Hematopoietic cell transplantation comorbidity index predicts transplantation outcomes in pediatric patients

Despite major advances in the field of hematopoietic cell transplantation (HCT) in recent years, life-threatening complications still occur. Quantifying this risk of toxicity for individual patients is challenging, but essential for accurate pre-HCT counseling. Without adequate means to gauge pre-HCT risk, it is difficult to determine the value of the therapy for individual patients, and making an accurate comparison of outcomes between centers is challenging.

Several indices have been used to quantify the impact of different comorbidities on overall disease-specific outcomes. The Charlson Comorbidity Index (CCI) is one such commonly used index which was developed by assigning weights to 19 chronic conditions according to their association with 1-year mortality. A subsequent validation cohort confirmed that there was a stepwise increase in cumulative mortality with each increased comorbidity level.^1,2  The CCI is frequently used in adults with a variety of chronic medical conditions and solid tumors, and has been used to predict nonrelapse mortality (NRM) successfully in patients undergoing HCT.^3,4  The sensitivity of the CCI in HCT patients is questionable, however, because many of the comorbidities included are rarely seen in HCT recipients and it misses many common HCT-related comorbidities such as recent infections.

To further refine an index for HCT patients, Sorror et al developed the hematopoietic cell transplantation–specific comorbidity index (HCT-CI).⁵  The HCT-CI is a modification of the CCI that incorporates more specific comorbidity definitions and the addition of frequently seen HCT specific comorbidities. After retrospective testing and validation in more than 1000 HCT patients, 17 comorbidities with a hazard ratio of > 1.2 were included in the HCT-CI. A head to head comparison of the HCT-CI and CCI showed that the HCT-CI was more sensitive and a better predictor of NRM and OS in HCT patients.⁵  Although some authors have questioned its universal applicability,^6,7  the HCT-CI has been subsequently validated with varying degrees of predictive ability in several independent adult HCT cohorts, including allogeneic, autologous, myeloablative (MA) reduced-intensity conditioning (RIC), and nonmyeloablative (NMA) patients.^8-12 

The HCT-CI includes important comorbidities with varying effects on treatment-related mortality and combines them into one score. It has been confirmed as a useful tool in adults, but has not yet been validated in children undergoing HCT. The aim of this analysis was to determine whether the HCT-CI score accurately predicts outcomes in pediatric patients undergoing HCT.

This was a retrospective cohort study of 252 pediatric patients (≤ 20 years) undergoing their first allogeneic HCT between January 2008 and May 2009 at 4 large transplantation centers (Cincinnati Children's Hospital Medical Center [CCHMC], Dana-Farber Cancer Institute/Children's Hospital Boston [DFCI/CHB], University of Minnesota Amplatz Children's Hospital [UMACH], and Texas Children's Cancer Center [TXCCC]). Consecutive pediatric patients were included regardless of diagnosis, donor source, conditioning intensity, or graft-versus-host disease prophylaxis, as long as prospectively collected comorbidity data were available for review. Conditioning regimens were classified according to the Center for International Blood and Marrow Transplant Research (CIBMTR) approach.¹³  Supportive care measures varied by institution, but in general included antimicrobial prophylaxis and infectious disease monitoring. All treatment protocols were reviewed and approved by the institutional review boards at each center and all patients/guardians provided signed informed consent in accordance with the Declaration of Helsinki.

Pretransplantation comorbidities were collected prospectively for each patient at the individual centers using the CIBMTR pretransplantation essential data (pre-TED) form. The comorbidities captured by this form include a history of mechanical ventilation, fungal infection, other infections, cardiac disorders, cerebrovascular disease, diabetes, altered hepatic function, inflammatory bowel disease, obesity, peptic ulcer disease, psychiatric disturbance, pulmonary abnormalities, renal insufficiency, and rheumatologic disorders (see Table 2 for specific comorbidity definitions). At the time of the analysis, each patient's records were examined retrospectively in a uniform fashion by one site representative to confirm that all comorbidities were captured accurately.

Excluding the mechanical ventilation and fungal infection history parameters, the comorbidities collected on the pre-TED form were then used to assign HCT-CI scores as originally described by Sorror et al.⁵  The adjusted hazard ratios (HRs; controlled for all coexisting comorbidities, age, conditioning intensity, and disease risk) and integer weights previously published by these authors were used for scoring. Comorbidities with an adjusted HR of 1.3-2.0 (arrhythmia, cardiac, inflammatory bowel disease, diabetes, cerebrovascular disease, psychiatric disturbance, mild hepatic, obesity and infection) were assigned a weight of 1, comorbidities with an adjusted HR of 2.1-3.0 (rheumatologic, peptic ulcer, moderate/severe renal, and moderate pulmonary) were assigned a weight of 2 and comorbidities with an adjusted HR of 3.0 or more (prior solid tumor, heart valve disease, severe pulmonary, and moderate/severe hepatic) were assigned a weight of 3. The HCT-CI was the sum of these integer weights. Patients then were assigned to 1 of 3 risk groups: low risk (score 0), intermediate risk (score 1-2), and high risk (score 3+).⁵  Other transplantation-related and outcome data were collected retrospectively at each center.

The primary endpoints for this analysis were the cumulative incidence of NRM and probability of overall survival (OS) at 1-year post HCT. NRM was defined as death after HCT without disease progression or relapse.

Kaplan-Meier curves were used to estimate the probability of OS and the cumulative incidence method was used to estimate the probability of NRM through 2 years posttransplantation, treating relapse as a competing risk.^14,15  The independent effect of the HCT-CI on OS and NRM was assessed by Cox regression and Fine and Gray models, respectively.^16,17  All factors were tested for the proportional hazards assumption before inclusion in the models. Factors included in the regression models were HCT-CI score (0 vs 1-2 vs 3+), age (0-1 vs 2-10 vs 11-20), conditioning (MA vs RIC/NMA), donor type (matched related donor vs mismatched related donor vs matched unrelated donor vs mismatched unrelated donor vs umbilical cord blood), and center. All P values were 2 sided. Analyses were performed using SAS 9.2 (SAS Institute) and R 2.4 statistical software.

Tables 1 and 2 show demographic, treatment, and comorbidity data by center. Underlying diagnoses included leukemia (n = 110, 44%), immune deficiency (n = 63, 25%), storage disorders (n = 24, 9%), aplastic anemia (n = 22, 9%), Fanconi anemia (n = 18, 7%), benign hematologic abnormalities (n = 9, 4%), and other (n = 6, 2%). The median age at transplantation was 6 years (range, 0.1-20). Donor sources included HLA-matched related donor (n = 74), HLA-mismatched related donor (n = 21), HLA-matched unrelated donor (n = 75), HLA-mismatched unrelated donor (n = 34), and umbilical cord blood (n = 48). Seventy-five percent received MA conditioning. The median follow-up for surviving patients was 343 days (range, 110-624).

The most common comorbidities seen in our cohort were infection (n = 18), pulmonary dysfunction (moderate, n = 29; severe, n = 12), and hepatic dysfunction (mild, n = 40; moderate/severe, n = 26). Peptic ulcer disease was not seen in our cohort. HCT-CI scores were distributed as follows: HCT-CI = 0 (n = 139), 1-2 (n = 52), and 3+ (n = 61). The distribution varied significantly by center as seen in Table 1.

The cumulative incidence of NRM at 1 year increased significantly with increasing HCT-CI score (HCT-CI score = 0: 10% [95% confidence interval, 5%-15%]; HCT-CI score = 1-2: 14% [95% confidence interval, 5%-23%]; and HCT-CI score = 3+: 28% [95% confidence interval, 16%-40%]; P < .01; Figure 1A). Other factors affecting the incidence of NRM at 1 year included a prior history of mechanical ventilation (no, 13% vs yes, 31%; P < .01) and proven fungal infection (no, 14% vs yes, 39%; P < .01). A multivariate analysis (Table 3) controlling for age, conditioning intensity, donor type, and center showed that those with HCT-CI scores of 1-2 and 3+ had a relative risk of NRM of 1.5 (95% confidence interval, 0.5-4.3, P = .48) and 4.5 (95% confidence interval, 1.7-12.1, P < .01) compared with those with a score of 0. As expected, those receiving MA conditioning also had a higher relative risk of NRM compared with those receiving RIC/NMA conditioning though this did not reach statistical significance. Age and center had no effect on NRM.

One hundred three patients with hematologic malignancies received MA HCT. Of these, 49 had an HCT score = 0, 33 had an HCT score = 1-2, and 21 had an HCT score of 3+. The cumulative incidence of NRM at 1 year increased significantly with increasing HCT-CI score (HCT-CI score = 0: 4% [95% confidence interval, 0%-9%]; HCT-CI score = 1-2: 9% [95% confidence interval, 1%-18%]; and HCT-CI score 3+: 39% [95% confidence interval, 18%-60%]; P < .01). A multivariate analysis (Table 4 “Patients with hematologic malignancies receiving MA HCT”) controlling for age, conditioning intensity, donor type, and center showed that those with HCT-CI scores of 1-2 and 3+ had a relative risk of NRM of 2.2 (95% confidence interval, 0.3-16.7, P = .44) and 10.2 (95% confidence interval, 1.3-82.6, P = .03) compared with those with a score of 0. Age and center had no effect on NRM.

One hundred eighty-nine patients received MA HCT. Of these, 109 had an HCT score of 0, 46 had an HCT score of 1-2, and 34 had an HCT score of 3+. The cumulative incidence of NRM at 1 year increased significantly with increasing HCT-CI score (HCT-CI score = 0: 10% [95% confidence interval, 4%-16%]; HCT-CI score = 1-2: 16% [95% confidence interval, 5%-27%]; and HCT-CI score 3+: 36% [95% confidence interval, 19%-53%]; P < .01). A multivariate analysis (Table 4 “All patients receiving MA HCT”) controlling for age, conditioning intensity, donor type, and center showed that those with HCT-CI scores of 1-2 and 3+ had a relative risk of NRM of 1.9 (95% confidence interval, 0.6-6.0, P = .30) and 5.5 (95% confidence interval, 1.9-15.9, P < .01) compared with those with a score of 0. Those patients aged 2-10 years had a lower relative risk (RR) of NRM (RR 0.4 [95% confidence interval, 0.1-1.0, P = .04]) than those aged 0-1 year. Center had no effect on NRM.

Sixty-one patients received RIC/NMA HCT. Of these, 30 had an HCT score of 0, 6 had an HCT score of 1-2, and 25 had an HCT score of 3+. The cumulative incidence of NRM at 1 year trended up with increasing HCT-CI score (HCT-CI score = 0: 7% [95% confidence interval, 0%-16%]; HCT-CI score = 1-2: 0%; and HCT-CI score 3+: 19% [95% confidence interval, 3%-35%]; P = .22), though this was not statistically significant. Because of the small number of patients in this subgroup, a multivariate analysis was not performed.

The probability of OS at 1 year decreased significantly with increasing HCT-CI score (HCT-CI score = 0: 88% [95% confidence interval, 80%-92%]; HCT-CI score = 1-2: 67% [95% confidence interval, 50%-79%]; and HCT-CI score = 3+: 62% [95% confidence interval, 48%-74%]; P < .01; Figure 1B). Other factors affecting OS at 1 year included a prior history of mechanical ventilation (no, 78% vs yes, 65%; P = .02) and proven fungal infection (no, 78% vs yes, 44%; P < .01). A multivariate analysis (Table 3) controlling for age, conditioning intensity, donor type, and center showed that those with HCT-CI scores of 1-2 and 3+ had a RR of death of 2.6 (95%CI 1.2-5.6, P = .01) and 4.6 (95%CI 2.1-9.7, P < .01), respectively, compared with those with a score of 0. Those receiving MA conditioning had a trend toward reduced OS. Age and center had no effect on OS.

In patients with hematologic malignancies receiving MA HCT, the probability of OS at 1 year decreased significantly with increasing HCT-CI score (HCT-CI score = 0: 87% [95% confidence interval, 71%-95%]; HCT-CI score = 1-2: 62% [95% confidence interval, 40%-77%]; and HCT-CI score = 3+: 46% [95% confidence interval, 24%-66%]; P < .01). A multivariate analysis (Table 4 “Patients with hematologic malignancies receiving MA HCT”) controlling for age, conditioning intensity, donor type, and center showed that those with HCT-CI scores of 1-2 and 3+ had a RR of death of 3.1 (95% confidence interval, 1.4-7.0, P < .01) and 4.6 (95% confidence interval, 1.9-11.0, P < .01), respectively, compared with those with a score of 0. Age and center had no effect on OS.

In all patients receiving MA HCT, the probability of OS at 1 year decreased significantly with increasing HCT-CI score (HCT-CI score = 0: 86% [95% confidence interval, 77%-92%]; HCT-CI score = 1-2: 63% [95% confidence interval, 45%-76%]; and HCT-CI score = 3+: 54% [95% confidence interval, 36%-70%]; P < .01). A multivariate analysis (Table 4 “Patients with hematologic malignancies receiving MA HCT”) controlling for age, conditioning intensity, donor type, and center showed that those with HCT-CI scores of 1-2 and 3+ had a relative risk of death of 3.5 (95% confidence interval, 1.111.0, P = .04) and 5.3 (95% confidence interval, 1.5-18.7, P < .01), respectively, compared with those with a score of 0. Age and center had no effect on OS.

In patients receiving RIC/NMA HCT, the probability of OS at 1 year trended down with increasing HCT-CI score (HCT-CI score = 0: 93% [95% confidence interval, 74%-98%]; HCT-CI score = 1-2: 100%; and HCT-CI score = 3+: 71% [95% confidence interval, 46%-86%]; P = .09), though this was not statistically significant. Because of the small number of patients in this subgroup, a multivariate analysis was not performed.

Since the original publication by Sorror et al in 2005 outlining the predictive ability of the HCT-CI,⁵  many groups have sought to validate the scoring system in a variety of disease and donor-specific settings.^6-12  None of these researchers focused on pediatric patients undergoing transplantation. The variation in the predictive ability of the HCT-CI among adult cohorts exemplified the need to test the scoring system in pediatrics, another very unique population. This is the first article to show that the HCT-CI scoring system predicts NRM and OS in pediatric patients undergoing allogeneic HCT and, consequently, is a useful tool to assess risk, guide counseling, and devise innovative therapies in the highest pediatric risk groups.

Our patient cohort was fairly heterogeneous in that it included pediatric patients with a variety of diagnoses who received both MA and RIC/NMA conditioning. Because previous reports have highlighted the variability in the predictive value of the HCT-CI by factors like disease and conditioning intensity, we also performed subanalyses in more homogenous groups (patients with hematologic malignancies receiving MA HCT, all patients receiving MA HCT, and all patients receiving RIC/NMA HCT). The predictive ability of the HCT-CI was confirmed in both MA subgroups. In addition, despite very limited numbers, there was a trend toward increasing NRM and decreasing OS with increasing HCT-CI score in the RIC/NMA subgroup. Larger analyses need to be performed in RIC/NMA patients to verify this trend. Overall, though, the predictive capability of the HCT-CI in pediatric patients appears to be generalizable.

Further refinement of the scoring system specifically for children, however, may be helpful given that 55% of the patients in our cohort had an HCT-CI score of 0 indicating that the scoring system lacks some sensitivity. This refinement could include an adjustment of the comorbidity definitions to better fit pediatric patients (eg, more accurately define renal dysfunction using glomerular filtration rate or a multiple of the upper limit of normal serum creatinine rather than having a strict cutoff of serum creatinine > 2 mg/dL). The addition of other pediatric specific factors, like congenital heart disease or multiple congenital abnormalities as is seen in genetic conditions like Fanconi anemia or Hurler syndrome, could also be considered. This revised score would then need to be tested and validated in a larger population of pediatric HCT patients as Sorror et al did in their original study.⁵ 

One limitation of this analysis is the difference in the number of comorbidities reported by each center. This could be secondary to a fundamental risk difference in the patient populations based on individual center expertise, conditioning intensity, and donor type, or because of differences in pre-HCT testing requirements at each center. For example, the utility of pulmonary function testing is controversial in pediatrics given that the results are dependent on patient understanding, cooperation, and effort. Only 2 of the 4 centers included in this analysis routinely do pulmonary function testing before transplantation. As a result, these centers had a larger number of recorded pulmonary comorbidities. Despite a significantly different number of comorbidities between centers, though, each center had similar NRM rates leading us to believe that there was no confounding effect on the relationship between HCT-CI and NRM. In addition, “center” was included as a factor in the regression analyses and was determined not to be a significant predictor or NRM or OS.

Another potential explanation for the difference in the number of comorbidities at each center is potential variability in how the individual scores were assigned by the treating physicians. To exclude this possibility, every patient's comorbidities were retrospectively reviewed and rescored in a uniform fashion by a designated person from each site. By doing this, several additional comorbidities were discovered at each site (total = +69; CCHMC = +11, DFCI/CHB = +27, UMACH = +17, TXCCC = +14). The majority of the added comorbidities (53/69 or 77%) were in the hepatic and pulmonary categories. Potential reasons for initially overlooking these comorbidities include unfamiliarity of the transplant physician with the specific comorbidity definitions and the assumption that small elevations in transaminases or decreases in pulmonary function testing in otherwise well-appearing children were not significant enough to note. Regardless of the reason, our discovery of inconsistent scoring illustrates that this may be a common occurrence and these scores should be double checked for accuracy. For a CI to be accurate and useful, it needs to be applied consistently within and among centers.

In conclusion, HCT-CI score predicts NRM and OS in pediatric patients undergoing HCT based on data from 4 large pediatric transplantation centers and may be used in the pre-HCT setting to help quantify the risk of mortality from transplantation-related complications. Refinement of the scoring system for pediatrics and subsequent prospective testing in larger populations may further increase its utility.

We thank James Arce from Texas Children's Cancer Center and Katherine Whitley from Dana-Farber Cancer Institute for their assistance with data collection and management.

Contribution: A.R.S. designed research, collected, analyzed, and interpreted data, and wrote the manuscript; N.S.M. designed research, analyzed and interpreted data, and reviewed the manuscript; M.L.M. designed research, analyzed and interpreted data, and reviewed the manuscript; T.E.D. performed statistical analysis and reviewed the manuscript; S.J. designed research, collected, analyzed, and interpreted data, and reviewed the manuscript; L.E.L. designed research, analyzed and interpreted data, and reviewed the manuscript; R.K. designed research, analyzed and interpreted data, and reviewed the manuscript; and S.M.D. designed research, analyzed and interpreted data, and reviewed the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Angela R. Smith, MD, MS, Assistant Professor Pediatric Blood and Marrow Transplantation, University of Minnesota, 420 Delaware St SE; MMC 484, Minneapolis, MN 55455; e-mail: smith719@umn.edu.