Institutional paired match comparison
A total of 75 patients were available to assess from the institutional database. Of these, 12 patients underwent postoperative adjuvant RT at the Peking Union Medical College Hospital. These were matched one to one to patients who did not undergo adjuvant RT suing the propensity matching. Patients underwent resection between 2003 and 2018. All patients had localized or oligometastatic disease and underwent surgery with curative intent. Median follow-up times for the ART and control groups were 23 months (range: 10-76) and 37 months (14-92). respectively.
Table 1 summarizes the baseline characteristics of patients by treatment group. There was no significant difference between the groups in term of sex, age, stage, receipt of mitotane, tumor size, endocrine syndrome, or surgical margin status. The majority of patients treated with adjuvant RT were treated after 2010 (91.7%). There was an average of 51 days from the date of surgery to initiation of RT (range, 16 to 199 days). Only one patient had a more than 100 days interval between surgery and initiation of RT. All patients were treated with three-dimensional conformal RT.
The OS distributions for case subjects and control subjects were significantly different (log-rank P = 0.149) (Fig. 1A). A total of 9 (58.3%) patients [ 5 (41.7%) treated with adjuvant RT and 4 (33.3%) treated without adjuvant RT] are known to have died. OS estimates at 3 were 62.7% vs 71.0%, respectively, with an adjusted HR of 2.81 (95% CI, 0.65 to 12.09; P=0.165).
Local recurrence was diagnosed in 2 of the ART group, and in 4 of the control group (P=0.64). Locoregional recurrence-free survival (LRFS) was similar between the two groups (log-rank P = 0.879) (Fig. 1B). Locoregional RFS estimates at 3 years were 73.6% vs 56.6%, respectively, with an adjusted HR of 0.87 (95% CI, 0.16 to 4.91; P = 0.88).
Any recurrence, including distant failures or death, occurred in 15 (71.4%) patients [7 (70%) treated with RT and 8 (72.7%) treated with resection alone]. There was an insignificant difference in DFS between patients who received adjuvant RT and those who underwent resection only (log-rank P value 0.085) (Fig. 1C). DFS estimates at 3 years were 37.5% vs 60.3%, respectively, with an adjusted HR of 2.58 (95% CI, 0.846 to 7.84; P = 0.096).
Meta-analysis for oncological outcomes
A total of 191 studies were retrieved based on our searching strategy. After reading the abstract, 16 studies were related to our aims (Fig. 2), and 11 of them were subsequently excluded with reasons give. The remaining 5 studies and our data were selected for our analysis (Table 2) [9-13]. A total of 238 participants were selected for the meta-analysis, of which 111 and 127 patients underwent ART after surgical resection and only surgical resection, respectively. All the studies included were retrospective design and the majority of them were from Europe and America.
The Newcastle-Ottawa Scale (NOS) was used to examine the quality of all included studies. Most studies (5 of 7) were marked as 7 in NOS as they failed to report enough follow-up time. The 1 remaining study received full scores in NOS, indicating they were high quality original studies (Table 3).
Overall survival
Of the six studies included in our meta-analysis, five reported OS event numbers for both ART group and control group (Fig. 3). The total patient population was 95 in both groups. There was significant heterogeneity among the five retrospective studies (Chi2 = 7.81, P = 0.099, I2 = 48.8%), and the fixed effects model was applied. Meta-analysis significantly favored OS with ART, with an OR of 2.41 (95% CI of 1.33, 4.38; P = 0.004).
Local recurrence-free survival
All the six studies reported event numbers for LRFS for both ART and control groups (Fig. 4). There were 111 patients in the ART group and 127 in the control group. Meta-analysis revealed significant heterogeneity among the six retrospective studies (Chi2 = 15.83, P = 0.007, I2 = 68.4%), and the random effects model was applied. The pooled analysis significantly favored ART for LRFS of ACC patients, with an OR of 4.08 (95% CI of 1.29, 2.14; P = 0.016). After performing a sensitive analysis, and found the heterogeneity was introduced by the MDACC study [10]. The imbalanced sample size and different follow-up time caused the referral bias. After that, combining the results of the five cohorts (190 patients) a statistical significance in favor of ART group was observed with a OR = 6.25 (95% IC 3.24–12.05, P < 0.001) (Fig. 5). There was low heterogeneity among the five studies (Chi2 = 5.69, P = 0.224, I2 = 29.6%) indicating that the pooled analysis is valid.
Recurrence-free survival
Five of six studies reported event numbers for recurrence-free survival for both ART and control groups (Fig. 6). There were 92 patients in the ART group and 93 in the control group. There was low heterogeneity among the five retrospective studies (Chi2 = 3.79, P = 0.435, I2 = 0%), and the fixed effect model was applied. Meta-analysis significantly favored ART for disease-free survival in ACC, with an OR of 2.27 (95% CI of 1.23, 4.18; P = 0.009).