Brain Imaging and Overall Survival after Allogeneic Hematopoietic Cell Transplantation

This work was carried out in collaboration between all authors. Authors BTC and HO designed the study, wrote the protocol and managed the data analysis. Author AD did the statistical analysis. Authors BTC, AOO and HO wrote the first draft of the manuscript and did the literature research. Authors CT and PP contributed to data collection and clinical correlation. All authors revised the manuscript and approved the final manuscript. ABSTRACT Aim: We conducted a retrospective review of all brain imaging studies in the first year after allogeneic haematopoietic cell transplantation (HCT) to determine (a) the percentage of patients with CNS neurological complications based solely on undergoing brain imaging, (b) transplant-related risk factors of undergoing brain imaging, and (c) overall survival in the patients with neurological compared to those transplant who did not have brain Comparisons between patient groups with brain imaging and without brain imaging were tested using the Pearson chi-square test. Survival analyses with at date of transplant and used Kaplan-Meier methods. Results: Of 543 HCT recipients, 128 patients (24%) underwent brain imaging during the first year after transplantation. There was a greater risk of brain imaging in unrelated donor transplants and in lymphoid as opposed to myeloid malignancies (respective hazard ratios 1.45 and 1.43, P =0.04). Overall survival was significantly worse in unrelated donor transplants (hazard ratio 1.42, P =0.003) and in cord blood transplants (hazard ratio 1.68, P =0.02). Landmark survival analysis of patients alive 1 year after HCT showed worse survival over the next 5 years in those who had brain imaging in the first post transplant year ( P <0.0001). Conclusion: These results suggest that development of neurological symptoms or a sign sufficient to prompt clinicians to order brain imaging early after HCT identifies a poor prognosis in transplant population.


INTRODUCTION
Reported incidences of neurological complications after allogeneic hematopoietic cell transplantation (HCT) vary widely in the transplantation literature. The highest reported incidence is 46% in adult patients who underwent unrelated-donor HCT [1], and 46% in paediatric patients who had their transplants in the early years between 1976 and 1983 [2]. A more recent study of paediatric transplants reported a much lower incidence of 15.8% [3].
Most clinical studies are based on retrospective chart reviews. One prospective study of 115 consecutive transplants in the early 1990s reported neurological complications in 17% of patients in the first 3 months after transplant [4]. A difficulty in making comparisons to studies that reported lower incidences of neurological complication [5,6] may be one of definition. The clinician may have restricted the diagnosis of neurological complication to more severe CNS problems or to structurally apparent diagnoses.
Recently we reported brain imaging results in 128 consecutive transplant patients who underwent brain magnetic resonance imaging (MRI) or computed tomography (CT) imaging in the first year after HCT [7]. Indications for brain imaging in these patients included encephalopathy or confusion (40%), headache (20%), focal neurological signs (12%), seizures (10%), trauma or fall (3%), other indications (16%). In the present report, we analyze data from these patients along with consecutive HCT patients seen at the same time from the same centre who did not have brain imaging. Our purpose is to determine (a) the incidence of central nervous system (CNS) neurological complications based solely on ordering of brain imaging, (b) transplant-related risk factors that lead to brain imaging, and (c) overall survival in those patients with neurological complications compared to those patients who did not have brain imaging.

MATERIALS AND METHODS
Subjects were 543 consecutive recipients (August 2004-August 2007) of allogeneic HCT for haematological malignancy or a related disorder. Donors were HLA-matched siblings or unrelated. Subjects were followed for overall survival for up to 6 years after HCT through October 2010. This study was approved by our Institutional Review Board.
Comparisons between patient groups with brain imaging and without brain imaging were tested using the Pearson chi-square test for categorical data, and the two-sample t-test for continuous measurements. Tests were two-sided and the cut-off for statistical significance was 0.05. Survival analyses with outcome time-to-brain-scan started at date of transplant and used Kaplan-Meier and Cox proportional-hazards methods. Regressions were reported as models with univariate predictors, and as reduced multivariable models including just the significant predictors. Risk ratios were reported with 95% confidence intervals. Timedependent Cox proportional-hazards model regressions with outcome of overall survival (time to death or last contact) were done along with Landmark Kaplan-Meier analysis starting at one year after transplant on all patients surviving one year.
The results of the categorical analysis on brain imaging data are shown in Table 2. Of the 543 patients, 128 (24%) had brain imaging during the first year after HCT. A total of 173 CTs and 103 MRIs were done in these 128 patients. Median time between transplant and the first brain image was 1.33 months, with a range of 0.03 to 11.93 months. Patients who received HCT from HLA-matched unrelated donors had a marginally higher percentage of brain imaging studies (27%, 71 of 263 patients) than those who received HCT from sibling donors (20%, 57 of 280 patients). This difference was not statistically significant by categorical analysis (P=0.07). But statistical significance was shown in one predictor of the two-predictor final multivariate Cox model with a hazard ratio of 1.45 ([95% CI 1.02-2.06], P=0.04). A greater number of patients with lymphoid (27%, 72 of 265 patients) as opposed to myeloid (20%, 56 of 278 patients) malignancies had brain imaging. This difference also was not significant by categorical analysis (P=0.06) but was significant as the other predictor of the final multivariate Cox model chosen by backwards elimination (hazard ratio 1.43 [95% CI 1.01-2.03], P=0.04). The univariate hazard ratio for recipients of cord-blood HCT compared to peripheral-blood transplants was 1.81 ([95% CI 0.97-3.38], P=0.06). No effect on the likelihood of undergoing brain imaging was found in categorical analysis with other parameters, including conditioning regimen (Non-myeloablative versus Fully ablative, P=0.28), type of stem cells transplanted (bone marrow, cord blood or peripheral blood, P=0.13), or disease status at the time of HCT (P=0.21). Also, counts of CD34+ or CD3+ cells in the HCT did not affect the likelihood of undergoing brain imaging (data not shown).

Risk of Mortality and Survival Analysis
Univariate Cox-regression analysis of overall survival in all 543 patients (  To determine the effect of brain imaging on overall survival in all 543 patients, we constructed a multivariate model chosen by backwards-elimination Cox. This final reduced model (P<0.0001) estimating hazard ratio of death had two predictor terms: patients who had brain imaging (treated as a time-dependent variable) having a hazard ratio of 3.59 (95% CI 2.81-4.59) over patients who did not have brain imaging (P<0.0001). Fig. 1 shows overall survival beginning year one post HCT in the 381 patients who survived the first year after transplant. Overall survival is graphed as a Landmark analysis starting one year posttransplant as there is no easy depiction of survival in the case of time-dependent variables starting at date of transplant. We take advantage of the fact that all brain imaging relevant to this study is complete by one year post-transplant, fixing the categories of comparison in advance of the start point of Landmark analysis. Survival is worse at year 2-6 in the 65 patients who had brain imaging in year 1 compared to the 316 patients who did not have brain imaging (P<0.0001).

DISCUSSION
We defined patients as having had a CNS complication if they underwent diagnostic brain imaging for any indication during the first year after HCT in this study. Based on this criterion, we report neurological complications in 24% of patients in the first year following HCT, an incidence similar to published results [2][3][4][5][6]8]. This method of determining the incidence of neurological complications eliminates selection bias in chart reviews and inconsistencies in hospital coding. However, a clear limitation is that the decision to obtain brain imaging is solely at the clinical discretion of the transplant physician. There may be practitioner-related differences between transplant centres or even at the same centre at various times. Minor CNS complications where brain imaging may be considered unnecessary would be missed and this method also fails to capture peripheral nervous system complications of transplantation. Despite these limitations, our result with the multivariate model is similar to published work reporting a higher incidence of neurological complications in unrelated-donor transplants [3].
Descriptive survival analysis in HCT studies generally use time-to-event measurements, from the date of transplant to the time of death or last contact. However, analysis of our study was complicated by the fact that brain imaging occurred at varying times after the date of transplant. To address this methodological complexity, we included time-dependent covariates in assessing the effect of brain imaging on survival. We also employed Landmark Kaplan-Meier analysis starting one year post-transplant on patients surviving one year to ensure that all relevant diagnostic brain imaging studies had been completed by the start date of descriptive survival analyses.
Consistent with the HCT literature, we found higher mortality with cord blood and unrelateddonor grafts compared to sibling-donor transplants. There was significantly higher mortality in patients with active disease or >1 relapse at the time of HCT. Nevertheless, the greatest difference in overall survival was in patients who had neurological complications. Similar higher mortality has been reported in retrospective clinical studies [4,6]. Central nervous system (CNS) pathology such as intracranial haemorrhage, abscess, and CNS metastases were considered the major cause of death in 17% autopsy series of HCT patients [9,10]. Clearly, serious structural brain lesions contribute to the mortality of transplant patients. However, many of the neurological complications in HCT patients (approximately 50% in the more inclusive clinical series) are toxic-metabolic encephalopathy or other non-structural CNS complications [8]. Only a third of the 128 patients who had brain imaging in our study had structural abnormalities found on brain imaging (cerebrovascular complications in 10 patients, CNS infection in 9, subdural fluid collection in 6, CNS tumour recurrence in 11, and drug toxicity in 5 patients) [7]. The relatively high mortality in the group of patients without brain structural abnormality may appear unexpected, because most toxic-metabolic encephalopathy that may prompt clinicians to obtain brain imaging is not in itself lifethreatening. However, as recent studies have shown, instances of severe metabolic derangement associated with encephalopathy do occur in transplant patients [11,12].
The association of early brain imaging with decreased overall survival that we report here is analogous to published studies of poor prognostic implications of pulmonary and hepatic abnormalities in HCT recipients [13,14]. Results in our patients clearly show that the clinical indication for brain imaging was itself a strong predictor of shorter survival. It is likely that by selecting patients with symptoms and signs sufficient to prompt the transplant physician to order brain imaging, we have identified patients who have evolved or are evolving a problem-laden post-HCT course with an ultimately poorer prognosis.

CONCLUSION
These results suggest that development of neurological symptoms or a sign sufficient to prompt clinicians to order brain imaging early after HCT identifies a poor prognosis in transplant population.

CONSENT
Not applicable.

ETHICAL APPROVAL
This retrospective study was approved by Institutional Review Board (IRB).