depicts OS and EFS for all trials reported. In the IFM90, IFM94, S9321, and TT1 trials, survival outcomes extended beyond 10 years. Thus, 10-year OS and EFS estimates were 33% and 14% in TT1, 26% and 9% in IFM94 (both arms combined), 19% and 6% in IFM90 (both arms combined), and 22% and 18% in S9321 (both arms combined), respectively. For the remaining trials, 5-year OS and EFS estimates were 78% and 71% in TT3, 66% and 50% in TT2 (both arms combined), 73% and 34% in IFM9902 (all three arms combined), and 42% and 13% in IFM9904, respectively. Reiterating Kaplan-Meier analyses from early phases through completion of patient accrual and progressively extending follow-up times revealed superimposable OS plots for all trials examined (Data Supplement Fig 1), suggesting consistency of patient characteristics throughout the accrual time periods, patient follow-up, and therapeutic management. Thus, we failed to detect instances where initially more promising results deteriorated with longer follow-up.
Kaplan-Meier plots of (A) overall and (B) event-free survival outcomes. IFM, Intergroupe Francophone du Myelome; TT, Total Therapy; S, Southwest Oncology Group; NR, not reported.
Results of individual randomized trials are portrayed in Data Supplement Figure 2. In IFM90, both OS and EFS remain significantly prolonged for the transplantation arm versus standard treatment arm (Data Supplement Fig 2A), whereas trends remain for both OS and EFS in favor of tandem versus single transplantation in IFM94 (Data Supplement Fig 2B). Both arms of S9321 remain virtually superimposable in terms of OS, but a trend has emerged in favor of transplantation for EFS (Data Supplement Fig 2C). Regarding TT2, a significant advantage in favor of the thalidomide arm has now emerged for OS (P = .04; Data Supplement Fig 2D). However, the original OS advantage of thalidomide in IFM9902 is no longer apparent (comparing the two arms without thalidomide v thalidomide-containing maintenance, P = .39; Data Supplement Fig 2E); the current analysis pertains to the 88% of patients who also had interphase fluorescence hybridization data.
Next, we performed univariate and multivariate analyses to determine which pretreatment parameters and which protocols were significantly linked to OS (). Independent adverse features included advanced age, low albumin and hemoglobin, and high B2M and lactate dehydrogenase (LDH); TT2, IFM9902, and TT3 protocols each conferred lower hazard ratio (HR) values (). Tandem transplantations as a group yielded superior results compared with single transplantations and standard-dose therapies (HR = 0.61, P < .001), and adding thalidomide to tandem transplantations was superior to tandem transplantations without thalidomide (HR = 0.69, P < .001; ).
Univariate and Multivariate Analyses of Baseline and Treatment Features Associated With Survival Outcomes in Context of Individual Trials
Univariate and Multivariate Analyses of Variables Associated With Postrelapse Survival, Including Baseline Characteristics, Length of Initial EFS, and Availability of Thalidomide and Bortezomib
To portray individual trial comparisons, protocol patients were matched on the four variables identified as independently affecting OS on multivariate analysis (albumin < 3.5 g/dL, B2M ≥ 3.5 mg/L, LDH ≥ upper limit of normal, and hemoglobin < 10 g/dL), which was accomplished in three sets of 228 patients each, with one set representing the three trials with independently superior OS (TT3, TT2, and IFM9902; ). The best outcomes were recorded in this category (TT3, TT2, and IFM9902), resulting in 8-year OS and EFS estimates of 62% and 31%, respectively, contrasting with similar inferior outcomes of 36% and 20%, respectively, for the two other categories (TT1; and IFM9904, IFM94, IFM90, and S9321; both P < .001; ).
(A,B) Pair-mate analyses of patients matched on albumin, β2-macroglobulin (B2M), lactate dehydrogenase (LDH), and hemoglobin (Hb). TT, Total Therapy; IFM, Intergroupe Francophone du Myelome; SWOG, Southwest Oncology Group.
Because of the long time span over which the protocols were executed, we examined postrelapse survival (PRS) to account for greater access to novel agents in more recently conducted trials. Thus, later trials' superior OS may have resulted from the availability of better salvage regimens rather than the impact of the original treatment. The following 5-year PRS estimates were recorded: 35% for IFM9902 and TT1, 29% for S9321, 27% for IFM9904 and TT2, 22% for IFM94, and 14% for IFM90 and TT3 (P < .001; ). IFM90 and TT1 participants had least access to thalidomide, which became available in 1997, but PRS after TT1 was far superior to PRS after IFM90. Patients in IFM94 and S9321 should have had equal access to thalidomide, perhaps explaining similar PRS. The availability of bortezomib and lenalidomide for salvage treatment around the year 2000 likely benefited patients treated on TT2, IFM9902, and IFM9904. The short PRS in patients who experienced treatment failure after TT3 may be attributable to the up-front use of all myeloma-active treatment ingredients, thus curtailing salvage efforts at relapse.
Post-relapse survival outcomes. IFM, Intergroupe Francophone du Myelome; TT, Total Therapy; S, Southwest Oncology Group.
We also performed a multivariate analysis to capture variables independently linked to PRS, including baseline characteristics (relapse characteristics were only available in TT trials), individual trials, and the length of preceding EFS (). Older age and higher levels of baseline B2M (> 5.5 mg/L) and LDH had adverse impacts on PRS; preceding EFS equal to or shorter than the median (803 days) was another adverse feature, whereas long EFS (fourth quartile, ≥ 1,280 days) reduced the hazard of PRS. Regarding individual protocols, IFM90 and TT3 both resulted in shorter PRS, whereas IFM9902 trial participation conferred superior PRS. Tandem transplantation significantly reduced the PRS hazard compared with single transplantation and standard chemotherapy, and thalidomide as part of tandem transplantations was an independent favorable feature for PRS. Availability of thalidomide at protocol start (IFM9902, TT2, and TT3) was associated with a low HR (HR = 0.71, P < .001), whereas the availability of bortezomib in TT3 was associated with poor PRS (HR = 2.04, P < .001). Long preceding EFS retained its favorable impact on PRS.