Proliferation to Apoptosis Tumor Cell Ratio as a Biomarker to Improve Clinical Management of Pre-Malignant and Symptomatic Plasma Cell Neoplasms

Proliferation and apoptosis of neoplastic cells are prognostic biomarkers in plasma cell neoplasms (PCNs). The prognostic capacity of proliferation to apoptosis ratio (Ratio-PA) in the era of immunomodulatory treatments is re-evaluated in 316 gammopathy of undetermined significance (MGUS), 57 smoldering multiple myeloma (SMM), and 266 multiple myeloma (MM) patients. Ratio-PA of 0.77 ± 0.12, 1.94 ± 0.52, and 11.2 ± 0.7 (p < 0.0001) were observed in MGUS, SMM, and MM patients. Ten-year overall survival (10y-OS) rates for patients with low/high Ratio-PA were 93.5%/77.3% p < 0.0001) for MGUS, 82.5%/64.7% (p < 0.05) for SMM, and 62.3%/47.0% (p < 0.05) for MM. For patients with low, intermediate, and high risk, 10y-OS for low/high Ratio-PA were 95.5%/72.9% (p < 0.0001), 74.2%/50.4% (p < 0.0001), and 35.3%/20.0% (p = 0.836), respectively. Ratio-PA was an independent prognostic factor for OS (HR = 2.119, p < 0.0001, Harrell-C-statistic = 0.7440 ± 0.0194) when co-analyzed with sex, age, and standard risk. In patients with Ratio-PAhigh, only first-line therapy with VRd/VTd, but not PAD/VCD, coupled with ASCT was associated with high 10y-OS (82.7%). Tumor cell Ratio-PA estimated at diagnosis offers a prognostic biomarker that complements standard risk stratification and helps to guide the clinical management of pre-malignant and symptomatic PCNs. Every effort should be made to provide first-line therapies including VTd or VRd associated with ASCT to patients with Ratio-PAhigh at higher risk of progression and death.

Cox regression analysis of total PCN showed that Ratio-PA was an independent prognostic factor for PFS (HR = 2.289, p < 0.0001, Harrell C-statistic = 0.7104 ± 0.0193) and OS (HR = 2.119, p < 0.0001, Harrell C-statistic = 0.7440 ± 0.0194) when analyzed together with sex (shorter PFS and OS for men than women), age (shorter PFS and OS for elders), and standard risk stratification (HR = 2.247, p < 0.0001 for PFS and HR = 3.280, p < 0.0001 for OS; Figure 3B). Comparable HR results were observed when MGUS, SMM, and MM stages were analyzed separately (see Figure 3C for details). for OS; Figure 3B). Comparable HR results were observed when MGUS, SMM, and MM stages were analyzed separately (see Figure 3C for details).

Ratio-PA Complements the Prognostic Capacity of Standard Risk Stratification in PCNs
The Ratio-PA showed prognostic capacity in pre-malignant and symptomatic PCNs not only by itself, but also complementing the standard risk stratification estimated in each type of PCN following current standard criteria, especially for low-(Risk-I) or intermediate-(Risk-II) risk patients. According to the low/high Ratio-PA, we observed 10y-PFS and 10y-OS rates of 85.4%/47.9% (p < 0.0001) and 95.5%/72.9% (p < 0.0001) for low-risk patients; 65.7%/35.4% (p < 0.0001) and 74.2%/50.4% (p < 0.0001) for intermediate-risk patients; and 23.5%/16.0% (p = 0.703) and 35.3%/20.0% (p = 0.836) for high-risk patients (Figure 4, left plots). It is noteworthy that 24.0% of standard low-risk patients who had Ratio-PA high showed PFS and OS curves close to those seen in intermediate-risk patients; and in the same way, 35.2% of intermediate-risk patients who had Ratio-PA high showed PFS and OS curves that were closed to those seen in high-risk patients.

Ratio-PA Complements the Prognostic Capacity of Standard Risk Stratification in PCNs
The Ratio-PA showed prognostic capacity in pre-malignant and symptomatic PCNs not only by itself, but also complementing the standard risk stratification estimated in each type of PCN following current standard criteria, especially for low-(Risk-I) or intermediate-(Risk-II) risk patients. According to the low/high Ratio-PA, we observed 10y-PFS and 10y-OS rates of 85.4%/47.9% (p < 0.0001) and 95.5%/72.9% (p < 0.0001) for low-risk patients; 65.7%/35.4% (p < 0.0001) and 74.2%/50.4% (p < 0.0001) for intermediate-risk patients; and 23.5%/16.0% (p = 0.703) and 35.3%/20.0% (p = 0.836) for high-risk patients ( Figure 4, left plots). It is noteworthy that 24.0% of standard low-risk patients who had Ratio-PA high showed PFS and OS curves close to those seen in intermediate-risk patients; and in the same way, 35

Delays in BM Sample Processing Reduces the Prognostic Capacity of the Ratio-PA
Given that both proliferation and apoptosis rates of BM-PC are functional biological properties, we evaluated the impact that delays in BM sample processing could have in the prognostic capacity of the Ratio-PA. Ratio-PA (3.7 ± 0.54, 2.45 ± 0.57, and 1.1 ± 0.23) and the proportion of patients with Ratio-PA high (39.7%, 31.3%, and 19.7%) decreased proportionally to the time elapsed from extraction (2 h, 4 h, and 24 h, respectively) ( Figure  5A). Differences in the distribution of patients by PCN stages or standard risk groups for samples processed at different times, which could have justified the differences observed in their Ratio-PA, were ruled out ( Figure 5B). Nonetheless, patients with Ratio-PA high showed lower PFS rates than patients with Ratio-PA low for samples processed in 2 h (81.3% vs. 39.5%, p < 0.0001), 4 h (68.3% vs. 42.8%, p < 0.05), or 24 h (70.9% vs. 48.3%, p < 0.01), although differences in the PFS and its statistic signification decreased in inverse proportion to the time elapsed until laboratory processing. Besides, Cox regression analysis confirmed that Ratio-PA was an independent prognostic factor in the PFS of PCN patients, but again with decreasing HR (3.189, 1.865, and 1.890), level of statistic signification (p < 0.0001, p < 0.01, and p < 0.05), and the Harrell C-statistic (0.7073 ± 0.0292, 0.6591 ± 0.0328, and 0.6728 ± 0.0402) as time until processing increased ( Figure 5C). Left, Kaplan-Meier, and Log-rank tests for progression-free survival (PFS) and overall-survival (OS) of total patients according to Ratio-PA (high or low) in the standard risk groups (Risk-I, -II, and -III).
Ten-year PFS and OS rates are indicated for each risk group in the plots. Tables on the right show same information for patients with monoclonal gammopathy of undetermined significance (MGUS), smoldering multiple myeloma (SMM), or multiple myeloma (MM). P estimated in the Log-rank test.

Delays in BM Sample Processing Reduces the Prognostic Capacity of the Ratio-PA
Given that both proliferation and apoptosis rates of BM-PC are functional biological properties, we evaluated the impact that delays in BM sample processing could have in the prognostic capacity of the Ratio-PA. Ratio-PA (3.7 ± 0.54, 2.45 ± 0.57, and 1.1 ± 0.23) and the proportion of patients with Ratio-PA high (39.7%, 31.3%, and 19.7%) decreased proportionally to the time elapsed from extraction (2 h, 4 h, and 24 h, respectively) ( Figure 5A). Differences in the distribution of patients by PCN stages or standard risk groups for samples processed at different times, which could have justified the differences observed in their Ratio-PA, were ruled out ( Figure 5B). Nonetheless, patients with Ratio-PA high showed lower PFS rates than patients with Ratio-PA low for samples processed in 2 h (81.3% vs. 39.5%, p < 0.0001), 4 h (68.3% vs. 42.8%, p < 0.05), or 24 h (70.9% vs. 48.3%, p < 0.01), although differences in the PFS and its statistic signification decreased in inverse proportion to the time elapsed until laboratory processing. Besides, Cox regression analysis confirmed that Ratio-PA was an independent prognostic factor in the PFS of PCN patients, but again with decreasing HR (3.189, 1.865, and 1.890), level of statistic signification (p < 0.0001, p < 0.01, and p < 0.05), and the Harrell C-statistic (0.7073 ± 0.0292, 0.6591 ± 0.0328, and 0.6728 ± 0.0402) as time until processing increased ( Figure 5C).

Tandem ASCT with VTd or VRd Are the Most Effective First-Line Treatments for Patients with Ratio-PA high
Finally, to explore which therapies of those used in current clinical practice were the most effective in patients with Ratio-PA high , we evaluated the 10y-OS rate and the relapse rate after sCR in symptomatic MM (n = 254), as well as in MGUS (n = 19) or SMM (n = 24) patients that progressed and required treatment (n = 297 total). Patients were classified according to the first-line treatment used ( Figure 6A). In patients eligible for ASCT, Ratio-PA low was associated with good 10y-OS rates (77.4% vs. 87.5%) for both PAD/VCD and VRd/VTd treatments, whereas in patients with Ratio-PA high , only VRd/VTd, but not PAD/VCD, were associated with high 10y-OS rates (82.7% vs. 55.2%). Patients not eligible for ASCT showed much lower 10y-OS rates, while non-significant differences were observed according to the Ratio-PA low/high (46.1% vs. 35.6%).
Higher 10y-OS rates observed in patients with Ratio-PA high treated with ASCT and VRd/VTd were related to a lower relapse rate than the one for patients treated with ASCT and PAD/VCD or No-ASCT (27.8% vs. 73.7% and 78.3%, p < 0.01). In the same line, high 10y-OS rates observed in patients with Ratio-PA low treated with ASCT, either with PAD/VCD or VRd/VTd, were associated with low relapse rates, 37.5% and 44.4% respectively, compared to the relapse rate of patients not eligible for ASCT (63.6%) ( Figure 6B).
Patients who did not relapse in the first six months after sCR showed higher 10y-OS rates compared to those who relapsed (86.0% vs. 51.6%, p < 0.0001). As expected, the Ratio-PA (low/high) did not have a significant impact on the 10y-OS rates of patients who did not relapse (87.0% vs. 84.6%); however, Ratio-PA low identified patients who, after relapse, showed higher 10y-OS rates than those with Ratio-PA high (72.7% vs. 41.0%, p < 0.05) ( Figure  6C). Finally, to explore which therapies of those used in current clinical practice were the most effective in patients with Ratio-PA high , we evaluated the 10y-OS rate and the relapse rate after sCR in symptomatic MM (n = 254), as well as in MGUS (n = 19) or SMM (n = 24) patients that progressed and required treatment (n = 297 total). Patients were classified according to the first-line treatment used ( Figure 6A). In patients eligible for ASCT, Ratio-PA low was associated with good 10y-OS rates (77.4% vs. 87.5%) for both PAD/VCD and VRd/VTd treatments, whereas in patients with Ratio-PA high , only VRd/VTd, but not PAD/VCD, were associated with high 10y-OS rates (82.7% vs. 55.2%). Patients not eligible for ASCT showed much lower 10y-OS rates, while non-significant differences were observed according to the Ratio-PA low/high (46.1% vs. 35.6%).
Higher 10y-OS rates observed in patients with Ratio-PA high treated with ASCT and VRd/VTd were related to a lower relapse rate than the one for patients treated with ASCT and PAD/VCD or No-ASCT (27.8% vs. 73.7% and 78.3%, p < 0.01). In the same line, high 10y-OS rates observed in patients with Ratio-PA low treated with ASCT, either with PAD/VCD or VRd/VTd, were associated with low relapse rates, 37.5% and 44.4% respectively, compared to the relapse rate of patients not eligible for ASCT (63.6%) ( Figure 6B).
Patients who did not relapse in the first six months after sCR showed higher 10y-OS rates compared to those who relapsed (86.0% vs. 51.6%, p < 0.0001). As expected, the Ratio-PA (low/high) did not have a significant impact on the 10y-OS rates of patients who did not relapse (87.0% vs. 84.6%); however, Ratio-PA low identified patients who, after relapse, showed higher 10y-OS rates than those with Ratio-PA high (72.7% vs. 41.0%, p < 0.05) ( Figure 6C). To further explore the impact of Ratio-PA in the outcome of patients requiring antimyeloma treatment, Supplementary Figure S1 shows the impact of Ratio-PA according to their eligibility for ASCT or their relapse status in the RISS stages. In ASCT-ineligible patients, only those with RISS-I and Ratio-PA high showed lower 10y-OS. In ASCT-eligible patients, RISS-I, RISS-II, and RISS-III stages with Ratio-PA high showed lower 10y-OS. In patients without relapse, Ratio-PA high did not lead to differences in the 10y-OS in patients with RISS-I, but the survival rate was lower in those with RISS-II and RISS-III. In relapsed patients with RISS-I and RISS-II, the Ratio-PA high led to lower 10y-OS. Nonetheless, these results should be confirmed in larger series due to the reduced number of patients in some subgroups.

Discussion
Plasma cell neoplasms with a prevalence >3% on population aged ≥ 50 (>5% on aged > 70) for MGUS [9,10] and close to 1% for MM are one of the most common pathologies in hematology clinics worldwide. Although most of them are asymptomatic, 1% yearly evolves to a symptomatic PCN that requires treatment, as do patients diagnosed of MM. Fortunately, effective risk stratification methods revised in recent years [7,8,30] have significantly contributed to rationalizing clinical follow-up and therapeutic management of these patients. However, substantial clinical heterogeneity in presentation and course is observed among patients within the same risk status or under comparable therapy regimens, underlining critical differences in neoplastic PC biology and its resistance to current anti-myeloma drugs. This study, performed in a large series of patients under current standard treatments and followed-up for a period of 10 years (mean ± SD, 55.5 ± 39 months), shows that proliferation and apoptosis rates of the neoplastic cells significantly determined the progression of the disease and the survival of patients with asymptomatic PCNs and also with symptomatic MM that required treatment. As suggested by our data, the net growth of the tumor estimated as a Ratio-PA of the BM-PCs could determine the To further explore the impact of Ratio-PA in the outcome of patients requiring antimyeloma treatment, Supplementary Figure S1 shows the impact of Ratio-PA according to their eligibility for ASCT or their relapse status in the RISS stages. In ASCT-ineligible patients, only those with RISS-I and Ratio-PA high showed lower 10y-OS. In ASCT-eligible patients, RISS-I, RISS-II, and RISS-III stages with Ratio-PA high showed lower 10y-OS. In patients without relapse, Ratio-PA high did not lead to differences in the 10y-OS in patients with RISS-I, but the survival rate was lower in those with RISS-II and RISS-III. In relapsed patients with RISS-I and RISS-II, the Ratio-PA high led to lower 10y-OS. Nonetheless, these results should be confirmed in larger series due to the reduced number of patients in some subgroups.

Discussion
Plasma cell neoplasms with a prevalence >3% on population aged ≥50 (>5% on aged >70) for MGUS [9,10] and close to 1% for MM are one of the most common pathologies in hematology clinics worldwide. Although most of them are asymptomatic, 1% yearly evolves to a symptomatic PCN that requires treatment, as do patients diagnosed of MM. Fortunately, effective risk stratification methods revised in recent years [7,8,30] have significantly contributed to rationalizing clinical follow-up and therapeutic management of these patients. However, substantial clinical heterogeneity in presentation and course is observed among patients within the same risk status or under comparable therapy regimens, underlining critical differences in neoplastic PC biology and its resistance to current antimyeloma drugs. This study, performed in a large series of patients under current standard treatments and followed-up for a period of 10 years (mean ± SD, 55.5 ± 39 months), shows that proliferation and apoptosis rates of the neoplastic cells significantly determined the progression of the disease and the survival of patients with asymptomatic PCNs and also with symptomatic MM that required treatment. As suggested by our data, the net growth of the tumor estimated as a Ratio-PA of the BM-PCs could determine the aggressiveness of the disease and its resistance to current treatments. Our results also reveal two issues of crucial relevance for the clinical management of these patients: firstly, patients with high net tumor growth (Ratio-PA high ) will respond mostly to first-line therapies including VTd or VRd associated with ASCT, while PCNs with low net tumor growth will show comparable good results with first-line therapies including PAD/VCD or VTd/VRd associated with ASCT; and secondly, proliferation and apoptosis of neoplastic PC should be analyzed in the shortest possible time from BM extraction (preferably <2 h), since time and/or transport alter these functional characteristics of the tumor cell, reducing the prognostic capacity of the Ratio-PA and consequently worsening the correct decision-making in both premalignant and symptomatic PCNs.
Indeed, delays in BM sample processing seemed to affect the prognostic capacity of Ratio-PA. Nonetheless, its independent predictive value remained valid, although it decreased, even in samples that were analyzed 24 h after extraction. This is in agreement with numerous studies in PCNs demonstrating a prognostic capacity of either the proliferation rate, [14][15][16][17][18][19] the apoptosis rate [20], or Ratio-PA [28,29]. Although, in light of our results, these studies may have underestimated the prognostic capacity of these parameters, it has been well-established that the PC proliferation rate is one of the most important independent prognostic factors, a specific marker of MM aggressiveness [31][32][33] and an early marker for disease progression and death [31,34]. However, our data suggest that the inverse relationship between BM-PC proliferation and apoptosis and particularly the balance established between these two parameters is what would ultimately determine the expandability and aggressiveness of the neoplastic PC, and would explain why Ratio-PA was the parameter with the highest predictive value in our PCN series. In fact, immunomodulatory therapies base their anti-myeloma action on the induction of PC apoptosis, which could help to explain why ASCT eligible patients treated with IMiDs showed the best outcome in our series [22,23]. It has been reported that the neoplastic population in MM consists of three clonally-related compartments [35]: (1) self-maintaining stem-cells, (2) growth fraction, and (3) differentiated non-proliferative PCs that will die by apoptosis and necrosis. High Ratio-PA would be indicative of an imbalance between these compartments, in favor of the proliferative one, which would justify a greater aggressiveness and a higher progression rate in these patients. Improvement of clinical outcomes in MM with the introduction of immunomodulatory drugs has caused some biomarkers to lose their prognostic value, and has made it necessary to redefine the stratification criteria to adjust their predictive capacity [7,8,30]. However, the clinical variability observed among patients with the same risk makes it necessary to complement the standard risk stratification with biologically meaningful biomarkers also reassessed for current therapies. In this study, the prognostic capacity of the Ratio-PA was re-evaluated for current treatments in MM, and it was found that not only does it continue to maintain its independent prognostic capacity, but it is also capable of complementing the RISS. Ratio-PA high identified 24.0% of low standard risk patients who showed rates of PFS and OS similar to patients with intermediate risk. Furthermore, Ratio-PA high identified 35.2% of intermediate-risk patients whose survival curves were closed to those at high standard risk. These patients could have benefited from first-line treatments reserved for high-risk patients. Moreover, if eligible for ASCT, these patients should receive regimens based on VTd or VRd associated to ASCT, the only ones that in our series seem to be effective against PC neoplasms with a high net tumor growth.
Unfortunately, Ratio-PA was unable to provide additional information on the prognosis of patients not eligible for ASCT, since all of them presented an unfavorable clinical course irrespective of the proliferation or apoptosis rates of their neoplastic cells. An exception was ASCT non-eligible patients with RISS-I, where Ratio-PA high identified a subset of patients at higher risk. However, a Ratio-PA low at diagnosis would point at 36% of cases that relapsed after sCR and then had prolonged survival. In contrast, Ratio-PA high predicted a cohort of patients with much poorer OS who could benefit from consolidation therapy, tandem ASCT, or intensive maintenance after relapse [36]. Recently, it has been described that evaluating PC proliferative index post-transplant is a powerful predictor of prognosis in myeloma patients failing to achieve a complete response [37]. Nonetheless, our data suggest that this is a characteristic that could be fixed from the diagnosis. Finally, Ratio-PA high did not impact the good clinical course of patients with RISS-I who maintained sCR beyond 6 months post-ASCT, although it could be pointing at RISS-II and RISS-III patients with worse clinical course, even though they maintained early sCR post-ASCT.
Although it should be confirmed in larger series, our results suggest that Ratio-PA could be helpful in the clinical management of SMM patients. Excluding ultra-highrisk SMM patients (risk of progression >80% in the first two years) who should be diagnosed as symptomatic MM [38,39], the optimal time to treat SMM patients remains controversial [40][41][42]. In our series, Ratio-PA high identified 29.8% of SMM patients with 35.3% 10y-PFS who could have benefited from an early treatment, [43] since safer and more effective therapies are now available. Watchful waiting was appropriate in an era of only alkylating agents and corticosteroids, but is no longer justifiable for these high-risk patients [39].
Ratio-PA analysis could also facilitate risk-adapted follow-up and clinical management in MGUS. MGUS progression to malignant PCNs occurs at a rate of 1% per year [11]. However, in our series, Ratio-PA high pointed out patients with a risk of progression and death 3.2 and 3.4 times higher, respectively. Shorter OS of MGUS patients with Ratio-PA high supports the possibility of a direct negative effect of the monotypic clone inducing severe organ damage, such as by favoring thrombosis, infection, or osteoporosis, amongst others; [44] therefore, it would be reasonable to assess the possibility of early treatment in these high-risk MGUS cases.
Altogether, these results show that the proliferation-to-apoptosis ratio (Ratio-PA) of PC, estimated at diagnosis by flow cytometry in the shortest possible time from BM extraction (preferably <2 h), offers a prognostic biomarker that complements standard risk stratification and helps to guide the clinical management of pre-malignant and symptomatic PCNs. Our results also show that every effort should be made to provide first-line therapies including VTd or VRd associated with ASCT to patients with Ratio-PA high , at higher risk of progression and death. Finally, Ratio-PA high predicted a cohort of patients who could benefit from consolidation therapy, tandem ASCT, or intensive maintenance when they relapse after stringent complete response.

Patients and Samples
EDTA anti-coagulated BM samples were obtained at diagnosis from 639 consecutive patients with PCNs in regular clinical practice from seven hospitals in the Region of Murcia, Spain. BM samples were also obtained for MRD assessment at three and six months after ASCT, and under the suspicion of loss of sCR. Patients were enrolled between 2010 and 2017, and followed-up until February 2021. This study was approved by the Research Ethics Committee Institutional Review Board (IRB-00005712). Written informed consent was obtained from all patients in accordance with the Declaration of Helsinki.
Following the IMWG criteria [8], patients were classified in 316 MGUS, 57 SSM, and 266 MM. Standard risk stratification (sRisk) in MGUS [30], SMM [7], and MM [8] was done following updated criteria. Patients were grouped as low (Risk-I), intermediate (Risk-II), and high (Risk-III) risk. In MGUS, disease progression was computed when it progressed to SMM or MM. In SMM and MM progression, complete response (CR) and relapse were estimated following the IMWG Uniform Response Criteria for Multiple Myeloma [39,45]. Treatments and management were at the discretion of the hematologists based on patient condition and tumor risk. Briefly, conventional first-line therapy for patients not eligible for ASCT included bortezomib, melphalan, and prednisone (VMP), bortezomib and dexamethasone (Vd), or more recently, lenalidomide and dexamethasone (Rd). In ASCT-eligible patients, first-line therapy included bortezomib, cyclophosphamide, dexamethasone (VCD) or bortezomib, doxorubicin, and dexamethasone (PAD) or more recently bortezomib, thalidomide, and dexamethasone (VTd) or bortezomib, lenalidomide, and dexamethasone (VRd), and ASCT conditioning with melphalan 200 mg/m 2 (dose ranging from 200 to 100-140 mg/m 2 in the case of renal impairment).
The proliferation rate of aberrant PC was estimated as the percentage of CD38/CD138 + cells in the Synthesis + G2/M phases of the cell cycle by using Cycloscope-MM (Cytognos, Salamanca, Spain). Apoptosis rate of aberrant PC was estimated as the percentage of Anexin-V + PCs minus the percentage of Anexin-V + mature B lymphocytes by using anti-Anexin-V V450 (BD).
Given that both proliferation and apoptosis rates of BM-PC are functional biological properties that can be compromised by the time elapsed from extraction to laboratory processing, the transport, and/or storage conditions, we compared the Ratio-PA values and its prognostic capacity in in-house samples (analyzed in less than 2 h, n = 293), and in external samples which were processed in less than 4 h (n = 173) or 24 h (n = 142), depending on whether samples were shipped on the same day or the day after the BM extraction. BM samples were kept at room temperature and protected from extreme temperatures during transport. As described above, the apoptosis rate of mature B lymphocytes was extracted from the PC apoptosis rate as an internal control so as to minimize the impact of process times.

Statistical Analysis
Statistical analyses were performed using SPSS version 15.0 (SPSS Inc, Chicago, IL, USA). Analysis of variance (ANOVA) and least significant difference (LSD) post hoc tests were used to analyze continuous variables. The receiver operating characteristic (ROC) was used to explore patient PFS and to determine the optimal cutoff values for proliferation and apoptosis rates, as well as for the proliferation-to-apoptosis ratio (Ratio-PA). PFS was estimated as months from the diagnosis date to disease progression or death. Survival curves were plotted according to the Kaplan-Meier method. The log-rank test was used to estimate significant differences. Multivariate analyses of prognostic factors for PFS and OS were performed using the Cox proportional hazards model (stepwise regression). The hazard ratio (HR) and 95% confidence interval were estimated. The Harrell C-statistic was obtained using STATA-14 (Somersd package, College Station, TX, USA). p < 0.05 was considered statistically significant.