Hematopoietic cell transplantation-comorbidity index and Karnofsky performance status are independent predictors of morbidity and mortality after allogeneic nonmyeloablative hematopoietic cell transplantation

Authors


Abstract

BACKGROUND

Elderly and medically infirm cancer patients are increasingly offered allogeneic nonmyeloablative hematopoietic cell transplantation (HCT). A better understanding of the impact of health status on HCT outcomes is warranted. Herein, a recently developed HCT-specific comorbidity index (HCT-CI) was compared with a widely acceptable measure of health status, the Karnofsky performance status (KPS).

METHODS

The outcomes of 341 patients were evaluated, conditioned for either related or unrelated HCT by 2-gray (Gy) total body irradiation given alone or combined with fludarabine at a dose of 90 mg/m2. Comorbidities were assessed retrospectively by the HCT-CI. Performance status before and toxicities after HCT were graded prospectively using the KPS and National Cancer Institute Common Toxicity criteria, respectively.

RESULTS

Weak Spearman rank correlations were noted between HCT-CI and KPS and between the 2 measures and age, number of prior chemotherapy regimens, and intervals between diagnosis and HCT (all r < 0.20). High-risk diseases correlated significantly with higher mean HCT-CI scores (P = .009) but not low KPS (P = .37). In multivariate models, the HCT-CI had significantly greater independent predictive power for toxicities (P = .004), nonrelapse mortality (P = .0002), and overall mortality (P = .0002) compared with the KPS (P = .05, .13, and .05, respectively). Using consolidated HCT-CI and KPS scores, patients were stratified into 4 risk groups with 2-year survivals of 68%, 58%, 41%, and 32%, respectively.

CONCLUSIONS

HCT-CI and KPS should be assessed simultaneously before HCT. The use of both tools combined likely refines risk-stratification for HCT outcomes. Novel guidelines for assessment of performance status among HCT patients are warranted. Cancer 2008. © 2008 American Cancer Society.

Allogeneic conventional hematopoietic cell transplantation (HCT) has been a potentially curative therapy for many patients with advanced hematologic malignancies.1, 2 However, myeloablative conditioning regimens have been associated with relatively high toxicities and nonrelapse mortality (NRM), which have limited allogeneic HCT to young patients or those in good medical condition. Nonmyeloablative or reduced-intensity conditioning regimens have been offered to elderly patients and/or those with significant organ compromise.3–6 To our knowledge, little is known regarding how to apply lessons learned from conventional allogeneic HCT in relatively young patients to nonmyeloablative HCT in elderly patients with comorbidities. In addition, variable rates of NRM have been reported after nonmyeloablative or reduced-intensity allogeneic HCT.4–10 Therefore, a better understanding of the impacts of patient age and health status on nonmyeloablative HCT outcomes is necessary for several reasons, including appropriate risk stratification and patient assignment to HCT protocols, individualization of patient care, and comparison of results of studies done at different centers.

Performance status scales were used as estimates of overall patient health at HCT, and have been found to predict outcomes after reduced-intensity HCT.11, 12 The Karnofsky performance status (KPS) was originally developed to measure self-care, daily activities, and evidence of disease.13 To our knowledge, the extent to which this scale reflects the burden of comorbidities in transplanted patients is not well known. Recently, comorbidities summarized by weighted HCT-specific comorbidity index (HCT-CI) scores have been shown to be strong predictors of NRM after allogeneic HCT.14–16 Herein, we investigated possible correlations between HCT-CI and KPS and between each of these 2 tools and other pretransplant variables. Furthermore, we tested the level of information that either tool added to the other in predicting allogeneic nonmyeloablative HCT outcomes.

MATERIALS AND METHODS

Patients were enrolled under protocols approved by the Institutional Review Boards of the collaborating institutions.

Patients

Records of 408 consecutive patients who underwent allogeneic HCT after nonmyeloablative conditioning between 2000 and 2006 at 4 collaborative academic institutions were reviewed for this study. Comorbidity data were missing for 4 patients and KPS data for 63 patients, resulting in 341 patients with complete comorbidity and KPS data who were included in the current analysis. Nonmyeloablative conditioning consisted of 2 grays (Gy) total body irradiation (TBI) alone (n = 50 patients) or 2 Gy of TBI preceded by fludarabine (n = 291 patients) at a dose of 90 mg/m2, and postgrafting immunosuppression consisted of mycophenolate mofetil (MMF) and cyclosporine (CSP), as previously described.3, 17, 18 All patients were enrolled in Fred Hutchinson Cancer Research Center (FHCRC) consortium studies. The primary differences between protocols included the use of human leukocyte antigen (HLA)-matched related or unrelated grafts, variations in the duration and intensity of CSP and MMF, and the addition of fludarabine to 2 Gy TBI. Patients were offered nonmyeloablative conditioning because they were ineligible for conventional HCT because of 1) age ≥50 years, 2) presence of preexisting significant medical problems, and/or 3) failing high-dose HCT. Exclusion criteria included pregnancy, cardiac ejection fraction of <35% or <40%, pulmonary diffusion capacity <35%, decompensated liver disease (fulminant hepatic failure or cirrhosis with portal hypertension), KPS percent of <50%, <60%, <70%, or <80%, and serologic evidence of infection with the human immunodeficiency virus. No exclusions were made for disease status, chemotherapy sensitivity, renal insufficiency, or active bacterial or fungal infections. The level of matching between patients and donors was assessed for HLA-A, HLA-B, and -C antigens by either intermediate-resolution DNA typing (to a level at least as sensitive as serology) or high-resolution techniques. HLA-matching for -DRB1 and -DQB1 was at the allele level.19 All patients received infection prophylaxis according to standard guidelines.20–24

Health and Disease Risk Assessment Instruments

Comorbidities were assessed retrospectively by comprehensive review of medical records and computer database systems. Scores were assigned using the definitions of 17 comorbidities included in the HCT-CI as previously described.16

Performance status was assessed prospectively as per the Karnofsky scale at the time of HCT by each patient's clinician.13 Information concerning KPS was collected for this analysis by review of medical records. No attempts were made to evaluate KPS retrospectively, and patients who lacked KPS information in their records were excluded. Of note, KPS was 1 of the exclusion criteria for a majority of nonmyeloablative protocols. Three patients were enrolled in protocols with no KPS exclusion criteria, whereas 14 patients were enrolled in protocols with KPS exclusion criteria of <80%, 44 patients with <70%, 160 patients with <60%, and 120 patients with <50%.

Toxicities after HCT were evaluated prospectively by each patient's clinician at the time of the event. Grades were assigned using the Common Toxicity Criteria (CTC) (version 3.0) of the National Cancer Institute. Toxicity grades were collected from the computer databases for this analysis. The CTC contained toxicity categories and each category contained adverse events. Adverse events were graded as: 0 indicates none; 1=, mild; 2, moderate; 3, severe; and 4, life-threatening or debilitating. Only grades 3 and 4 nonhematologic toxicities were considered for this report. Toxicities believed to be related to acute graft-versus-host disease were not graded under corresponding organ toxicities.

Risks of disease relapse/progression were classified retrospectively according to the published categorization for nonmyeloablative patients.25

Statistical Methods

Survival curves and probabilities were estimated using the Kaplan-Meier method, whereas cumulative incidence curves and probabilities for post-HCT toxicities and NRM were estimated as previously described.26 Univariate comparisons of proportions were performed with the chi-square test. Multivariate hazards ratios (HRs) for NRM and survival outcomes were estimated from Cox regression models, treating NRM and disease relapse/progression as competing risks when appropriate. Multivariate P values for a given variable were based on adjustment for all other variables in the model. All correlations were evaluated by Spearman rank correlation. Correlation coefficients of ≤0.30 were considered weak.27

RESULTS

Pretransplant Characteristics, Comorbidities, and Performance Status

A majority (70%) of patients were aged >50 years and 33% were aged >60 years (Table 1). Approximately 23% of patients had failed autologous HCT and 2% had failed allogeneic HCT, whereas 3% of patients had prior planned autologous HCT. A majority of patients had acute myeloid leukemia, myelodysplastic syndromes, lymphomas, chronic lymphocytic leukemia, or multiple myeloma. Nearly half of the patients received related grafts (47%), all of which were HLA-matched except for 2 patients who were given single HLA-antigen mismatched grafts. Fifty-three percent of patients had unrelated grafts, 142 of which were HLA-matched, 21 were 1-HLA antigen mismatched, and 19 were 1-HLA allele mismatched. The majority of patients had diseases with either a standard (61%) or high risk (29%) for relapse.25, 28 Patients had a median HCT-CI score of 2 (range, 0–9) and a median KPS of 85% (range, 50–100%). Approximately 75% and 46% of patients had HCT-CI scores of >0 and ≥3, respecitively, whereas 82% and 38% of patients had a KPS of <100% and ≤80%, respectively (Fig. 1).

Figure 1.

Distribution of pretransplant (A) hematopoietic cell transplantation-comorbidity index (HCT-CI) scores and (B) Karnofsky performance status (KPS) percentages among patients treated with allogeneic HCT after nonmyeloablative conditioning.

Table 1. Patients and Disease Characteristics
CharacteristicsNo. of patients (n = 341)
  • HCT indicates hematopoietic cell transplantation; AML, acute myeloid leukemia; ALL, acute lymphocytic leukemia; MDS, myelodysplastic syndromes; CLL, chronic lymphocytic leukemia; MM, multiple myeloma; NHL, non-Hodgkin lymphoma; HD, Hodgkin disease; G-PBMC, granulocyte-colony-stimulating factor-peripheral blood mononuclear cells; FHCRC, Fred Hutchinson Cancer Research Center.

  • *

    Other diagnoses included chronic myeloid leukemia, Fanconi anemia, immunodeficiency syndromes, neuroblastoma, myeloproliferative disease, renal cell carcinoma, and Waldenstrom macroglobulinemia.

  • Low risk included ALL in first complete remission (CR), CLL and MM in CR, low-grade and mantle cell lymphoma either in or not in CR, high-grade lymphoma in CR, myelofibrosis, and Waldenstrom macroglobulinemia. Standard risk included AML in any CR, MDS-refractory anemia, CML in first chronic phase, and CLL and MM not in CR. High-risk included AML not in CR, transformed AML, ALL beyond first CR, MDS other than refractory anemia, chronic myelomonocytic leukemia, CML beyond first chronic phase, high-grade NHL not in CR, HD, and renal cell carcinoma.25

Median age (range), y56 (1–74)
Median follow-up (range), mo17 (0.9–73)
Median no. of preceding treatment regimens (range)3 (0–10)
Median (mo) between diagnosis and HCT (range)20.8 (3.2–413)
Prior high-dose HCT, %
 Planned autologous3
 Failed autologous23
 Failed allogeneic2
Diagnosis, %
 AML23
 ALL6
 MDS10
 CLL10
 MM18
 NHL19
 HD8
 Others*6
Recurrence risk, %
 Low10
 Standard61
 High29
Hematopoietic cell source, %
 G-PBMC99
 Bone marrow1
Donor, %
 Related47
 Unrelated53
Institutions, %
 FHCRC and University of Washington82
 Veterans Affairs Puget Sound Health Care System7
 Emory University5
 Oregon Health and Science University3
 University of Utah2
 Rocky Mountain Cancer Center1

Correlation Studies

By Spearman rank correlation, there was a weak inverse correlation between HCT-CI scores and KPS percentages (r = −0.18; P = .001) (Fig. 2A). Weak Spearman rank correlations (Fig. 2B-D) were also detected between HCT-CI scores and age (r = 0.08; P = .12), number of preceding chemotherapy regimens (r = 0.08; P = .14), and months between diagnosis and HCT (r = −0.10; P = .08), respectively. Similarly, KPS percentages had weak correlations with each of the 3 characteristics (r = < 0.01 [P = .99]; r = −0.02 [P = .69]; and r = 0.01 [P = .90], respectively). There were statistically significantly higher mean HCT-CI scores among patients with high-risk (2.7) versus standard (2.2) and low-risk diseases (2.0, P = .009 by test for trend across risk category). By comparison, no statistically significant differences in mean KPS were detected among the 3 disease-risk categories (87.3%, 86.6%, and 85.7%, respectively; P = .37). High comorbidity scores (≥3) and low KPS percentages (≤80%) were found among 60% and 38% of patients with high-risk diseases (Fig. 3), respectively.

Figure 2.

Correlation between hematopoietic cell transplantation-comorbidity index (HCT-CI) scores and (A) Karnofsky performance status (KPS) percentages, (B) age, (C) number of preceding chemotherapy regimens, and (D) months between diagnosis and HCT. Dots represent individual patients, dashed lines indicates least squares regression line.

Figure 3.

The distribution of hematopoietic cell transplantation-comorbidity index (HCT-CI) scores of 0 to 2 versus ≥3 and Karnofsky performance status (KPS) percentages of > 80% versus ≤80% among patients with high-risk disease. Patients with high-risk disease had a higher incidence of high comorbidity scores but lower incidence of poor KPS.

Impact of HCT-CI and KPS on Nonmyeloablative HCT Outcomes

Overall, patients had 53% and 18% cumulative incidences of grade ≥3 and grade 4 toxicities at 120 days, respectively. The 2-year NRM and overall survival were 24% and 52%, respectively. Higher HCT-CI scores (≥3 vs 0–2) were associated with statistically significantly higher incidences and rates of grades ≥3 (P = .001) and grade 4 toxicities (P = .004), NRM (P < 0.0001), and mortality (P < .0001) (Fig. 4). By comparison, lower KPS percentages (≤80% vs > 80%) were associated with higher ≥grade 3 toxicities (P = .01), higher NRM (P = .04), and higher mortality (P = .01), but not higher grade 4 toxicities (P = .24). Multivariate Cox regression models were designed to study independent effects of pretransplant factors on HCT outcomes (Table 2). Factors analyzed were HCT-CI scores, KPS percentages, age, number of preceding chemotherapy regimens, disease risk, cytomegalovirus (CMV) serum status, and months between diagnosis and HCT. Both high HCT-CI scores (P = .004) and low KPS percentages (P = .05) were found to be independently predictive of ≥grade 3 toxicities, whereas only HCT-CI scores were found to significantly predict grade 4 toxicities (P = .03). High HCT-CI scores (P = .0002), greater age (P = .01), and greater numbers of preceding chemotherapy regimens (P = .004) were found to be statistically significantly associated with increased NRM, whereas low KPS percentages had a trend toward increased NRM (P = .13). Higher mortality was found to be independently associated with higher HCT-CI scores (P = .0002), lower KPS percentages (P = .05), greater age (P = .0006), greater numbers of prior regimens (P = .002), and diseases with higher recurrence risks (P = .0002). CMV serum status and interval between diagnosis and HCT were not found to be significant in predicting HCT outcomes.

Figure 4.

Cumulative incidences of (A) grades 3 to 4 toxicity, (B) nonrelapse mortality, and (C) Kaplan-Meier probabilities of survival among patients treated with allogeneic nonmyeloablative hematopoietic cell transplantation (HCT) as stratified by HCT-comorbidity index (HCT-CI) scores (0-2 vs ≥3) and Karnofsky performance status (KPS) percentages (>80% vs ≤80%).

Table 2. Multivariate Analyses of Risk Factors for Grades 3 to 4 and Grade 4 Toxicities*, Nonrelapse Mortality, and Survival Among Patients Treated With Allogeneic HCT After Nonmyeloablative Conditioning
Risk factors (No. of patients)Grades 3–4 toxicitiesGrade 4 toxicitiesNRMSurvival
HRPHRPHRPHRP
  • HCT indicates hematopoietic cell transplantation; NRM, nonrelapse mortality, HR, hazards ratio; CI, comorbidity index; KPS, Karnofsky performance status.

  • Cytomegalovirus serostatus and the interval between diagnosis and HCT were not found to be statistically significant in predicting any of the 4 outcomes.

  • *

    Toxicities were graded using the Common Toxicity Criteria (version 3.0) of the National Cancer Institute.

HCT-CI scores0–2 (183)1.0 1.0 1.0 1.0 
≥3 (156)1.58.0041.83.032.55.00021.97.0002
KPS percentages>80% (211)1.0 1.0 1.0 1.0 
≤80% (128)1.36.051.21.481.45.131.42.05
Age, y≤50 (104)1.0 1.0 1.0 1.0 
>50 (235)1.12.521.12.692.07.011.99.0006
No. of preceding chemotherapy regimens0–3 (188)1.0 1.0 1.0 1.0 
≥4 (151)0.78.151.43.202.12.0041.77.002
Disease riskLow/standard (240)1.0 1.0 1.0 1.0 
High (99)1.12.521.43.201.40.222.01.0002

Consolidation of HCT-CI and KPS

Patients with HCT-CI scores of 0–2 or ≥3 were further stratified based on KPS percentages of >80% versus ≤80% (Table 3). Accordingly, patients were divided into 4 groups that had increasing risks for worse outcomes: Group I included 38% of the patients, Group II included 16%, Group III included 25%, and Group IV included 21%. Patients in the 4 groups had Day 120 ≥grade 3 toxicities of 41%, 57%, 59%, and 66%, respectively (P = .002); 2-year NRM of 16%, 17%, 30%, and 39%, respectively (P < .0001); and 2-year survivals of 68%, 58%, 41%, and 32%, respectively (P < 0.0001) (Fig. 5).

Figure 5.

Cumulative incidences of (A) grades 3 to 4 toxicity, (B) nonrelapse mortality, and (C) Kaplan-Meier probabilities of survival among patients treated with allogeneic nonmyeloablative hematopoietic cell transplantation (HCT) as stratified into 4 risk groups based on a consolidated HCT-comorbidity index (HCT-CI) and Karnofsky performance status (KPS) scale. Group I (solid black line) includes patients with HCT-CI scores of 0 to 2 and a KPS of >80%. Group II (dotted black line) includes patients with HCT-CI scores of 0 to 2 and a KPS of ≤80%. Group III (solid blue line) includes patients with HCT-CI scores of ≥3 and a KPS of >80%. Group IV (dotted blue line) includes patients with HCT-CI scores of ≥3 and a KPS of ≤80%.

Table 3. Consolidation of HCT-CI Scores and KPS Percentages Into 4 Risk Groups and Distribution of Patients Among These Groups
Risk groupsDefinitionsPatients (n = 341) %
HCT-CIKPS
  1. HCT-CI indicates hematopoietic cell transplantation-comorbidity index; KPS, Karnofsky performance status.

I0–2>80%38
II0–2≤80%16
III≥3>80%25
IV≥3≤80%21

DISCUSSION

Nonmyeloablative conditioning regimens have allowed allogeneic HCT to be offered to older patients.29, 30 Older cancer patients often simultaneously have comorbidities, impaired functional status, and decreased social support as potential causes for poor treatment outcomes.31, 32 In addition, younger, medically infirm patients have been treated by nonmyeloablative protocols. It has become important to understand the biologic impact of impaired health status on nonmyeloablative HCT outcomes. Accurate tools are warranted to reliably separate medically fit from medically unfit patients. A recently developed HCT-CI has been shown to predict NRM and survival more accurately than several of the other comorbidity indices.16, 33–41 Herein, we assessed the predictive powers of both the HCT-CI and a measure of physical function, namely, the KPS.

We found that HCT-CI scores and KPS percentages were only weakly correlated with each other, indicating that each of these tools captured different levels of the patients' health status. This was not surprising because the HCT-CI was developed to map individual organ function impairments, whereas the KPS was designed to measure self-care, daily activity, and evidence of disease. Our results agreed with the previously reported weak correlation between comorbidity and performance status in both geriatric and solid malignancy settings.42, 43 This finding suggested the use of both comorbidity and performance status tools to evaluate patients at the time of HCT.

In this study, increasing age was found to be correlated with neither higher comorbidity scores nor lower performance status, which contrasts with reports from geriatric practice and conventional chemotherapy trials for hematologic malignancies.44–46 This discrepancy might be because of inclusion criteria for nonmyeloablative protocols, which enrolled patients aged >50 years regardless of comorbidities, whereas younger patients were enrolled if they had significant impairment of health status. Clearly, studying the impact of chronologic age on nonmyeloablative HCT outcomes without considering comorbidities could be misleading. This might explain the variable results in previous studies that did not consider comorbidities and demonstrated that age either did11, 47 or did not12, 48–50 affect HCT outcomes.

Patients with high-risk disease for recurrence frequently had increased comorbidities. This could not be explained by greater number of preceding chemotherapy regimens because no correlations between these numbers and HCT-CI scores were found. Differences in the intensity of chemotherapy regimens given to patients with aggressive compared with indolent malignancies might be an explanation for the findings. Another explanation could be the reported linkage between comorbidities such as prior malignancy,51 diabetes,52, 53 obesity,54, 55 autoimmune diseases,56–58 or smoking-associated lung disease59, 60 and aggressive malignancies (reviewed previously61).

Several previous analyses of prognostic factors for HCT outcomes did not include assessment of health status.62–68 Others, using performance status as the sole measure of the patients' medical impairments, either reported correlations with both NRM and survivals,10, 49, 69, 70 NRM but not survival,11 disease relapse and progression-free survival but not NRM,71 or none of the outcome measures.72 Failure to account for comorbidities and/or absence of uniform guidelines to assess performance status may have been responsible for the variable results.

In the current report, both HCT-CI scores and KPS percentages were found to be independently correlated with toxicities, NRM, and overall mortality. However, the HCT-CI had greater predictive power than the KPS in multivariate analyses. This might partly be because of the fact that the KPS, but not the HCT-CI, has been used as a protocol exclusion criterion. Another explanation could be the greater spread of data from the HCT-CI, as opposed to the KPS, which virtually guaranteed greater predictive power of the former. Contrary to our results, 2 previous reports from a single institution demonstrated that performance status was of higher prognostic significance than both the Charlson comorbidity index and the Kaplan-Feinstein scale.73, 74 We have reported that the HCT-CI has higher sensitivity and better ability to predict outcomes than these other 2 indices,16, 33 which might explain the current results.

By consolidating the HCT-CI with a measure of physical function represented by KPS, we were able to stratify current patients evenly into 4 risk groups with 2-year survivals of 68%, 58%, 41%, and 32%, respectively (P < .0001). One previous report merged the Kaplan-Feinstein scale with an Eastern Cooperative Oncology Group (ECOG) performance status scale to analyze outcomes in 105 patients given reduced-intensity HCT.74 The merged scales stratified patients into low-risk (73%) and high-risk groups (27%) with 6-month NRMs of 15% and 50%, respectively. No description of the impact on overall survival was provided.

In conclusion, the HCT-CI and the KPS assessed different levels of medical health status. Merging the information attained by the 2 scoring systems refined the outcome predictions for patients given nonmyeloablative HCT. In the future, it would be of interest to assess the impacts of both the HCT-CI and the KPS on conventional HCT outcomes.

Acknowledgements

We thank the data coordinators (Heather Hildebrant and Jennifer Freese) and the study nurses (Mary Hinds, John Sedgwick, Michelle Bouvier, and Joanne Greene) for their invaluable help in making the study possible. Bonnie Larson, Helen Crawford, Karen Carbonneau, and Sue Carbonneau provided assistance with article preparation. We also thank the Transplant Teams

Ancillary