|Home | About | Journals | Submit | Contact Us | Français|
It is unknown if diminished pulmonary function early after allogeneic hematopoietic transplant is associated with poor long-term outcomes.
To determine if posttransplant lung function is associated with 5-year non-relapse mortality and the development chronic graft-versus-host disease
Retrospective analysis of 2158 patients who had routine pulmonary function testing 60–120 days after transplant between 1992 and 2004. Cox regression was used to assess the hazard ratio for 5-year non-relapse mortality. A second analysis assessed the hazard ratio for the development of chronic graft-versus-host disease.
Lung function score was the primary exposure and was calculated according to FEV1 and DLCO. Individual pulmonary function parameters were secondary exposures. The primary outcomes were 5-year non-relapse mortality and the development of chronic graft-versus-host disease.
Most patients had normal lung function following transplant. A higher lung function score, signifying greater impairment, was associated with an increased risk of mortality [category 1 HR 1.47 (1.17–1.85); category 2 HR 3.38 (2.53–4.53); category 3 HR 7.80 (4.15–14.68)]. A similar association was observed for all individual pulmonary function parameters. Low FEV1 was associated with the subsequent development of chronic graft-versus-host disease [FEV1 70–79% HR 1.26 (1.01–1.57); 60–69% HR 1.48 (1.10–2.01); < 60% HR 2.02 (1.34–3.05)]. Models using either lung function score or individual pulmonary function parameters performed about equally well as judged by the c-statistic.
Impaired lung function at day 80 posttransplant was associated with a higher risk of non-relapse mortality. A low FEV1 following transplant was associated with developing chronic graft-versus-host disease within one year.
Pulmonary complications have been observed in 40 to 60% of patients after allogeneic hematopoietic cell transplant (HCT). Complications may be infectious or noninfectious such as pulmonary edema, diffuse alveolar hemorrhage, idiopathic pneumonia syndrome, or bronchiolitis obliterans (1). Pulmonary complications have been associated with decreased posttransplant pulmonary function, including reductions in forced expiratory volume in 1 second (FEV1), forced vital capacity (FVC), total lung capacity (TLC), or a decrease in diffusion capacity for carbon monoxide (DLCO) (2–11). Many factors may affect pulmonary function after HCT including the underlying disease process, conditioning regimen, infections, and the development of acute or chronic graft-versus-host disease (GVHD) (7–9, 12)
While previous studies have characterized the clinical scenarios associated with decreased lung function after transplantation, most cohorts were small and primarily descriptive. One large study published in 1995 reported a two-fold increase in risk of non-relapse mortality associated with development of restrictive lung defects after transplantation, but restrictive lung disease was not associated with developing chronic GVHD (5). Despite the limitations of these studies, they suggest that post transplant lung function changes may be useful for assessing a patient’s risk for various transplant related outcomes, such as development of GVHD and mortality.
In 2005, the National Institutes of Health (NIH) Consensus Development Project on Criteria for Clinical Trials in Chronic GVHD proposed new recommendations to improve the diagnosis and grading of chronic GVHD. Calculation of a lung function score (LFS) was recommended to grade the extent of lung function compromise after a diagnosis of chronic GVHD had been established (13). However, the LFS is based upon an algorithm developed originally for grading pretransplant lung function. The relationship between lung function, mortality, and chronic GVHD has not been rigorously analyzed using posttransplant lung function. To address this gap, we conducted a retrospective cohort study of patients who underwent HCT during a 12-year period to examine the relationship of posttransplant pulmonary function testing (PFT) with 5-year mortality. We further evaluated the relationship between posttransplant lung function and the development of chronic GVHD according to NIH criteria.
All patients who had HCT at Fred Hutchinson Cancer Research Center (FHCRC) or Seattle Cancer Care Alliance between January 1992 and December 2004 were potentially eligible. Patients who were younger than 15 years, died before pulmonary function testing (PFT), or did not have PFT were excluded (Figure 1). All clinical data except for chronic GVHD status were prospectively collected and retrospectively analyzed. Chronic GVHD data according to NIH criteria was collected retrospectively. The patient's underlying disease state was categorized as low, intermediate, or high risk as previously described (14, 15). Donor match status was determined according to donor–recipient HLA compatibility. Stem cell sources were classified as bone marrow, peripheral blood stem cell, or a combination of both. Conditioning regimens were classified as nonmyeloablative or myeloablative. Subjects in the myeloablative group were subdivided as receiving either a total body irradiation (TBI) or non–TBI based regimen. Acute GVHD was graded based on stages of organ involvement using standard criteria (15, 16). Ethnicity was self-reported. Using clinical records, all patients were followed from transplant until death or January 04, 2008. This study was approved by the institutional review board at FHCRC.
All pulmonary function testing was performed at our Center, according to American Thoracic Society guidelines (17), using the Sensormedics 2100 (Sensormedics Co., Yorba Linda, CA) from January 1992 to August 1999, and the Sensormedics V-Max 22 with Autobox 6200 from September 1999 to December 2004. Published equations for adults were used to determine predicted values of FEV1, FVC, TLC and DLCO (18). All DLCO measurements were corrected for the hemoglobin measurement obtained closest to the time the diffusion capacity was measured (19). All PFT values, except FEV1/FVC ratio, were expressed as a percentage of predicted values and assessed categorically. PFT categories were defined as normal (≥80%), mildly abnormal (70–79%), moderately abnormal (60–69%) or severely abnormal (<60%). Per NIH recommendations, the lung function score (LFS) was calculated according to the day 80 FEV1 and DLCO, each of which was mapped to a category as follows: (≥80% = 1, 70–79% = 2, 60–69% = 3, 50–59% = 4, 40–49% = 5, and <40% = 6) (13, 14). Scores for FEV1 and DLCO were then summed, and categorized 0 to 3 as defined by NIH recommendations [LFS score 2 = category 0 (normal); LFS score 3–5 = category 1 (mildly abnormal); LFS score 6–9 = category 2 (moderately abnormal); or LFS score 10–12 = category 3 (severely abnormal)].
Chronic GVHD data was available for a subset of the cohort. As part of a separate study, 2602 patients with a history of a myeloablative transplant between 1992 and 2005 had undergone a retrospective chart review to establish a diagnosis of chronic GVHD according to current NIH guidelines (13). Patients who had undergone this chronic GVHD assessment and who had day 80 PFT were eligible for the chronic GVHD analysis.
All analyses were performed using STATA 10.0 (StataCorp, College Station, TX). Two tailed P-values <0.05 were considered statistically significant. All data were analyzed as categorical variables except age (continuous). Robust standard errors using a sandwich estimator were calculated for all analyses. Models were compared using Harrell’s C-statistic. The C-statistic is the proportion of predicted outcomes and observed outcomes that are concordant. The primary exposure was LFS category and primary outcome 5-year non-relapse mortality. Secondary exposures were individual PFT parameters. Cox proportional hazards models were used to evaluate the association between non-relapse mortality and lung function. All multivariable analyses were adjusted for covariates that may be associated with lung function, chronic GVHD, and mortality. These variables included age, sex, disease risk, conditioning regimen, HLA status, acute GVHD, and prior CMV infection in the donor and patient as determined by serological testing. Patients who had recurrent malignancy before PFT were excluded from hazard models (n=61). Survival was censored at 5 years from transplant or at a diagnosis of recurrent malignancy.
A secondary analysis used development of chronic GVHD within one year of PFT as the primary outcome. LFS category was the primary exposure. Patients who had recurrent malignancy before PFT were excluded. Those with a diagnosis of chronic GVHD prior to, or within one week of PFT were also excluded. Survival was censored at death, the onset of recurrent malignancy, or one year following PFT.
Between January 1992 and December 2004, 3548 patients underwent HCT. After excluding subjects < 15 years of age (12%), those who died before day 80 PFT (19%) and those without day 80 PFT data (8%), 2158 patients were included in the analysis (Figure 1). Patients were followed for a median (IQR) of 1312 (251–1826) days. The median (IQR) time from transplant until the day 80 PFT was 78 (76–83) days.
Table 1 summarizes the clinical characteristics of the full cohort and the cohort who underwent an analysis for GVHD. The majority of patients had normal lung function (Table 2). The mean FEV1, FVC, and TLC were 88 ± 15%, 93 ± 16%, and 97 ± 15% respectively. The mean DLCO was 81 ± 18% and was less than 80% in 49% of patients. Twelve percent of patients met criteria for obstructive lung disease, defined by a FEV1/FVC ratio < 0.7. Of patients with obstructive lung disease, the FEV1 was ≥ 80% in nearly one-half of patients and < 60% in 12% of patients. Twelve percent of patients had restrictive lung disease, defined as TLC < 80% predicted. Of patients with restrictive lung disease, the FEV1 was ≥ 80% in 16% and < 60% in 21% of patients. Forty-two percent of patients were in LFS category 0 (Table 2). These patients had both a FEV1 and DLCO ≥ 80% predicted. Forty-two percent of patients were in LFS category 1. These patients had either a mildly abnormal FEV1, DLCO, or both. Nine percent of patients were in LFS category 2, and 1% were in category 3. Patients in LFS category 3 had severely abnormal lung function representing both a FEV1 and DLCO no higher than 60% predicted. Five-year all-cause mortality was 40% and 5-year non-relapse mortality was 20%. Five-year relapse was 28%.
Seventy-two patients (4%) were diagnosed with chronic GVHD before or within one week of PFT, and were excluded from the chronic GVHD analysis. The remaining 1650 patients evaluated for chronic GVHD were similar to the entire cohort except a slightly higher percentage had a bone marrow stem cell source (69%), and very few (<1%) had a non-myeloablative conditioning regimen (Table 1). The distribution of day 80 PFTs for this cohort is summarized in Table 3. Overall 846 (51%) developed chronic GVHD and 726 were diagnosed within 1 year of PFT.
Diminished lung function at day 80 posttransplant was associated with an increased risk of non-relapse mortality at five years as measured by lung function score (Table 4). Patients in LFS category 1 had a nearly 50% increased risk of death. This risk increased to more than 3-fold for patients in category 2 and nearly 8-fold in category 3. All hazard ratio estimates were statistically significant. The trend of increased mortality with increasing LFS category was statistically significant (p < 0.0005). After adjustment for age, sex, disease risk, conditioning regimen, HLA compatibility, acute GVHD, donor and patient CMV status, and pre-HCT lung function, hazard ratios were largely unchanged [category 1 HR 1.37 (1.08–1.75); category 2 HR 3.02 (2.11–4.33); category 3 HR 8.56 (4.54–16.14)]. The trend remained statistically significant (p<0.0005).
There was also a stepwise increase in risk of mortality associated with individual PFT parameters (Table 4). Compared to a normal FEV1, a mildly decreased FEV1 was associated with a 76% increased risk of mortality [HR 1.76 (1.38–2.24)]. The HR increased to 2.33 (1.68–3.25) and 5.79 (4.28–7.85) for a moderately and severely decreased FEV1, respectively. A similar increase in HR was seen for FVC. Although the relationship was not as distinct for TLC categories, hazard ratio estimates for each category were also significantly increased. For DLCO, this association was only significant for categories 2 and 3 [category 2 HR 2.04 (1.57–2.66); category 3 HR 1.96–3.43)]. Similar to LFS, the trend towards increased mortality with decreasing pulmonary function was statistically significant for all PFT parameters (p<0.0005). Figure 2 demonstrates 5-year cumulative incidence of non-relapse mortality as a function of PFT parameters and LFS categories. After adjustment, hazard ratio estimates were largely unchanged and remained statistically significant for all PFT parameters except DLCO category 1. There was no significant relationship between mortality and FEV1/FVC ratio [HR 1.11 (0.92–1.36)].
A moderately or severely decreased LFS category was associated with a significantly increased risk of developing chronic GVHD within one year of PFT [Category 2 HR 1.38 (1.02–1.86); Category 3 HR 2.93 (1.00–8.55) (Table 5). The trend for an increased risk of developing chronic GVHD as LFS increased was also significant (p=0.037). This was primarily due to a strong association between FEV1 and chronic GVHD as DLCO was not associated with chronic GVHD in any category. A mildly abnormal FEV1 was associated with a 26% increased risk of chronic GVHD. This risk increased to 48% for a moderately abnormal FEV1 and was more than 2-fold higher for patients with a severely abnormal FEV1. After adjustment for age, sex, disease risk, conditioning regimen, acute GVHD status, and pre-HCT lung function, hazard ratios were largely unchanged. The trend of an increased hazard ratio with decreasing FEV1 was significant in both adjusted and unadjusted analyses (p<0.0005). Similar to FEV1, there was a significant trend of an increased risk of chronic GVHD with decreasing FVC (p=0.001). Within categories of FVC this association was only significant when FVC was less than 60% predicted [HR 2.39 (95% CI 1.48–3.86)]. TLC was not associated with an increased risk of developing chronic GVHD.
To evaluate the performance of the LFS and individual PFT parameters as potential predictors of 5-year non-relapse mortality and chronic GVHD, we compared the Harrell’s c-statistic for each of the above time to event multivariate models. Overall, models using either the LFS or individual PFT parameters performed similarly. With respect to adjusted 5-year mortality, the C-statistic was 0.74 for the LFS, FEV1, and FVC models and 0.73 for TLC and DLCO models. C-statistics were also similar between adjusted chronic GVHD models (C-statistic = 0.60 for LFS and FEV1; C-statistic = 0.59 for FVC, TLC, DLCO).
Although previous studies have evaluated lung function after HCT (2–11), few studies have critically examined whether lung function observed within the first 100 days after transplantation is associated with long term clinical events. Instead, several previous studies have suggested that pulmonary function abnormalities within the first 6 months after transplant may be the result of peritransplant events, may be reversible, and may have no prognostic value (2). Our results show the contrary, indicating that diminished lung function at day 80 following transplant was significantly associated with poor long term outcomes. Most importantly, patients with the most severely abnormal lung function, as measured by the LFS, had a nearly 8-fold higher risk of 5-year non-relapse mortality. An increased risk was also seen with individual pulmonary function parameters. Our results also show that early posttransplant pulmonary dysfunction may identify patients at risk of developing chronic GVHD. While this relationship was present for the LFS, the clearest association was with FEV1. These results suggest that diminished lung function posttransplant should not be dismissed as a transient finding. Moreover, routine pulmonary function testing following transplant may identify a group of higher risk patients who need to be followed closely for posttransplant complications and signs of chronic GVHD.
Models using either the LFS or individual PFT parameters performed similarly. These results suggest that individual PFT parameters, especially FEV1, may be as informative as the LFS, making it reasonable to assess spirometry alone for routine monitoring of lung function after HCT. This practice would result in a decrease in cost and increase in accessibility of lung function monitoring, and thus increase the likelihood that monitoring will be adopted more widely as standard clinical practice for the management of all HCT patients.
Crawford et al. previously showed that diminished lung function after transplant was associated with late mortality (5). A restrictive lung defect at day 80 or a decrease in TLC of 15% or greater from baseline was associated with a 2-fold increased risk of mortality. FEV1/FVC ratio and DLCO were not significantly associated with mortality. In our study, all PFT parameters except for FEV1/FVC ratio were associated with mortality. Several reasons might explain why our results differ. Crawford et al. classified an abnormal DLCO as < 80%, a threshold that may have been too high to identify patients with clinically significant pulmonary compromise. Our study was larger and divided patients with an abnormal DLCO as mild, moderate, and severe. We did not find a significant association for patients with a mildly abnormal DLCO (70–80%), but an association was present for a DLCO less than 70%. Furthermore, Crawford et al. described a cohort who had HCT before 1990. Posttransplant care has changed significantly since then. Nonmyeloablative conditioning regimens are now frequently used and associated with fewer pulmonary complications (20), and preemptive antifungal and anti-CMV treatment are also more readily available (21). Both changes have improved outcomes after HCT.
Crawford et al. did not detect an association between TLC and chronic GVHD (5). We also did not find an association between TLC and chronic GVHD. Chronic GVHD of the lungs manifests as bronchiolitis obliterans and airflow obstruction, which may explain why FEV1, and not TLC, was associated with the development of chronic GVHD in our study.
Pulmonary dysfunction following HCT has several possible causes. Patients are at increased risk of respiratory infections due to prolonged immunosuppression (1, 21). Chemotherapeutics may have direct toxic effects such as damage to vascular endothelium or alveoli resulting in a decreased DLCO. Chest wall, mediastinal, or total body irradiation may have short and long-term effects on pulmonary function (22). Acute and chronic GVHD have been previously associated with diminished posttransplant pulmonary function (4, 7, 9). However, previous studies were unable to ascertain whether diminished pulmonary function was the result of chronic GVHD or was an early marker of chronic GVHD. Additionally, these studies were conducted before NIH consensus definitions. Strengths of our study were ascertainment of the date chronic GVHD was diagnosed, restriction of our analysis to patients without a diagnosis of chronic GVHD at day 80, and the use of current NIH criteria for the diagnosis of chronic GVHD.
Our results should be interpreted with some caution. First, day 80 PFTs were not available for the entire cohort. A post hoc analysis of patients who survived to at least 60 days, and therefore could have had PFT but did not (n=527, 15%), showed that 81% of these patients died. It may be that patients missing PFT carried a higher burden of illness and were either unable to have testing, or providers were reluctant to subject them to testing. If true, this practice would tend to bias our results away from the null hypothesis and strengthen our findings. Second, our single center results might not apply to other centers. In the GVHD analysis, we used current NIH diagnostic criteria in an attempt to minimize this limitation. Third, the diagnosis of chronic GVHD according to NIH criteria was retrospective and subject to misclassification; and some patients may have had undiagnosed chronic GVHD when pulmonary function was evaluated. Finally, we were unable to determine the cause of non-relapse mortality as patients are routinely discharged from our Center around day 100. Many deaths then occur outside the Seattle region.
In summary, we have shown that decreased pulmonary function after transplant was associated with an increased risk of 5-year non-relapse mortality. Decreased FEV1 was also associated with the development of chronic GVHD within the first year after the day 80 PFT. Pulmonary dysfunction can be graded with the NIH recommended LFS, but a simple assessment of FEV1 can provide similarly useful clinical information. Future studies should focus on validation of our findings and should examine these associations using additional pulmonary function assessments at one year or later following transplant. In the meantime, these data provide evidence supporting the NIH recommendation that pulmonary function should be routinely monitored after HCT. Patients who have significantly abnormal lung function should be followed closely for complications or signs of chronic GVHD.
Supported in part by a NIH training grant (T32 HL 007287) (E.C.W); ISCIII code PI-061698, ISCIII code BA-06/90061, CIBER of Respiratory Diseases CB06/06/0043, Spain, and RTIC code C03/11, and SOCAP Enfermería y Fisioterapia (M.O.-L., and A.R.-S.); Fundação de Amparo a Pesquisa do Estado de Sao Paulo (06/59475-4) (A.V.); R01 CA106512 (P.V.C); CA 18024 and HL 36444 (P. J. M.); CA 118953-01A1 and CA 78902 (M.E.D.F.); R01 HL 088201 (J.W.C),
The funding sources had no role in the study design, collection, analysis, and interpretation of the data, writing of the report, or in the decision to submit the report for publication.
No authors have any financial relationship with a company that has a direct financial interest in the subject matter or products discussed in the submitted manuscript, or with a company that produces a competing product.