Pre-treatment C-reactive Protein (CRP) is a Predictor for Allogeneic Hematopoietic Cell Transplantation Outcomes

Abstract

We tested the independent prognostic impact of two commonly used biomarkers, C-reactive protein (CRP) and interleukin-6 (IL-6), on allogeneic hematopoietic cell transplantation (HCT) outcomes. Consecutive patients who underwent a uniform reduced intensity regimen of fludarabine, melphalan and alemtuzumab were evaluated. Cryopreserved sera drawn before conditioning was available to measure CRP and IL-6 levels from 81 and 79 patients, respectively. Patients having CRP values above the median of 18.5 mg/L had significantly more grade 3 to 4 hepatic toxicity (P = 0.08), longer HCT hospital stay (P = 0.005), more aGVHD (P = 0.003), greater non-relapse mortality (NRM) (P =0.01) and inferior overall survival (P =0.02). Higher baseline CRP also correlated with more grade 3 to 4 infectious toxicity (P = 0.09). In contrast, pre-HCT IL-6 levels above the median of 78.3 pg/mL, did not confer a statistically significant increased risk of toxicities or mortality. Elevated hematopoietic cell transplantation-comorbidity index (HCT-CI) did not predict for any measure of HCT morbidity. After adjustment for disease status, comorbidity, performance status and age, elevated CRP concentration remained predictive of NRM. CRP holds promise as a readily available serum biomarker that can be measured prior to HCT conditioning to enhance estimates of transplantation tolerance.

Introduction

Numerous advances including, but not limited to, reduced intensity conditioning (RIC), have allowed extension of allogeneic hematopoietic cell transplantation (HCT) to older and less healthy individuals. Even after RIC HCT, morbidity and mortality can be considerable, ranging from 10 – 40% ^1–4. In addition to age, comorbidity scores and performance status have emerged as powerful prognostic factors for HCT complications ^1,2,5,6. For example, the hematopoietic cell transplantation-comorbidity index (HCT-CI) has gained acceptance as a useful tool to measure comorbidity and estimate non-relapse mortality (NRM) ^6,7. More reproducible and less complex tools are still needed to better predict HCT related complications to appropriately balance the benefits of disease control against the serious regimen related morbidity and mortality, particularly for older and/or less fit patients.

Serum inflammatory biomarkers have emerged as powerful prognostic factors for outcome in a variety of settings. C-reactive protein (CRP) and interleukin-6 (IL-6) are two of the most commonly analyzed serum biomarkers. IL-6 is a pro-inflammatory cytokine that mediates systemic inflammation ⁸ whereas CRP is a hepatically synthesized acute phase reactant. Higher serum concentrations of CRP and IL-6 independently predict poor functional status ^9–13, increased risk of cardiovascular outcomes ^14,15 and death ^16–18. The prognostic value of higher CRP levels also applies to cancer patients ^19,20. Preliminary evidence suggests that rising or elevated inflammatory biomarkers after HCT conditioning may predict for HCT related complications^21–24 and relapse ²⁵.

To our knowledge, the overall prognostic importance of inflammatory biomarkers prior to HCT conditioning has not been well-described. We therefore examined the independent impact of pre-conditioning CRP and IL-6 values on HCT outcomes.

Materials and Methods

We analyzed 112 consecutive patients enrolled on a prospective trial employing a RIC allograft with in vivo T-cell depletion for hematologic malignancies ²⁶. Previously, we described the impact of performance status and the Kaplan-Feinstein comorbidity scale on outcome in a subset of these patients ⁵. Details of the conditioning regimen and supportive care have been published previously ²⁶. The preparative regimen consisted of five days of fludarabine 30 mg/m²/day IV (150 mg/m² total) and alemtuzumab 20 mg/day IV (100 mg total) followed by melphalan 140 mg/m² IV. Tacrolimus was administered for post-transplantation immunosuppression. Allografts were obtained from G-CSF mobilized peripheral blood stem cells except occasional unrelated donors requesting a marrow harvest. All patients were hospitalized for the preparative regimen and for the immediate post-transplantation care. Patients were discharged after neutrophil engraftment and if medically stable. The University of Chicago Institutional Review Board approved the protocol, and all patients provided written informed consent.

Pre-transplant Comorbidity Assessment and Performance Score: Comorbidity was tabulated using the HCT-CI as published by Sorror ⁶. Following this classification scheme, scores of 3 or greater are categorized as high comorbidity, and values < 3 are classified as low comorbidity. Pre-HCT performance status (PS) scores used the Eastern Cooperative Oncology Group score of 0 – 4. Protocol eligibility required a score of 2 or less so values are restricted to 0, 1 or 2.

Transplantation-Related Morbidity

Morbidity was measured through day 100 by tabulating duration of initial hospitalization, grade 3 or greater infectious and hepatic toxicities, and acute graft-versus-host disease (aGVHD). Infectious and hepatic toxicities were obtained by reviewing the medical records focusing on grade 3 or greater toxicity following the Common Toxicity Criteria from the National Cancer Institute (http://ctep.cancer.gov/forms/CTCAEv3.pdf). Acute GVHD grading used published criteria ²⁷. Non-relapse mortality (NRM) was defined as death without disease progression.

Biomarkers

Inflammatory biomarkers were assayed from cryopreserved sera obtained prior to transplantation conditioning using enzyme linked immunosorbent assays (ELISA) following the manufacturers’ guidelines. CRP was quantified using the CRP Quantitative Assay (Biomeda ™, Burlingame, CA). The manufacturer’s reference range is 0 to 8.00 mg/L). IL-6 (in pg/mL) was measured using the TiterZyme®EIA (Assay Designs, Inc., Ann Arbor, MI) system.

Statistical Analysis

Wilcoxon rank sum test was used to compare continuous variables between groups because most variables were not normally distributed. Chi-squared or Fisher's exact tests were used to compare proportions between groups. For ease of interpretation, CRP and IL-6 levels were dichotomized at the median, but the effect of log-transformed values on outcome was also explored. The date of the stem cell infusion was considered day 0 for all time-to-event analyses except for the length of hospital stay where the day of hospitalization for conditioning represented day 0. Overall survival (OS) probabilities were estimated using the method of Kaplan and Meier ²⁸, and OS was compared between groups using the log-rank test. Cox proportional hazards models were used for univariate analysis of continuous predictor variables, as well as for multivariate analyses. The estimated hazard ratios (HR) together with the 95% confidence intervals (CI) and p-values are reported based on Cox regression ²⁹. Cumulative incidence estimates were obtained using the competing risks approach where deaths occurring prior to the event of interest (e.g. relapse or morbidity) were treated as a competing event ³⁰. To account for competing risks, Cox proportional hazards models were used as described by Lunn and McNeil ³¹. Statistical analysis used Stata version 9 (College Station, Tx) and R (http://www.r-project.org/).

Results

Table 1 details baseline characteristics. The median age was 52 years (range 19 – 70 years) and AML/MDS was the frequent most indication for HCT. High comorbidity (i.e. HCT-CI ≥ 3) was found in 42% and a PS of 2 in 9%. Median follow-up for survivors was 42 months.

Table 1 .

Baseline Characteristics, N=112

Characteristic	Value	(%)
Median age at transplantation in years	52
≥50 years	67	(60)
Diseases
AML/MDS	65	(58)
NHL and Hodgkins lymphoma	18	(16)
CLL	3	(3)
ALL	4	(4)
CML	4	(4)
Myeloma	3	(3)
Other*	11	(7)
Active Disease at Transplant Donor Type	52	(46)
Matched Related (n=64) or Unrelated (n=39)	103	(92)
1 Antigen/Allele Mismatched Related (n=4) or Unrelated (n=5)	9	(8)
Comorbidity by HCT-CI
0	31	(28)
1–2	34	(30)
≥ 3	47	(42)
Performance Status (PS)
0	76	(68)
1	26	(23)
2	10	(9)
Biomarkers **
CRP, mg/L
median	18.5
Mean	40.5
IL-6, pg/mL
Median	78.3
Mean	123

HCT-CI- Hematopoietic Cell Transplantation-Comorbidity Index

PS- Eastern Cooperative Oncology Group Performance Score

CRP- C-reactive protein

IL-6- Interleukin-6

^*Other disease includes aplastic anemia and/or PNH (N = 3), chemotherapy induced aplasia (N = 1), mast cell leukemia (N = 2), dendritic cell tumor (N = 1), sickle cell anemia (N = 1), myelofibrosis (n = 2), prolymphocytic leukemia (N=1)

^**C-reactive protein and Interleukin-6 were measured in 81 and 79 patients, respectively

Biomarkers

Cryopreserved sera drawn prior to HCT conditioning were available for analysis of CRP levels in 81 patients (72%). Two patients did not have adequate sera to also analyze IL-6 levels (N = 79, 71%). Coefficients of variation for CRP and IL-6 levels were 4.6% and 4.9%, respectively. CRP values ranged from 0.17 to 180 mg/L with a mean of 40.5 mg/L +/ − 1.7 and a median of 18.5 mg/L. IL-6 values ranged from 10 to 2258 pg/mL, with a mean of 123 pg/mL and a median of 78.3 pg/mL.

Elevated CRP levels showed a positive correlation with active disease status at transplantation (P = 0.059) and worse PS (P = 0.019). No correlation was found between CRP levels and HCT-CI (P = 0.64) or recipient age (P = 0.94). IL-6 levels did not correlate with disease status at transplantation (P = 0.60), PS (P = 0.48), older age (P= 0.82), or high HCT-CI (P = 0.24). There was no significant correlation between high HCT-CI and active disease at HCT (P = 0.18).

Outcomes

The median and mean duration of hospitalization for transplantation was 23 and 27 days, respectively. By day 100, the cumulative incidence of grade 3 or 4 infectious toxicities was 50% (95% CI: 40 – 59), and the incidence of hepatic toxicities was 30% (95% CI: 22 – 39). The cumulative day 100 incidence of acute grade II-IV GVHD was 24% (95% CI: 17 – 32). NRM at day 100 and day 180 was 17% (95% CI: 11 – 25) and 23% (95% CI: 16 – 31), respectively. Estimated overall survival (OS) was 83% at day 100 (95% CI: 75 – 89), 78% at day 180 (95% CI: 61 – 78), and 55% (95% CI: 45 – 64) at one year.

Prognostic Factors-Univariate Analysis

Table 2 summarizes the univariate analysis of the effect of standard prognostic factors on transplantation related morbidity and mortality. We did not include donor type in this analysis as our previous analysis using this conditioning regimen showed no difference between unrelated and related donors on outcome (P = 0.99) ⁵.

Table 2.

Univariate Association of Prognostic Factors and Transplantation Outcomes

	Grade 3 or 4 Toxicity by day 100				Duration of Transplant Hospital Stay		aGVHD		Non-Relapse Mortality		Overall Survival
	Infection		Hepatic
Factor
	HR	P	HR	P	HR	P	HR	P	HR	P	HR	P
Age ≥ 50 years	0.84	0.52	1.1	0.81	1.3	0.19	0.98	0.95	2.0	0.06	1.8	0.02
Active Disease	1.2	0.53	1.5	0.30	1.9	0.001	0.70	0.33	2.2	0.03	3.1	<0.001
HCT-CI ≥ 3	0.97	0.92	1.0	0.93	1.4	0.11	0.65	0.27	1.4	0.37	2.0	0.004
PS ≥ 1	1.0	0.91	2.0	0.05	1.4	0.10	2.4	0.02	3.6	<0.001	2.5	<0.001
CRP, above median	1.8	0.09	3.1	0.01	2.0	0.005	4.1	0.003	3.1	0.01	2.0	0.02
IL-6, above median	1.5	0.24	1.9	0.14	1.3	0.27	1.3	0.75	1.7	0.22	1.3	0.41

aGVHD- Acute graft-versus-host disease

HCT-CI- Hematopoietic Cell Transplantation-Comorbidity Index

PS- Eastern Cooperative Oncology Group Performance Score

CRP- C-reactive protein

IL-6- Interleukin-6

HR - Hazard ratio

Toxicity

Elevated CRP above the median was predictive of morbidity and mortality after HCT (figure 1). There was a non-significant trend for an association between elevated CRP values and infectious toxicities (HR = 1.8, 95% CI: 0.9 – 3.4, P = 0.09), and higher CRP values conferred an increased risk of hepatic toxicities (HR = 3.1, 95% CI: 1.3 – 7.4, P = 0.01). Patients with elevated CRP levels also had longer initial transplantation hospitalization (HR= 2.0, 95% CI: 1.2 – 3.1, P = 0.005) and developed more acute grade II-IV GVHD (HR=4.1, 95% CI: 1.6 – 10.4, P = 0.003). Only 8 patients out of 112 suffered grade III to IV aGVHD, and five of the six with evaluable CRP data had an elevated CRP value. In contrast as shown in table 2, elevated IL-6 levels prior to HCT conditioning were not associated with worse outcome. Worse PS at the time of transplantation was associated with more hepatic toxicity (P = 0.05) and aGVHD (P = 0.016), but not with prolonged hospital stay or more frequent infections. However, high HCT-CI comorbidity score was not associated with an increased risk of developing transplantation morbidity.

Figure 1.

Non-relapse mortality (NRM) and Overall Survival (OS)

Univariate analyses of potential predictors of NRM and OS are reported in Table 2. Patients with elevated CRP values were more likely to suffer NRM (HR 3.1, 95% CI: 1.3 – 7.5). The association with NRM persisted when the log-transformed CRP variable was used (P = 0.03). The cumulative incidence of NRM by day 180 for patients with higher CRP levels was 32% as opposed to only 10% for those with lower CRP levels (P = 0.013). By contrast, high IL-6 levels were only marginally associated with greater NRM (P = 0.22). Similarly, there was no association between log-transformed IL-6 and NRM (P = 0.63). IL-6 values above the median only did not reveal a statistically significant trend for higher day 180 NRM (P = 0.22). In unadjusted analyses, elevated CRP values predicted for inferior OS (P = 0.02) whereas IL-6 levels did not (P = 0.41). In an additional analysis of patients with both CRP and IL-6 measurements (n = 79), 32% of evaluable patients had elevation of both biomarkers above the median (data not shown). This dual elevated biomarker phenotype was strongly associated with an increased risk of NRM (P = 0.0087). The weak correlation of the biomarkers with active disease at HCT did not translate into increased disease-related deaths (P = 0.42 for elevated CRP levels and P = 0.98 for elevated IL-6 levels) suggesting prognostic relevance primarily restricted to transplantation related toxicity rather than relapse.

Among other prognostic factors, worse performance status (P = 0.0001), and active disease status (P =0.03) were associated with NRM whereas high HCT-CI score, older age, and mismatched donor at HCT were not. All prognostic markers showed strong associations with inferior overall survival, with the exception of 1 antigen/allele mismatch donors, possibly related to the small number of patients in this group.

Multivariate Analysis

The independent effect of each biomarker on NRM and OS was assessed by adjusting for age, HCT-CI, PS, and active disease at transplantation (table 3). Donor mismatch was excluded due the low number of patients (n = 9). Elevated CRP concentration continued to predict NRM (P = 0.04) and trended toward reduced OS (P = 0.09). IL-6 levels showed no significant effect on NRM or OS. Dual biomarker elevation (i.e. both IL-6 and CRP values above the median) was associated with a significantly heightened risk of NRM in multivariate analysis (P = 0.007).

Table 3.

Multivariate Analysis of Prognostic Factors

	Non-relapse Mortality		Overall Survival
Characteristic	HR	P value	HR	P value
Age ≥ 50 years	1.2	0.68	1.2	0.63
Disease Status (active)	1.6	0.24	2.6	0.002
HCT-CI ≥ 3	1.0	0.98	1.8	0.06
PS ≥ 1	2.1	0.08	1.5	0.07
CRP above median*	2.5	0.04	1.7	0.09
IL-6 above median*	1.2	0.19	1.2	0.50

^*Multivariate models include one biomarker, i.e. either CRP or IL-6

HCT-CI- Hematopoietic Cell Transplantation-Comorbidity Index

PS- Eastern Cooperative Oncology Group Performance Score

CRP- C-reactive protein

IL-6- Interleukin-6

Discussion

Advances such as RIC and better supportive care have led to a gradual loosening of HCT eligibility criteria, and an expanding pool of older and less healthy HCT candidates ³². Even after RIC in relatively select individuals, NRM varies widely ^1–4,33. The dilemma has generated widespread interest in more accurate pre-transplantation markers of tolerance and NRM. Several studies have shown that comorbidity and performance status are useful, though crude tools ^1,2,5,6, suggesting the need for additional prognostic markers of tolerance.

Inflammatory biomarkers such as CRP and IL-6 have emerged as promising prognostic factors for numerous health outcomes in non-transplantation cohorts. HCT studies have focused on inflammatory biomarkers post-conditioning and/or serially evaluated to predict outcome ^34,35. Higher levels post-HCT have been associated with increased transplantation related complications in some studies ^34,36,37 and with worse survival in others ^24,38. Schots and colleagues analyzed IL-6, interleukin-8 and tumor necrosis factor-alpha levels serially in 84 patients after HCT. Elevated inflammatory markers early after transplantation predicted for major complications ²¹. Min and colleagues described an association between lower mean CRP values after HCT with less GVHD and more disease relapse ²⁵. In the largest series to date, Fuji and investigators demonstrated among 224 patients, that higher peak CRP values after conditioning correlated with more acute GVHD, infections, and NRM ²⁴. But decisions about whether to proceed to HCT as well as modifications of the conditioning regimen or treatment plan mandate pre-treatment risk assessment rather than biomarker kinetics following conditioning.

To our knowledge, this is the first study to assess the independent prognostic value of pro-inflammatory biomarkers measured prior to HCT conditioning on outcome. We found that elevated levels of CRP, but not IL-6, predicted for increased morbidity and mortality. Higher CRP levels robustly predicted for worse tolerance including hepatic toxicity, length of transplantation hospital admission, aGVHD, NRM, and OS. Higher CRP also tended to track with severe GVHD. Six patients had grade III to IV aGVHD and baseline CRP measured; an elevated CRP level was found in five of the six. A statistically non-significant trend also existed for infectious toxicities (P = 0.09). Further, the impact on NRM persisted after adjusting for other standard prognostic variables, including the HCT-CI. In contrast, elevated IL-6 levels only showed a non-significant trend for worse outcomes. The biomarkers’ impact appeared restricted to toxicity and death due to NRM; neither biomarker influenced death from disease relapse. Thus, CRP may be a relatively specific marker of HCT tolerance.

The cause for worse outcomes among patients having elevated baseline CRP remains unclear. In addition to measuring the inflammatory cascade, CRP has direct pro and anti-inflammatory properties ^39,40. One could postulate that pre-transplantation inflammation increases susceptibility to regimen-related toxicity and to the deleterious effects of GVHD. If this hypothesis is correct, then anti-inflammatory strategies may be beneficial for patients with elevated baseline biomarkers. For example, HMG CoA reductase inhibitors (a.k.a. statins) may have a beneficial immunomodulatory profile and statins have recently been suggested to reduce GVHD after HCT ⁴¹. Previous studies using pentoxifylline as an anti-inflammatory strategy have been very disappointing, but were applied indiscriminately and therefore may not have been focused on a high-risk population ^42,43. A simpler albeit incomplete concept may be that CRP represents a general measure of health status. For example, CRP values, independent of comorbidity, are associated with disability and diminished function in older adults ^44,45. The correlation between higher CRP values and worse PS (P = 0.019) but not comorbidity or age further supports this hypothesis. Irrespective, this readily available and inexpensive biomarker predicted NRM independent of standard prognostic factors, suggesting a promising role in pre-transplantation evaluation. More puzzling is why IL-6 did not confer similar adverse risks as elevated CRP levels since IL-6 serves as a potent and rapid stimulus of CRP synthesis ⁸. The better predictive capacity of CRP levels may reflect the specific and amplified role of CRP as a downstream product of inflammation. IL-6 is a highly regulated pro-inflammatory cytokine exhibiting pleiotripic effects ⁴⁶. IL-6 has multiple stimuli, changes less dramatically in response to inflammation, and exhibits a diurnal variation, all of which may obscure the prognostic impact ⁴⁷.

The largest study limitation relates to the generalizability since we used a non-myeloablative T-cell depleted regimen. Other limitations should be noted. The normal range of CRP depends on the population evaluated, with levels from less than 3 to less than 10 mg/L considered normal ^48,49. In this study, the median level was quite high at 18 mg/L. In a recent study evaluating CRP kinetics among HCT patients, the median baseline CRP levels were much lower at 3 mg/L although the baseline range of 0 – 200 mg/L was comparable to that found in our patients of 0.17 to 180 mg/L²⁴. Likewise, IL-6 values were markedly elevated in our study with a median of 78 pg/mL compared to median values of 2 – 4 pg/mL derived from epidemiologic surveys in other cohorts ^15,16,50. For clinical simplicity in this exploratory analysis, biomarker values were dichotomized above and below the median. Refining thresholds or categorization schemes (e.g. quartiles) in further studies may allow better prognostic discrimination. We believe analyzing levels at a standard time point among a homogeneous group of patients may further clarify the prognostic potential. For example, analyzing CRP levels evaluated two weeks after achieving first remission among HCT eligible AML patients. Although IL-6 levels did not afford much value as a single predictor, a strong association with NRM, beyond that found for higher CRP levels alone, was detected among patients having higher levels of both IL-6 and CRP. For this reason, future validation studies should include levels of IL-6 as well as CRP levels. Numerous other inflammatory markers studied in other settings such as cystatin-C, pro-calcitonin, serum amyloid A, tumor-necrosis factor-alpha, and interleukin-8 also warrant investigation.

Because we measured multiple domains of toxicity, our study supplements the growing literature on prognostic factors for HCT tolerance. We again note, as previously reported, that worse performance status predicts for NRM and OS ⁵. We also previously found that higher comorbidity, as measured by the Kaplan-Feinstein scale, functioned as a weak prognostic factor. Sorror and colleagues recently developed the HCT specific comorbidity index (HCT-CI) which is a more sensitive comorbidity measure than the Kaplan-Feinstein scale, leading to inclusion of the HCT-CI in this study ^6,7. In a recent update of 408 patients undergoing nonmyeloablative conditioning, the Seattle group confirmed high HCT-CI scores and worse PS adversely impacted morbidity, NRM, and OS ⁵¹.

Surprisingly, we found no significant predictive value of the HCT-CI for toxicity or NRM (HR = 1.4, P = 0.37). The cause for these discordant results remains unclear although preliminary data from another institution also found that PS, but not high HCT-CI, predicted for NRM ⁵². We have recently shown that the addition of alemtuzumab for in vivo T-cell depletion to fludarabine and melphalan reduced organ toxicity and GVHD at the expense of relapse compared to the same regimen without alemtuzumab ⁵³. The alemtuzumab containing regimen may have diminished the impact of pre-existing comorbidity and organ dysfunction on NRM ^26,54. This raises the intriguing hypothesis that a low toxicity and low GVHD regimen such as fludarabine, melphalan, alemtuzumab may be beneficial in patients manifesting significant comorbidity. This should not diminish enthusiasm for the HCT-CI; in our series the HCT-CI still predicted for inferior OS and may be valuable for anticipating tolerance for regimens incorporating standard GVHD prophylaxis.

In conclusion, elevated levels of CRP but not IL-6, prior to HCT conditioning, independently predicted for increased toxicity and NRM after allogeneic transplantation using a reduced intensity conditioning regimen. The data warrant confirmation in patient populations receiving other conditioning regimens and methods of GVHD prophylaxis. If validated, CRP could be rapidly integrated into risk assessment tools before transplantation and may provide a simple and inexpensive test to promote comparisons across study cohorts and within registry studies. We envision that future research will allow precise determination of transplantation tolerance based upon three domains: biomarkers, comorbidity, and functional status. Arming clinicians and patients with a reliable estimate of morbidity and mortality should facilitate transplantation decision making and should permit testing therapeutic strategies that optimally balance HCT related morbidity and mortality against disease relapse.