Inflammatory Markers and Development of Symptom Burden in Patients with Multiple Myeloma during Autologous Stem Cell Transplantation

High-dose chemotherapy followed by autologous hematopoietic stem cell transplant (AuSCT) has become part of the standard treatment regimen for patients with multiple myeloma (1). As aggressive cancer therapies, both AuSCT and allogeneic stem cell transplant (AlloSCT) are associated with the occurrence of high symptom burden in the initial weeks posttransplant, including fatigue, pain, disturbed sleep, poor appetite, and drowsiness (2, 3), despite the differences in therapeutic agents and stem cell sources. These symptoms are difficult to manage, due in part to limited knowledge of the pathophysiology underlying their development.

Mounting evidence from both animal models and human research suggests that dysregulation of inflammatory processes may play a pivotal role in producing chronic systemic conditions, such as poor appetite, pain, fatigue, and disturbed sleep (4–6). Our studies of patients who rapidly developed severe symptoms in response to aggressive cancer therapy (chemoradiation for gastrointestinal or lung cancer and AlloSCT for hematologic malignancies) have shown that a significant increase in serum inflammatory cytokines such as interleukin (IL)-6 and soluble tumor necrosis factor receptor 1 (sTNF-R1) is temporally associated with the development of severe symptoms during treatment (3, 7, 8). Consistent behavioral observations in these clinical studies parallel the cytokine-induced sickness behaviors, such as hyperalgesia, disturbed sleep, reduced water and food intake, and lack of activity, seen in animal models (5).

In the current study, we studied patients with multiple myeloma undergoing high-dose chemotherapy with AuSCT by temporally tracking changes in inflammatory markers and their relationship with the development of severe symptoms. AuSCT presents a unique opportunity for studying the cytokine dysregulation induced by high-dose chemotherapy without interference from the acute graft-versus-host response associated with AlloSCT (3).

Patients were enrolled from 2008 to 2011 in the outpatient clinic of the Department of Stem Cell Transplantation at The University of Texas MD Anderson Cancer Center in Houston, Texas. Eligible patients had a confirmed pathologic diagnosis of multiple myeloma, were at least 18 years old, and most were scheduled to receive high-dose melphalan (200 mg/m²) as their conditioning regimen for AuSCT. Patient characteristics and clinical parameters [age, sex, cancer stage, Eastern Cooperative Oncology Group performance status (ECOG PS; ref. 9, Charlson comorbidity score (10), induction treatment history, and body mass index (BMI)] were recorded during the study.

The study was approved by the MD Anderson Institutional Review Board, and all participants gave written informed consent. The study is registered at ClinicalTrials.gov (NCT00688168).

The MD Anderson Symptom Inventory (MDASI) assesses the severity of 13 common cancer-related symptoms (pain, fatigue, nausea, vomiting, dry mouth, shortness of breath, lack of appetite, difficulty remembering, drowsiness, disturbed sleep, sadness, distress, and numbness) and six items related to symptom interference with functioning (11). For this study, we used the psychometrically validated multiple myeloma module of the MDASI (MDASI-MM), which contains the 19 MDASI items plus 7 multiple myeloma–specific items (bone aches, muscle weakness, sore mouth/throat, rash, difficulty concentrating, constipation, and diarrhea) (12).

Patients rated symptom severity over the previous 24 hours on a 0 to 10 scale ranging from “not present” to “as bad as you can imagine.” MDASI-MM assessments were conducted at baseline (18 days before the transplant; day −18), twice a week from baseline to 30 days after AuSCT (day +30), and then weekly until 100 days after AuSCT (day +100). The assessments were completed on paper or, when the patient was unavailable in person, via phone interview with study staff.

We assessed IL-1β, soluble IL-1 receptor antagonist (sIL-1RA), soluble IL-1 receptors 1 (sIL-1R1) and 2 (sIL-1R2), soluble IL-2 receptor antagonist (sIL-2RA), IL-4, IL-6 and its soluble receptor sIL-6R, IL-8, IL-10, IL-12p70, IL-17, TNF-α, sTNF-R1, sTNF-R2, interferon (IFN)-γ, C-reactive protein (CRP), vascular endothelial growth factor (VEGF), macrophage inflammatory protein (MIP)-1α, and monocyte chemotactic protein (MCP)-1. These inflammatory markers have been identified as potentially relevant in animal and human sickness–behavior studies (3, 5, 6, 13, 14).

Blood draws without anticoagulant were scheduled for the morning on the day of apheresis, at mobilization and conditioning day, twice a week from stem cell transplant (day 0) to day +30, and at routine clinic visits thereafter to day +100. After serum isolation by centrifugation (2–4 hours after phlebotomy), the samples were stored at −80°C for batch analysis. Inflammatory markers were measured in 25 μL of serum using the MILLIPLEX map assays (EMD Millipore Corporation) and a Luminex 100 Analyzer (Luminex Corporation) (15). IL-1β, IL-4, IL-6, IL-8, IL-10, IL-12p70, IFN-γ, and TNF-α were measured using the Millipore magnetic Human High-Sensitivity Cytokine/Chemokine Panel. IL-17, VEGF, MCP-1 and MIP-1α were quantified using the Millipore Human Cytokine/Chemokine Panel 1. The Millipore nonmagnetic Human Soluble Cytokine Receptor Panel was used to measure sIL-1R1, sIL-1R2, sIL-2RA, sIL-6R, sTNF-R1, and sTNF-R2. CRP, a by-product of inflammation, was measured in 1:27,000 diluted serum sample before analysis using the Human Cardiovascular Disease Panel 2. sIL-1RA was measured in 100 μL of serum by enzyme-linked immunosorbent assay (R&D Systems, Inc.).

To analyze the development of patient-reported symptom outcomes in relation to objective measures, we constructed loess curves to depict the mean level of symptom severity, white blood cell (WBC) count, and serum inflammatory marker levels from baseline to day +100 after AuSCT (where day of stem cell transplant was designated as day 0). Nonlinear mixed modeling was used to determine WBC nadir, symptom peaks, and the peaks and nadirs in serum levels of inflammatory markers from baseline to day +30. Statistical significance was set at 0.0038 (0.05/13.00) to adjust for multiple comparisons. We focused on the first 30 days posttransplant because most of the marker samples were collected during that time period and because graphed data for most of the symptoms and markers plateaued after day +30. Mean MDASI-MM symptom severities and mean inflammatory marker concentrations between the day −18 to day +30 period (before and the first month after AuSCT) and the period of symptom and marker plateau (day +31 to day +100) were reviewed using the Wilcoxon rank sum test.

Descriptive analysis was used to present patient characteristics. Markers for which more than 10% of the samples had undetectable values were not included in the analysis. For each marker retained for analysis, samples with undetectable values were imputed to have one-half the lowest nonzero value from the detectable samples for that marker. To give context to the variability of the cytokine assay and symptoms, a waterfall plot was constructed to examine the changes in the five most severe symptoms and in serum IL-6 levels from baseline to nadir (16).

To determine the relationship of the strength of inflammatory changes and increases in symptom severity, mixed-effect models were constructed to identify the critical serum inflammatory marker(s) that were temporally associated with symptoms with dynamic changes in the first 30 days, included as time-dependent variables (17). We used both MDASI-MM single-symptom scores and a mean component score of the five most severe symptoms as outcome measures for each patient. Time-independent variables that might have an impact on symptom or inflammatory response (age, sex, cancer stage, bortezomib induction treatment status, stem cell dose, baseline mood score, and BMI) were included in all models. The baseline mood score was the average of pre-AuSCT MDASI-MM “sadness” and “distress” scores, based on a previous study by McGregor et al. (18) indicating that distress had a significant effect on blood count recovery. Each inflammatory marker was included in a univariate model on a transformed natural log scale. Those that showed significant effects at P < 0.05 in the univariate models were included in a multivariate model for each symptom outcome. Statistical significance for each multivariate model was set as 0.05/n (where n is the number of cytokines included in the model). Estimates were calculated for natural log-transformed inflammatory marker variables to show how the symptom outcome would change when the marker changed by any ratio.

Of 70 eligible patients, 7 declined to participate. Table 1 presents the demographic and clinical characteristics of the remaining 63 participants enrolled pre-AuSCT. Four of the 63 patients withdrew from the study by day +30 and 9 patients withdrew between days +31 and +100. WBC nadir was observed on day +8 (range, 4–10 days). No disease relapse or death occurred during the first 100 days after AuSCT. Most of the patients had received bortezomib-based induction therapy (53, vs. 8 on lenalidomide). Almost all patients received melphalan 200 mg/m² as conditioning; 3 patients also received other agents in combination with melphalan.

The overall missing-data rate for MDASI-MM symptom ratings was 7% between baseline and day +30 and 15% between days +31 and +100, stemming primarily from missed follow-up calls to patients. Comparing average mean MDASI-MM scores from baseline to day +30, the five most severe symptoms were fatigue, disturbed sleep, pain, lack of appetite, and drowsiness; this ranking remained consistent throughout the study.

Loess curves present the average WBC count and the projected severity of these five symptoms on a 0 to 10 scale from baseline to day +100 (Fig. 1). Mixed-effects modeling demonstrated that lack of appetite displayed the largest change over time (linear terms for days since AuSCT, estimate = 3.86, SE of the mean = 0.42, P < 0.0001). We also examined day of peak symptom severity for the five most severe symptoms during the first 30 days after AuSCT (Table 2). The individual symptoms and the component score showed significant peaks around day +8. Pain, one of the most severe symptoms, remained relatively constant over the observation period.

All cytokines that were below the limit of detection in more than 10% of the samples were excluded from the analysis. MCP-1, IL-8, TNF-α, sIL-1RA, sIL-1R2, sIL-2RA, sIL-6R, sTNF-R1, sTNF-R2, and CRP (each undetectable in <2% of the samples), MIP-1α (5.5%), IL-10 (6.5%), and IL-6 (9%) were included in the analyses.

Loess curves depict the average levels of inflammatory markers from baseline to day +100 (Fig. 2). The markers presented in the figure were those found to have significant associations with symptom outcomes in both the univariate and multivariate analyses. Table 3 shows that the 13 testable inflammatory markers reached peak or nadir on various days after AuSCT. Because of these fluctuations, we concentrated on data from baseline to day +30 using nonlinear mixed modeling. Significant peak values were detected for IL-6 on day +9, CRP on day +13, and IL-10 on day +18 (all P < 0.0001). A significant nadir was seen for sIL-6R on day +8 (P < 0.0001) and for sIL-1RA on day +5 (P = 0.003).

No significant peak or nadir was found for IL-8; however, the Wilcoxon rank sum test showed significantly lower levels of IL-8 during the first 30 days after AuSCT than from day +31 to day +100 (P = 0.0003). Significantly higher concentrations of sTNF-R1 (P = 0.006) and sIL-2RA (P < 0.0001) were found in the first 30 days after AuSCT than in the later period. No significant change over time was observed for MCP-1, sIL-1R2, or TNF-α.

Figure 3 presents a waterfall plot of the change scores from baseline to nadir. Most patients had increased symptom burden that was paralleled by increased proinflammatory response. The only outlier had an increased IL-6 level at baseline of 2000.0 pg/mL, corresponding with the patient's baseline clinical status (hypertensive, tachypneic, slightly tachycardic, and hyperglycemic). However, this patient also showed parallel changes of improved conditions with decreased symptoms and decreased cytokines over time.

Univariate analysis identified serum IL-6, sIL-6R, IL-10, CRP, MIP-1α, sIL-1R2, and sTNF-R1 (all P < 0.05) as significantly related to one or more of the five most severe symptoms during the first 30 days after AuSCT (Table 4). Age, sex, BMI, bortezomib versus other induction treatment, stem cell dose, pre-AuSCT mood score, and cancer stage were used as covariates, but there was no observed significant effect on symptom outcomes. The identified markers were retained as candidates for the multivariate analysis.

The multivariate analysis (mixed effect modeling) identified IL-6 (P = 0.001) and MIP-1α (P = 0.013) as the critical inflammatory markers related to the development of the five most severe symptom outcomes during the first 30 days after AuSCT, controlling for the same set of covariates. Worse pre-AuSCT mood score was significantly associated with more severe symptom outcomes (P = 0.003). The other covariates did not have a significant effect.

For individual symptom outcomes, we found significant positive relationships between change in IL-6 concentration level and change in drowsiness (P = 0.009) and between sTNF-R1 (P = 0.004) and IL-10 (P = 0.01) related to higher pain severity; inverse relationships were found between MIP-1α and disturbed sleep (P = 0.0002), and sIL-6R and fatigue (P = 0.039). Pre-AuSCT mood score was the only patient or clinical factor significantly related to changes in symptom severity (fatigue and drowsiness), although younger patients (≤60 years of age) consistently reported more severe pain than did older patients across the observation period (estimate = 0.106; P = 0.007).

The current study suggests that IL-6 plays an important role in the development of high symptom burden during the initial month after stem cell transplant (4, 13, 19, 20). A cluster of symptoms (fatigue, disturbed sleep, lack of appetite, pain, and drowsiness) was consistently the most severe among all MDASI-MM symptoms assessed from day −18 to day +100 of AuSCT. Methodologically, our longitudinal study was sufficiently powered to use mixed-effects modeling to detect the temporal association between symptom scores and expression of inflammatory markers, while controlling for confounding variables (17, 21–23).

A high-dose melphalan conditioning regimen triggers a severe drop in WBC count and a resulting deficiency that could lead to infectious complications. This, along with residual disease, psychologic distress, and prolonged side effects from induction therapy, could contribute to inflammatory and symptom responses. Regardless of the effect of these factors, patients collectively shared not only the same severe symptoms, but also a similar symptom trajectory that was at its worst at WBC nadir after AuSCT. There was no significant change over time in pain severity, but because pain was persistently severe compared with other symptoms, we included it in the symptom component score.

The univariate analysis identified that serum IL-6, sIL-6R, IL-10, CRP, MIP-1α, sIL-1R2, and sTNF-R1 was significantly related to development of the five most severe symptoms during the same period. In examining the timing of peak or nadir of inflammatory marker expression compared with the peak in symptom burden at day +8 during WBC nadir (range, day 4–10), we observed the closest correlations for IL-6 (positively) and IL-6R and MIP-1α (inversely). IL-6 has been identified as a primary marker associated with symptom-burden peak in both AuSCT and AlloSCT, which suggests that its role is more relevant to the nadir effect caused by treatment than to disease during the critical first month posttransplant, although IL-6 is a well-known disease growth factor in multiple myeloma (24). In view of the low WBC count, it is possible that the immune system is not the source of IL-6; IL-6 might alternatively be produced by other organ cells such as hepatocytes or endothelial cells. sIL-1RA had an earlier nadir (day +5) than the symptom peak and a more rapid decline than IL-6, indicating a more sensitive response to the transplant. To our knowledge, our study is the first to identify inverse associations between sIL-6R and IL-6 and between sIL-6R and fatigue severity. sIL6R is known to increase IL-6–induced signaling by forming a complex with IL-6 and gp130 on the membranes of cells. The decrease in the level of sIL-6R may protect the individual from even more proinflammatory signaling and subsequent development of more severe symptoms.

How circulating cytokines signal the brain to trigger multiple severe symptoms in humans is unknown, although in animal studies it has been demonstrated that systemic cytokine production can influence behavior by inducing symptoms of sickness and depression (5, 25). Further, we cannot explain why levels of either TNF-α or its receptors were unchanged at the time point of chemotherapy-induced WBC nadir. Given that the peak response of CRP and IL-10 occurred after the symptom peak, these mediators are unlikely to initiate the development of severe symptoms (26).

The study has several limitations. We designed our study to coincide with patients' routine blood-draw schedules, so as to reduce the likelihood of missing data. Therefore, the timing of blood draws during the day may not have been consistently optimal and could have been affected by circadian patterns of cytokine release. Also, we collected a limited number of samples (totaling approximately 15% of all samples) during the day +30 to +100 period, due to fewer regularly scheduled follow-up visits during this time period. Finally, samples were kept for 2 to 4 hours at room temperature but were not tested for stability thereafter; this approach should be examined in future studies, as many cytokines are unstable. Additional studies might also examine observer-rated transplant-associated toxicities to gain a better understanding of such toxicities and symptom burden.

In summary, defining the temporal development of inflammatory activity in concert with the development of high symptom burden is another step toward establishing a better understanding of the nature of treatment-induced symptom burden. Anti-inflammatory intervention may be further tested in clinical studies to establish the role of inflammation inhibition in relation to amelioration of severe symptoms, so long as this can be done without compromising the efficacy of the antitumor treatment. Reduction of therapy-associated symptoms should make cancer treatment more tolerable and therefore more accessible to those patients who might not otherwise be considered for aggressive cancer therapy.

R.Z. Orlowski has received commercial research grants from Celgene, Bristol-Myers Squibb, Onyx Pharmaceuticals, and Millennium: The Takeda Oncology Company. He is also employed as a consultant/advisory board member with Celgene, Onyx Pharmaceuticals, Millennium: The Takeda Oncology Company, and Bristol-Myers Squibb. No potential conflicts of interest were disclosed by the other authors.

Neither the National Cancer Institute (NCI) nor the NIH had any role in the study design, data collection, analysis, interpretation, or preparation of the report.

Conception and design: X.S. Wang, J.M. Reuben, R.Z. Orlowski, L.A. Williams, C.S. Cleeland

Development of methodology: X.S. Wang, J.M. Reuben, T.R. Mendoza, C.S. Cleeland

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.N. Cohen, J.M. Reuben, R.Z. Orlowski, M.H. Qazilbash, L.A. Williams, C.S. Cleeland

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X.S. Wang, Q. Shi, N.D. Shah, C.J. Heijnen, J.M. Reuben, M.H. Qazilbash, V.E. Johnson, T.R. Mendoza

Writing, review, and/or revision of the manuscript: X.S. Wang, Q. Shi, N.D. Shah, C.J. Heijnen, J.M. Reuben, R.Z. Orlowski, M.H. Qazilbash, L.A. Williams, T.R. Mendoza, C.S. Cleeland

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.M. Reuben

Study supervision: X.S. Wang, L.A. Williams

The authors thank the editorial assistance of Jeanie F. Woodruff, BS, ELS, and data collection and management from Venus M. Ilagan, BA, CCRP, Jackie Joy, Brooke White, Gary M. Mobley, BS, MA, and Mary H. Sailors, PhD.

This study was supported by grants from the NCI of the NIH, including NCI P01 CA124787 (to C.S. Cleeland) and MD Anderson Cancer Center Support Grant NCI P30 CA016672.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.