Page 178 - 2019_07 resto del Mondo-web

T. Dittrich et al.
tion of the MAYO3b staging system compared to that of the other two models. However, AF retained its signifi- cant prognostic value in each condition.
All evaluated staging systems provide a high prognostic value
Using Kaplan-Meier estimates, we could illustrate the ability of each analyzed staging system to identify sub- groups of patients with distinct risks. There was virtually no overlap of the 95% confidence intervals of the different stages for each classification in the entire cohort (Figure 2A). With respect to the spread of the median overall sur- vival, it seems that the MAYO3b system is superior in detecting the patients with the best and worse prognoses in the entire cohort and in each subgroup.
In the next step, we evaluated the overlap and respec- tive prognosis of patients within each stage of the staging systems. The results of this analysis are shown in Table 3, in which proportions and median overall survival of patients are reported according to each stage of the MAYO2004/3b staging systems and subgrouped further by the MAYO2012 staging system. Both the MAYO3b and MAYO2012 staging systems were able to re-divide stages of the other system, identifying patients with better or worse outcomes. For example, 47% of the patients in MAYO2012 stage I were attributed to MAYO2004/3b stage II and 28% of the patients in MAYO2004/3b stage I were attributed to MAYO2012 stage II. In both cases, the higher stages were associated with an approximately 34 months shorter overall survival (Table 3).
To further evaluate and compare the prognostic value of the different staging systems, we computed time-depen- dent prediction errors, illustrating the overall model per- formance and time-dependent concordance indices, which mirror the systems’ discriminative accuracy (Figure 3). All staging systems showed a high performance value (Figure 3A). However, the MAYO3b and MAYO2012 sys- tems were superior to the MAYO2004 system with respect to discriminative accuracy in this analysis.
The overall performance of each staging system is almost entirely preserved in patients with impaired renal function
In the subgroup of patients with eGFR<50, all of the analyzed staging systems retained highly significant glob- al P values (Figure 2B). There was, however, only a poor discrimination of patients in stages I and II for all systems, as indicated by not significantly different median overall survivals (Figure 2B, Online Supplementary Table S5). This did not translate into a marked reduction of the predictive value according to prediction error curves, as well as con-
cordance indices, which was comparable to that of the entire cohort for each staging system (Figure 3B).
The performance of all staging systems is reduced
in patients with atrial arrhythmias, but best preserved in the MAYO3b system.
In the subgroup of patients with AF, the analyzed stag- ing systems retained statistically significant global P-val- ues, while there were overlaps between some stages (Figure 2C). Prediction error curves and concordance index plots did, however, confirm a reduced performance of all staging systems when compared to the performance in the entire cohort (Figure 3C). The MAYO2012 score appeared to be more affected than MAYO2004 and MAYO3b, especially with respect to discriminative accu- racy for prediction beyond 18 months. Thus, the MAYO3b system clearly outperforms the MAYO2012 system in the subgroup of patients with AF.
Discussion
This large retrospective study was aimed to: (i) evaluate the influence of two conditions with high prevalence among patients with AL amyloidosis patients, impaired renal function (eGFR<50 mL/min/1.73 m2, “eGFR<50”) as well as atrial arrhythmia or pacemaker rhythm (“AF”), on the applicability of the most widely used AL amyloidosis cardiac staging systems (MAYO2004, MAYO3b and MAYO2012); (ii) evaluate the additional prognostic value of eGFR<50 and AF in the context of these staging sys- tems; and (iii) compare the overall performance of these staging systems in the entire cohort and in the subgroups with eGFR<50 and AF.
(I) Patients with eGFR<50 or AF had a dismal prognosis; the risk of death was almost doubled by either comorbid- ity. The underlying reason seems to be that both condi- tions are surrogate parameters for heart involvement and accompanied by other unfavorable characteristics such as older age, lower Performance Status and, consequently, reduced eligibility for chemotherapy (Table 1). Interestingly, prerenal kidney injury due to heart involve- ment appears to be one of the most common reasons for reduced eGFR (Online Supplementary Results, Online Supplementary Figure S5B) and the survival of patients with decreased kidney function is mainly determined by heart involvement status and not negatively influenced by a higher degree of kidney organ involvement (Online Supplementary Results, Online Supplementary Figure S6A).
We verified in our cohort of AL patients that impaired renal function as well as AF are associated with increased
Table 3. Comparison of the MAYO2004/3b and MAYO2012 staging systems.
MAYO2012 stage I MAYO2012 stage II MAYO2012 stage III MAYO2012 stage IV
N. Median OS 95% CI N. Median OS 95% CI N. Median OS 95% CI N. Median OS 95% CI
MAYO2004/3bstageI 141 131.7 130-NR 54 97.8 62-NR - - - - - -
MAYO2004/3bstageII 125 97.5 70-NR 191 54.2 45-88 135 33 27-54 7 15.1 8-NR MAYO2004stageIII - - - 27 72.1 25-NR 183 17.1 14-24 360 5.9 5-8 MAYO3bstageIIIa - - - 27 72.1 25-NR 112 24.4 17-46 143 17.5 11-29 MAYO3b stage IIIb - - - - - - 71 7.7 6 - 18 217 3.4 3 - 5
N: number of patients; OS: overall survival; 95% CI: 95% confidence interval; NR: not reached.
1456
haematologica | 2019; 104(7)