable disease-free survival with conventional therapy, the rate of PFS seen after non-myeloablative allogeneic HCT is encouraging and provides proof of the principle that immunological GVL effects alone can control DLBCL in some cases. Most relapses occurred within the first 3 months after HCT, and we observed a plateau in the relapse curve at 1 year after HCT, suggesting that the GVL effect is durable where it is present. While late deaths did occur, these were due to metastatic colon cancer and myocardial infarction in patients who were disease-free, and reflect the age and general risk factors of this population rather than disease relapse or specific complications of non-myeloablative HCT.

Previous studies of reduced-intensity or non-myeloablative HCT for NHL have generally included small numbers of patients with DLBCL. Many of these studies used fludarabine-based conditioning with the addition of an alkylating agent and in some cases alemtuzumab. Progression-free survival at 2 or 3 years in these studies has generally ranged from 13% to 34% for patients with DLBCL (Robinson et al, 2002; Faulkner et al, 2004; Morris et al, 2004; Dean et al, 2005; Armand et al, 2008), though a 2004 single-institution report from the M.D. Anderson group described nine of 10 patients with DLBCL alive at a median of 23 months after reduced-intensity allogeneic HCT (Escalon et al, 2004). While differences in patient selection and characteristics make direct comparison between published cohorts difficult or impossible, some generalized findings have emerged and are confirmed in this relatively large multicentre cohort of patients with DLBCL. Relapses in these patients tend to occur early after HCT, reflecting both the aggressiveness of the underlying disease and the window of weeks to months necessary for robust GVL effects to become clinically evident. In particular, chemorefractory disease at HCT appears to be a common and logical risk factor for relapse and mortality.

Patient selection is clearly an important factor in applying non-myeloablative conditioning to an aggressive malignancy such as DLBCL. Patients with rapidly growing bulky lymphoma were excluded from our protocols as unlikely to benefit, and those who were transplanted with chemorefractory disease had poor outcomes. Given our findings and those of other groups discussed above, non-myeloablative allogeneic HCT is best suited for patients whose relapsed DLBCL remains at least temporarily and partially chemosensitive. For patients with rapidly progressive or chemorefractory relapsed DLBCL, other investigational cytoreductive approaches, such as radioimmunoconjugate therapy, may be preferable before attempting allogeneic HCT.

Our analysis included only patients with de novo (untransformed) aggressive B-cell lymphoma. We recently reported on an additional 16 patients with transformed NHL who underwent non-myeloablative allogeneic HCT using identical protocols with a 3-year PFS of 21% (Rezvani et al, 2008). The 3-year cumulative incidences of relapse in the de novo and transformed groups were similar (41% and 38% respectively); the somewhat poorer survival in the transformed cohort was a result of higher NRM, and might be due to heavier and lengthier pretreatment received during the indolent phase of the disease. Our cohort also includes six patients who received tandem autologous-allogeneic HCT. Three of these six patients were alive and disease-free at last follow-up. In these three patients, it is possible that long-term disease-free survival is a consequence of high-dose therapy and autologous HCT rather than the allogeneic GVL effect.

The observed association between ≥4 lines of treatment before allogeneic HCT and improved survival with decreased NRM is somewhat counterintuitive. We explored confounding factors, such as the possibility that patients who received more therapy before HCT might consequently have had less disease at the time of HCT, or might have had fewer comorbidities and thus been treated more aggressively before HCT. However, we were unable to find significant interactions in this cohort. The association may indicate that receipt of <4 lines of chemotherapy is a marker for more aggressive or refractory disease where conventional therapy was ineffective and patients were referred more quickly for allogeneic HCT. Additionally, the choice of four lines as a cut-off point was partially data-driven, and the association weakened significantly when lines of treatment was treated as a continuous variable, so the finding should be interpreted cautiously in that light. Given conflicting findings in other published cohorts and the lack of a compelling explanation for this finding, allogeneic HCT should not be delayed solely on the basis of this association.

Given the concern over the potential long-term complications of allogeneic HCT, particularly chronic GVHD, it is encouraging that the majority of surviving patients were able to discontinue immunosuppressive medication, and that disabling chronic GVHD was very rare. While the Karnofsky performance status is, at best, an indirect measure of quality of life, the median Karnofsky status at last follow-up of surviving patients of 95% indicates that this approach is associated not merely with survival but with recovery of meaningful functional status.

Our results provide evidence that non-myeloablative allogeneic HCT can produce durable immunologically mediated disease-free survival with excellent performance status in patients with relapsed DLBCL who otherwise have a dismal prognosis. However, patient selection is a vital factor. Patients whose disease can be temporarily controlled or cytoreduced are most likely to benefit, while those with rapidly progressive chemorefractory disease are unlikely to benefit and may be more appropriately treated with other investigational approaches. Efforts to safely intensify this conditioning regimen, for example through the addition of rituximab, may further reduce the relapse rate and extend these benefits to a wider range of patients.


Jennifer Freese, Heather Hildebrant, Gresford Thomas, and Courtney McNamara assisted with data retrieval, and Bonnie Larson and Helen Crawford assisted with manuscript preparation. We also thank the patients who participated in these protocols and the nurses, physicians, staff, and families who cared for them.

Supported by grants CA78902, CA18029, CA15704, and HL088021 from the National Institutes of Health, Bethesda, MD, USA.