PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
 
Ann Hematol. Author manuscript; available in PMC 2012 April 1.
Published in final edited form as:
PMCID: PMC3053415
NIHMSID: NIHMS265166

Prognostic factor analyses of myeloma survival with intergroup trial S9321 (INT 0141): examining whether different variables govern different time segments of survival

Abstract

Multiple myeloma (MM) survival plots usually display steeper initial and shallower subsequent slopes reflecting differences in disease biology and likely prognostic factors (PF). S9321 trial was selected to determine PF operative at baseline and subsequent 3, 4, 5, and 7-year landmarks (LM-0, LM-3, LM-4, LM-5, and LM-7). With a median follow-up of 8.2 years, survival was similar in transplant and standard therapy arms, justifying data pooling. Median survival for 775 eligible patients is 48 months. According to proportional hazards models, seven of 12 investigated baseline variables retained independent significance for LM-0, of which only two (beta-2-microglobulin and age) extended out to LM-7; the remaining five comprised features of disease aggressiveness (lactate dehydrogenase, calcium, platelet count, C-reactive protein) and host co-morbidity (performance status). Our observations of LM dependency of PF can be exploited toward advancing myeloma therapy by stratifying patients according to whether early or late portions of the survival history are being targeted.

Keywords: Myeloma, Transplant trials, IFM trials, Total therapy trials, SWOG 9321

Introduction

The prognosis of patients afflicted with multiple myeloma (MM) varies widely and depends on baseline laboratory features related to tumor burden [bone marrow plasmacytosis, anemia, extent of bone disease, beta-2-microglobulin (B2M)], disease aggressiveness [lactate dehydrogenase (LDH); the presence of cytogenetic abnormalities (CA) and, more recently, molecular genetics such as interphase fluorescence in situ hybridization (FISH) and gene expression profiling (GEP); cytokine activity such as interleukin-6 (IL6) reflected by levels of C-reactive protein (CRP) and albumin], and host features [age, performance status (PS), renal and cardio-pulmonary functions which can be aggravated by M-protein products causing AL or light chain deposition disease] (for review, see [1, 2]). The recently introduced International Staging System (ISS) classification combining serum levels of B2M and albumin has been widely adopted [3] and has replaced the first staging system named after Durie and Salmon [4]. Still, within tumor stages, considerable variation remains in clinical outcomes which have been extended markedly over the past 5 to 10 years as a result of autotransplant-supported high-dose melphalan treatment [59] and the more recent introduction of novel agents such as thalidomide, lenalidomide, and bortezomib employed together with dexamethasone, each other, and with standard genotoxic agents [1014].

Kaplan–Meier survival plots often follow biphasic or triphasic patterns, raising therefore the issue as to whether different baseline parameters might govern the different time segments of survival outcomes. Here we have examined survival outcomes on Inter-group trial INT-141 [Southwest Oncology Group (SWOG) protocol S9321] [15] according to different landmarks to investigate whether the differently shaped segments of the overall survival curves are “governed” by different sets of baseline variables.

Patients and methods

The protocol details of S9321 have been updated recently [16], and the median follow-up is now 8.2 years. As the two randomized study arms (standard chemotherapy, single autotransplant) exhibited super-imposable overall survival curves, they were collapsed for the purpose of this investigation. Patients on the allotransplant arm were excluded. Survival curves were generated using the product-limit method [17] and univariate and multivariate regression analyses were performed using the Cox proportional hazards model [18], with baseline prognostic factors dichotomized at established cut-points. A recently developed method for deciding on the number of exponential curves that best fit an observed survival curve, and the placements of the change-points for those exponentials [19], was used to influence the best landmark values for further analyses. Hazard ratios over time were estimated by dividing estimates of hazard rates [20] for those over and under prognostic factor cut-points. The protocol had been reviewed and approved by individual institutional review boards. All patients had signed a written informed consent before protocol enrollment, in keeping with the Helsinki Declaration.

Results

Overall survival, portrayed in Fig. 1, is characterized by slope segments with progressively decreasing steepness, herein modeled for OS by two such segments with a change-point at 52 months, justifying the analyses of serial LM. Univariately significant laboratory variables associated with OS varied with the LM examined (Table 1). Of 12 baseline parameters examined, 10 affected survival significantly in the context of LM-0, with only trends observed for albumin (constituting ISS stage together with B2M) and osteolytic bone lesions (component of the Durie–Salmon staging system). B2M impacted survival from all LM while the prognostic implications of CRP, platelet count, hemoglobin, calcium, and performance status were restricted to LM-0. Age affected prognosis from LM-0, LM-3, LM-4, and LM-7 and LDH from LM-0 and LM-5. In the multivariate model (Table 2), B2M’s consistent survival impact across all LM was confirmed while, of the seven LM-0-significant baseline variables, five did not extend beyond LM-0; older age appeared significant for LM-3 and LM-7 with a strong trend for LM-4. According to R2 statistics, values of nearly 18% and 19%, respectively, for LM-0 and LM-7 suggest that almost 20% of variability in survival outcome can be accounted for. In the case of the other intervening LM analyses, R2 values of less than 10% reflect a relatively poor performance of the multivariate models pertaining to these time segments. Similar conclusions were reached using continuous as opposed to dichotomized cut-points for prognostic factors (data not shown).

Fig. 1
Overall survival from registration to start of protocol therapy. Overall survival fit with two exponentials with a cut-point at 52 months
Table 1
Univariate analysis: results of parameters associated with overall survival applying landmarks at baseline (LM-0) and at 3, 4, 5, and 7 years after treatment start (LM-3, LM-4, LM-5, LM-7)
Table 2
Multivariate analysis: results of parameters associated with overall survival applying landmarks at baseline (LM-0) and at 3, 4, 5, and 7 years after treatment start (LM-3, LM-4, LM-5, LM-7)

To illustrate the impact of the LM variable, Fig. 2 portrays hazard ratio (HR) estimates over time. With the exception of age, HR values declined over time, especially for PS and platelet count. The consistently significant adverse impact of B2M across all LM considered was also reflected in the shallower decline of its HR values. Figure 3 portrays Kaplan–Meier plots dated from early to later LM relevant to the presence of two variables shown to govern all segments of survival history (B2M, age) or LM-0 only (calcium, LDH). In the case of B2M and age, the risk factor number-dependent progressively shorter survival pertained across LM (left panels) whereas, for calcium and LDH, such prognostic discrimination was lost beyond LM-0 (right panels).

Fig. 2
Hazard ratio (HR) values of the significant prognostic factors (PF) are portrayed over time
Fig. 3Fig. 3
Kaplan–Meier plots from treatment initiation according to two variables affecting different segments of survival: Left panels serum B2M and age affected survival independently across all landmarks (LM) examined; the best outcome was observed when ...

Discussion

To our knowledge, this is the first report to have examined the issue of whether different baseline prognostic variables “govern” different portions of Kaplan–Meier survival plots. Applying progressively increasing LM from LM-0 to LM-7 in this protocol with a median follow-up of live patients of 8.2 years, we indeed discovered that of seven independently significant parameters affecting survival when examined from LM-0, only two carried forward to later LM (B2M and age). Thus, the rapid loss of prognostic implications from LM-3 onward of calcium, LDH, CRP, platelet count, and performance status is consistent with a rapid attrition of such patients due to disease aggressiveness (calcium, LDH, CRP, platelet count) or co-morbidities (reflected in the performance status). Conversely, B2M governed all phases of the survival curve, reflecting the importance of initial tumor burden (and yet uncovered mechanisms) for the entire patient lifespan.

Conclusion

Multiple myeloma survival plots usually display steeper initial and shallower subsequent slopes, reflecting differences in disease biology, which in turn may be governed by different baseline prognostic factors. Applying serial landmark techniques to 775 newly diagnosed patients enrolled in SWOG trial S9321 with a median follow-up of more than 8 years, LDH and calcium governed the early survival phase, while B2M was the sole feature that retained independent significance for later survival from 3, 4, 5, and 7-year landmarks.

We are presently investigating whether such LM-dependent prognostic variables also apply to more contemporary and overall more effective therapies as our Total Therapy 2 and 3 protocols [8, 9]. Consideration of potentially LM-dependent variables in clinical trial design should aid in identifying therapeutic interventions specifically targeting LM-0-restricted highly aggressive disease, while reserving longer outcome-targeting strategies, such as reducing tumor burden, to later phases of therapy.

Acknowledgment

This study was supported in part by the following PHS Cooperative Agreement grant awarded by the National Cancer Institute, DHHS to the Southwest Oncology Group: CA38926.

Footnotes

Financial disclosures None

Contributor Information

Bart Barlogie, Myeloma Institute for Research and Therapy, University of Arkansas for Medical Sciences, 4301 West Markham #816, Little Rock, AR 72205, USA, barlogiebart/at/uams.edu.

Vanessa Bolejack, Cancer Research and Biostatistics, Seattle, WA, USA.

Michael Schell, Moffitt Cancer Center, Tampa, FL, USA.

John Crowley, Cancer Research and Biostatistics, Seattle, WA, USA.

References

1. van Rhee F, Anaissie E, Angtuaco E, et al. Chapter 109. Myeloma. In: Kaushansky K, editor. Williams hematology. 8th edn. New York: McGraw-Hill; 2010. (in press)
2. Kyle R, Rajkumar SV. Multiple myeloma. N Engl J Med. 2004;351:1860–1873. [PubMed]
3. Greipp PR, San Miguel J, Durie BGM, et al. International staging system for multiple myeloma. J Clin Oncol. 2005;23:1–9. [PubMed]
4. Durie BGM, Salmon SE. A clinical system for multiple myeloma. Correlation of measured myeloma cell mass with presenting clinical features, response to treatment and survival. Cancer. 1975;36:842–854. [PubMed]
5. Attal M, Harousseau JL, Facon T, et al. Single versus double autologous stem-cell transplantation for multiple myeloma. N Engl J Med. 2003;349:2495–2502. [PubMed]
6. Child JA, Morgan GJ, Davies FE, et al. High-dose chemotherapy with hematopoietic stem-cell rescue for multiple myeloma. N Engl J Med. 2003;348:1875–1883. [PubMed]
7. Cavo M, Tosi P, Zamagni E, et al. Prospective, randomized study of single compared with double autologous stem-cell transplantation for multiple myeloma: Bologna 96 clinical study. JCO. 2007;25:2434–2441. [PubMed]
8. Barlogie B, Tricot G, Rasmussen E, et al. Thalidomide and hematopoietic-cell transplantation for multiple. N Engl J Med. 2006;354:1021–1030. [PubMed]
9. Nair B, van Rhee F, Shaughnessy JD, et al. Superior results of Total Therapy 3 (2003–33) in gene expression profiling-defined low-risk multiple myeloma confirmed in subsequent trial 2006–66 with bortezomib, lenalidomide and dexamethasone (VRD) maintenance. Blood. 2010;115:4168–4173. [PubMed]
10. Singhal S, Mehta J, Desikan R, et al. Antitumor activity of thalidomide in refractory multiple myeloma. N Engl J Med. 1999;341:1565–1571. [PubMed]
11. Richardson PG, Blood E, Mitsiades C, et al. A randomized phase 2 study of lenalidomide therapy for patients with relapsed or relapsed and refractory multiple myeloma. Blood. 2006;108:3458–3464. [PubMed]
12. Palumbo A, Bringhen S, Liberati A, et al. Oral melphalan, prednisone, and thalidomide in elderly patients with multiple myeloma: updated results of a randomized controlled trial. Blood. 2008;112:3107–3114. [PubMed]
13. Rajkumar SV, Jacobus S, Callander N, et al. Lenalidomide plus high-dose dexamethasone versus lenalidomide plus low-dose dexamethasone as initial therapy for newly diagnosed multiple myeloma: an open-label randomized controlled trial. Lancet Oncol. 2010;11:29–37. [PMC free article] [PubMed]
14. San Miguel JF, Schlag R, Khuagena NK, et al. Bortezomib plus melphalan and prednisone for initial treatment of multiple myeloma. N Engl J Med. 2008;356:906–917. [PubMed]
15. Barlogie B, Kyle RA, Anderson KC, et al. Standard chemotherapy compared with high-dose chemoradiotherapy for multiple myeloma: final results of phase III US intergroup trial S9321. JCO. 2006;24:929–936. [PubMed]
16. Barlogie B, Attal M, Crowley J, et al. Long-term follow-up of autotransplantation trials for multiple myeloma: update of protocols conducted by the Intergroupe Francophone du Myelome, Southwest Oncology Group and University of Arkansas for Medical Sciences. JCO. 2010;28:1209–1214. [PMC free article] [PubMed]
17. Kaplan EL, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc. 1958;53:457–481.
18. Cox DR. Regression models and life tables. J R Stat Soc B. 1972;34:187–202.
19. Han G, Schell MJ, Kim J. Improved survival modeling using a reduced piecewise exponential approach. J Am Stat Assoc. 2010 (in press) [PMC free article] [PubMed]
20. Mueller HG, Wang JL. Hazard rate estimation under random censoring with varying kernels and bandwidths. Biometrics. 1994;50:61–76. [PubMed]