RESULTS
Between 2009 and 2018, a total of 469 newly diagnosed oral cancer
patients receiving wide resection and neck dissection were recruited.
After propensity score matching, 400 oral cancer patients were enrolled
for our analysis. The mean age was 55.2+ 11.1 years and 92% were
male (Table 1). In period 1, patients treated by target group surgeons
were more likely to be older, had tongue cancer, and received adjuvant
radiotherapy. In period after 2015 Oct, patients treated by target group
surgeons were more likely to have tongue cancer.
The average LN yield per half-year by different departments was
illustrated in Figure 2. As shown in Table 2, compared with control
group in period 1, the average LN yield was lower in target group (22.8
vs 33.4, p < 0.001). The average LN yield was also
lower in target group between different neck dissection types (21.2 vs
29.3, p<0.001 in END group; 24.3 vs 38, p<0.001 in RND group). However, compared with control in period
2, the average LN yield was similar between the target group and control
group (37.4 vs 33.9, p =0.099); 34.2 vs 33.5 (p=0.7) in END group
and 41.5 vs 35 (p=0.11) in RND group. This finding means there was
significant increase of LN yield in target after 2015 Oct, but no
statistical difference of LN yield in control group. The
difference-in-difference analysis through linear model estimated the
impact of LNPA on LN yield. In Table 3, the interaction item, target
group\(\times\)period 2, with a coefficient of 13.24 (p<0.001)
represented the improvement of LN yield in target group after launching
LNPA since 2015 Oct.
We further analyzed the regional recurrence by diagnosis month. The
regional recurrence rate between target and control groups was presented
in Table 4 and Figure 3. During period before 2015 Oct, regional
recurrence was a little higher among patients treated in target group
(8.5% vs 5.9%, p=0.45). In period after 2015 Oct, regional recurrence
in target group was a little lower than control group (2.5% vs 3.7%,
p=0.65). Difference-in-differences analysis in Table 5 showed a
borderline protective effect of LNPA for cN0 patients with a reduced
regional recurrence rate of 9.4% (p =0.087).