DOI: https://doi.org/10.21203/rs.3.rs-2527472/v1
Radical prostatectomy (RP) is one of the most commonly used treatments for patients with prostate cancer (PCa)1. According to European guidelines for PCa, extended pelvic lymph node dissection (PLND) is recommended in patients with a greater than 5% nomogram derived lymph node (LN) invasion probability2. The diagnostic value of PLND for PCa staging is acknowledged, which can help guide the adjuvant treatment. However, the therapeutic value of extended PLND remains controversial3. Furthermore, it is also associated with higher risk of perioperative complications such as increased blood loss, lymphoceles and thromboembolic events4,5. Several previous studies have shown that more extensive PLND is associated with a better cancer-specific survival (CSS) in PCa patients6–10. While other studies did not support this conclusion11–15. Among these studies, some compared outcome in the only patients with pathological positive or negative LN. However, pathological LN status is unknown when the decision is made to perform PLND or extended PLND.
A recent systematic review has concluded that PLND and its extent are associated with worse intraoperative and perioperative outcomes, whereas a direct therapeutic effect is still not evident from the current literature16. And robust and powered clinical trials are needed.
Based on these advances, it is clinically significant to investigate the prognostic value of extent of PLND. Based on SEER database, the accurate extent of PLND was unavailable, therefore we used the removed lymph node count (RLNC) as an indicator for the extent of PLND. We hypothesized that in higher risk of lymph node metastasis patients, more RLNC would contribute to better survival, so we include only patients with D’Amico high risk PCa. To test the hypothesis, we establish the prediction model in patients with RLNC over 6 to predict lymph node metastasis risk (LNMR) for every patient, and then explore the prognostic value of RLNC in patients with different LNMR.
In 96875 high risk PCa patients treated with RP and PLND, a total of 5261 patients (5.43%) had LN metastasis. Interquartile range (IQR) for age was 57–67 years, with a median of 62 years. Of all, 57.4% harbored cT2c stage, and 35.9% harbored T3 stage. Median PSA value was 6.4 (IQR 4.8–9.9) ng/mL. The median RLNC was 6, and the 75th percentile of RLNC was 11 (Table 1). Stratification according to RLNC ≥ 11 (25.8%) versus RLNC<11 (74.2%) revealed that patients who underwent PLND with more RLNC harbored less favorable tumor characteristics(Table 2). Specifically, they had higher rate of cT3/4 stage (47.1 vs 35.1%), GS ≥ 8 (23.4 vs 16.7%) and higher level of PSA value (mean: 10.5 vs 9.2 ng/mL). The mean predicted LNMR (0.06 ± 0.09) in RLNC<6 group was significantly higher than the proportion of metastatic LN (2.26%), because LNMR were predicted using the model established from the data in RLNC ≥ 6 patients.
Characteristics |
|
---|---|
Age |
62 (57–67) |
PSA value |
6.40 (4.8–9.9) |
Year of diagnosis |
|
2004 |
4844 (5.00%) |
2005 |
4619 (4.77%) |
2006 |
5270 (5.44%) |
2007 |
6101 (6.30%) |
2008 |
6091 (6.29%) |
2009 |
6326 (6.53%) |
2010 |
6849 (7.07%) |
2011 |
7035 (7.26%) |
2012 |
6154 (6.35%) |
2013 |
5947 (6.14%) |
2014 |
6032 (6.23%) |
2015 |
6815 (7.03%) |
2016 |
7812 (8.06%) |
2017 |
7916 (8.17%) |
2018 |
4391 (4.53%) |
2019 |
4673 (4.82%) |
Race |
|
White |
78881 (81.43%) |
Black |
12273 (12.67%) |
Asian |
5081 (5.24%) |
Unknown |
640 (0.66%) |
Marital status |
|
Married |
73324 (75.69%) |
Single |
18746 (19.35%) |
Unknown |
4805 (4.96%) |
T stage |
|
T1 |
66 (0.07%) |
T2A |
1023 (1.06%) |
T2B |
438 (0.45%) |
T2C |
55622 (57.42%) |
T2NOS |
2805 (2.90%) |
T3 |
34768 (35.89%) |
T4 |
2153 (2.22%) |
Gleason score |
|
6 |
16852 (17.40%) |
7 |
62227 (64.23%) |
8 |
8347 (8.62%) |
9 |
9449 (9.75%) |
LN examined |
6 (3–11) |
LN metastasis |
|
0 |
91614 (94.57%) |
1 |
5261 (5.43%) |
PCa = Prostate cancer; PLND = Pelvic lymph node dissection; PSA = Prostate specific antigen; LN = Lymph node |
Characteristics |
RLNC < 6 |
RLNC > = 6 |
RLNC < 11 |
RLNC > = 11 |
---|---|---|---|---|
Number |
47096 |
49779 |
72383 |
24492 |
Age |
61.5 ± 6.8 |
61.8 ± 6.7 |
61.6 ± 6.7 |
61.9 ± 6.7 |
PSA value |
9.0 ± 10.0 |
10.1 ± 11.4 |
9.2 ± 10.3 |
10.5 ± 11.9 |
RLNC |
3(2–4) |
10(7–15) |
4(2–7) |
16 (13–21) |
Predicted LNMR |
0.06 ± 0.09 |
0.08 ± 0.11 |
0.07 ± 0.09 |
0.09 ± 0.11 |
Race |
||||
White |
37853 (80.37%) |
41028 (82.42%) |
58632 (81.00%) |
20249 (82.68%) |
Black |
6484 (13.77%) |
5789 (11.63%) |
9528 (13.16%) |
2745 (11.21%) |
Asian |
2488 (5.28%) |
2593 (5.21%) |
3771 (5.21%) |
1310 (5.35%) |
Unknown |
271 (0.58%) |
369 (0.74%) |
452 (0.62%) |
188 (0.77%) |
Marital status |
||||
Married |
35977 (76.39%) |
37347 (75.03%) |
54994 (75.98%) |
18330 (74.84%) |
Single |
8852 (18.80%) |
9894 (19.88%) |
13843 (19.12%) |
4903 (20.02%) |
Unknown |
2267 (4.81%) |
2538 (5.10%) |
3546 (4.90%) |
1259 (5.14%) |
T stage |
||||
T1 |
31 (0.07%) |
35 (0.07%) |
47 (0.06%) |
19 (0.08%) |
T2A |
511 (1.09%) |
512 (1.03%) |
793 (1.10%) |
230 (0.94%) |
T2B |
213 (0.45%) |
225 (0.45%) |
316 (0.44%) |
122 (0.50%) |
T2C |
29557 (62.76%) |
26065 (52.36%) |
43851 (60.58%) |
11771 (48.06%) |
T2NOS |
1253 (2.66%) |
1552 (3.12%) |
1982 (2.74%) |
823 (3.36%) |
T3 |
14650 (31.11%) |
20118 (40.41%) |
23973 (33.12%) |
10795 (44.08%) |
T4 |
881 (1.87%) |
1272 (2.56%) |
1421 (1.96%) |
732 (2.99%) |
Gleason score |
||||
6 |
9974 (21.18%) |
6878 (13.82%) |
13995 (19.33%) |
2857 (11.67%) |
7 |
29668 (62.99%) |
32559 (65.41%) |
46318 (63.99%) |
15909 (64.96%) |
8 |
3800 (8.07%) |
4547 (9.13%) |
5968 (8.25%) |
2379 (9.71%) |
9 |
3654 (7.76%) |
5795 (11.64%) |
6102 (8.43%) |
3347 (13.67%) |
LN positive |
||||
0 |
46031 (97.74%) |
45583 (91.57%) |
70037 (96.76%) |
21577 (88.10%) |
1 |
1065 (2.26%) |
4196 (8.43%) |
2346 (3.24%) |
2915 (11.90%) |
CSS endpoint |
||||
Alive |
45962 (97.59%) |
48599 (97.63%) |
70615 (97.56%) |
23946 (97.77%) |
Cancer specific death |
1134 (2.41%) |
1180 (2.37%) |
1768 (2.44%) |
546 (2.23%) |
OS endpoint |
||||
Alive |
42154 (89.51%) |
45498 (91.40%) |
65061 (89.88%) |
22591 (92.24%) |
Dead |
4942 (10.49%) |
4281 (8.60%) |
7322 (10.12%) |
1901 (7.76%) |
10-year CSS |
||||
Alive |
46184 (98.06%) |
48809 (98.05%) |
70954 (98.03%) |
24039 (98.15%) |
Dead |
912 (1.94%) |
970 (1.95%) |
1429 (1.97%) |
453 (1.85%) |
PCa = Prostate cancer; RLNC = Removed lymph node count; PSA = Prostate specific antigen; LNMR = Lymph node metastasis risk; LN = Lymph node; CSS = Cancer specific survival; OS = Overall survival |
To stratifying patients with different risk of LN metastasis, a multivariable logistic regression model was established to predict LN metastasis. The training sample included the data from patients with RLNC ≥ 6. Age of diagnosis, T stage, baseline PSA, and Gleason score were included as independent variable. The area under receiver operating characteristic curve for training sample is 0.829 (Fig. 2). The LNMR was calculated for every patient according to the prediction model.
Stratification according to the LNMR subgroups revealed that 75.7% (73359), 14.4% (13912), 5.5% (5334), and 4.4% (4270) of all patients were with LNMR < 0.1, 0.1 ≤ LNMR < 0.2, 0.2 ≤ LNMR < 0.3 and LNMR ≥ 0.3. And the metastatic LN count were 1475 (2.01%), 1401 (10.07%), 1026 (19.24%) and 1359 (31.83%) (Table 3).
LNMR |
< 0.1 |
>=0.1, < 0.2 |
>=0.2, < 0.3 |
>=0.3 |
P-value |
---|---|---|---|---|---|
N |
73359 |
13912 |
5334 |
4270 |
|
Age |
61.41 ± 6.74 |
62.23 ± 6.61 |
62.92 ± 6.54 |
62.09 ± 6.89 |
< 0.001 |
PSA value |
7.20 ± 6.85 |
13.45 ± 10.92 |
17.42 ± 17.47 |
27.21 ± 22.80 |
< 0.001 |
RLNC |
5 (3–10) |
7 (3–12) |
7 (4–13) |
8 (4–15) |
< 0.001 |
Race |
< 0.001 |
||||
White |
60177 (82.03%) |
11058 (79.49%) |
4292 (80.46%) |
3354 (78.55%) |
|
Black |
9134 (12.45%) |
1865 (13.41%) |
691 (12.95%) |
583 (13.65%) |
|
Asian |
3550 (4.84%) |
899 (6.46%) |
329 (6.17%) |
303 (7.10%) |
|
Unknown |
498 (0.68%) |
90 (0.65%) |
22 (0.41%) |
30 (0.70%) |
|
Marital status |
< 0.001 |
||||
Married |
56324 (76.78%) |
10125 (72.78%) |
3876 (72.67%) |
2999 (70.23%) |
|
Single |
13398 (18.26%) |
3076 (22.11%) |
1209 (22.67%) |
1063 (24.89%) |
|
Unknown |
3637 (4.96%) |
711 (5.11%) |
249 (4.67%) |
208 (4.87%) |
|
T stage |
< 0.001 |
||||
T1 |
20 (0.03%) |
40 (0.29%) |
2 (0.04%) |
4 (0.09%) |
|
T2A |
450 (0.61%) |
501 (3.60%) |
60 (1.12%) |
12 (0.28%) |
|
T2B |
137 (0.19%) |
248 (1.78%) |
39 (0.73%) |
14 (0.33%) |
|
T2C |
55305 (75.39%) |
290 (2.08%) |
27 (0.51%) |
0 (0.00%) |
|
T2NOS |
838 (1.14%) |
1607 (11.55%) |
279 (5.23%) |
81 (1.90%) |
|
T3 |
15419 (21.02%) |
10663 (76.65%) |
4733 (88.73%) |
3953 (92.58%) |
|
T4 |
1190 (1.62%) |
563 (4.05%) |
194 (3.64%) |
206 (4.82%) |
|
Gleason score |
< 0.001 |
||||
6 |
16852 (22.97%) |
0 (0.00%) |
0 (0.00%) |
0 (0.00%) |
|
7 |
51463 (70.15%) |
9300 (66.85%) |
1324 (24.82%) |
140 (3.28%) |
|
8 |
3584 (4.89%) |
3012 (21.65%) |
1138 (21.33%) |
613 (14.36%) |
|
9 |
1460 (1.99%) |
1600 (11.50%) |
2872 (53.84%) |
3517 (82.37%) |
|
LN metastasis |
< 0.001 |
||||
0 |
71884 (97.99%) |
12511 (89.93%) |
4308 (80.76%) |
2911 (68.17%) |
|
1 |
1475 (2.01%) |
1401 (10.07%) |
1026 (19.24%) |
1359 (31.83%) |
|
CSS |
< 0.001 |
||||
Alive |
72420 (98.72%) |
13397 (96.30%) |
4913 (92.11%) |
3831 (89.72%) |
|
Cancer-specific death |
939 (1.28%) |
515 (3.70%) |
421 (7.89%) |
439 (10.28%) |
|
OS |
< 0.001 |
||||
Alive |
67224 (91.64%) |
12385 (89.02%) |
4545 (85.21%) |
3498 (81.92%) |
|
Dead |
6135 (8.36%) |
1527 (10.98%) |
789 (14.79%) |
772 (18.08%) |
|
10 year-CSS |
< 0.001 |
||||
Alive |
72655 (99.04%) |
13486 (96.94%) |
4975 (93.27%) |
3877 (90.80%) |
|
Dead |
704 (0.96%) |
426 (3.06%) |
359 (6.73%) |
393 (9.20%) |
|
PCa = Prostate cancer; LNMR = Lymph node metastasis risk; PSA = Prostate specific antigen; RLNC = Removed lymph node count; LN = Lymph node; CSS = Cancer specific survival; OS = Overall survival |
A smooth curve was plotted by LNMR and 10 year-CSM stratified by RLNC ≥ 11 and RLNC<11 or RLNC ≥ 6 and RLNC<6(Supplemental Fig. 1). Based on the smooth curve, as with the LNMR increasing, more favorable CSS occurred in patients with higher RLNC, especially in the group of RLNC ≥ 11 versus RLNC<11. When the 10-year CSM rates were plotted against RLNC according to different LNMR subgroups, the negative correlation between RLNC and CSM was more significant in LNMR ≥ 0.3 subgroup (Fig. 3).
Next, we analyzed the prognostic significance of RLNC on CSS and OS in the four different LNMR subgroups (Table 4). Multivariate COX regression analyses showed that in subgroup of patients with LNMR ≥ 0.3, more RLNC showed significant survival benefit with regard to CSS and OS when the cutoff value of RLNC is 11 (CSS: HR: 0.78 (0.63, 0.96) p = 0.0213; OS: HR: 0.85 (0.72, 0.99) p = 0.0407). When the cutoff value of RLNC was 6, there were no prognostic significance for OS and CSS in each subgroup. When the RLNC was continuous coded, there were prognostic value for CSS (HR: 0.9887 (0.9761, 1.0014) p = 0.0808) and OS (HR: 0.9891 (0.9796, 0.9987) p = 0.0256), although for CSS the result didn’t reach statistical significance. The interaction analysis between RLNC and LNMR (LNMR ≥ 0.3 and < 0.3) were performed and there was only significant interaction between cutoff value of 11 RLNC and LNMR (LNMR ≥ 0.3 and < 0.3) for CSS (interaction p = 0.0429).
LNMR |
RLNC6 |
RLNC 11 |
RLNC continuously coded |
---|---|---|---|
CSS |
0.94 (0.86, 1.04) 0.2365 |
1.03 (0.95, 1.12) 0.4637 |
0.9972 (0.9914, 1.0031) 0.3515 |
< 0.1 |
1.09 (0.96, 1.24) 0.1692 |
1.01 (0.86, 1.19) 0.8817 |
1.0020 (0.9921, 1.0119) 0.6997 |
>=0.1, < 0.2 |
0.97 (0.82, 1.16) 0.7486 |
0.91 (0.74, 1.11) 0.3403 |
0.9970 (0.9851, 1.0090) 0.6202 |
>=0.2, < 0.3 |
0.99 (0.82, 1.20) 0.9353 |
1.05 (0.85, 1.29) 0.6780 |
0.9961 (0.9833, 1.0092) 0.5602 |
>=0.3 |
0.99 (0.82, 1.19) 0.8938 |
0.78 (0.63, 0.96) 0.0213 |
0.9887 (0.9761, 1.0014) 0.0808 |
Interaction p value with LNMR |
0.6183 |
0.0429 |
0.1545 |
OS |
0.97 (0.93, 1.01) 0.1630 |
0.94 (0.90, 0.99) 0.0265 |
0.998 (0.995, 1.002) 0.3346 |
< 0.1 |
0.98 (0.93, 1.03) 0.4196 |
0.98 (0.91, 1.04) 0.4615 |
1.0019 (0.9980, 1.0058) 0.3439 |
>=0.1, < 0.2 |
0.91 (0.82, 1.00) 0.0560 |
0.87 (0.77, 0.99) 0.0305 |
0.9937 (0.9863, 1.0012) 0.0991 |
>=0.2, < 0.3 |
1.03 (0.89, 1.19) 0.6834 |
0.97 (0.83, 1.13) 0.6849 |
0.9937 (0.9841, 1.0034) 0.1999 |
>=0.3 |
0.94 (0.82, 1.09) 0.4369 |
0.85 (0.72, 0.99) 0.0407 |
0.9891 (0.9796, 0.9987) 0.0256 |
Interaction p value with LNMR |
0.7517 |
0.1972 |
0.055 |
LNMR = Lymph node metastasis risk; RLNC = Removed lymph node count; CSS = Cancer-specific survival; OS = Overall survival |
The diagnostic value of PLND for PCa staging is acknowledged. however, there are still debate with regard to the survival benefit of extended PLND. Before the present study, there have been a variety of retrospective studies comparing the oncological outcomes of extended PLND versus standard PLND. In a recent systematic review, the comparison of 21 retrospective studies on no PLND versus any form of PLND revealed no significant difference in favor of PLND for biochemical recurrence, metastasis free survival or CSS16. 11 of 13 studies comparing extended PLND and standard PLND in terms of biochemical recurrence did not demonstrate significant difference. While two studies showed a benefit from extended PLND in specific subgroups: intermediate-risk patients (96% versus 90%, p = 0.017) and pN1 patients with < 15% of retrieved lymph nodes affected (43% versus 10%, p = 0.01)17,18.
One recent prospective randomized trial compared the oncological outcomes between extended PLND and standard PLND, finding extended PLND provides better pathological staging and better biochemical recurrence free survival only in ISUP grade groups 3–5. This study provided a high level evidence compared with previous studies. However, CSS data were not available because of short follow-up and limited sample size14. Another prospective randomized trial found that extended PLND had no significant association with lower biochemical recurrence rate. However, in this study, the differences in nodal count and the rate of positive nodes between extended and standard PLND were smaller than expected15.
Another multicenter retrospective study included large sample of intermediate risk and high risk prostate cancer patients with long follow up, but didn’t find significant difference on biochemical recurrence, metastasis free survival and CSS between patients with and without extended PLND12. Despite of the retrospective nature and possible selection bias, these results suggested the negative therapeutic value of extended PLND was not probably due to limited sample size and short follow up.
Abdollah et al and Preisser et at found more extensive PLND improves survival in patients with node-positive and node-negative prostate cancer6,10. However, on one hand, because we can’t obtain the lymph node status before surgery, these results couldn’t have guidance on clinical decision; On the other hand, according to Will Rogers phenomenon, the lymph node status is affected by the range of PLND. This reason may potentially affect these conclusions.
Most previously studies had negative results whether in retrospective or prospective studies. We assumed that the significant value of extended PLND is more likely in patients with higher risk of LN metastasis, so we used the SEER data to confirm the hypothesis. As most studies obtained negative results including intermediate and high risk patients, we only included the patients with high risk prostate cancer in our study. The SEER data is retrospective and with limited information available, so we established the prediction model according to patients’ characteristics in the database to accurately evaluate the LN metastasis risk. We also excluded patients received adjuvant radiotherapy in the database. This can eliminate the radiotherapy’s effect on survival, and better reflect the pure effect of PLND on survival; on the other hand, because PLND may also affect the prognosis by guiding adjuvant therapy (including radiotherapy), such exclusion criteria might exclude these effects on the survival. The SEER database didn’t include the data about the extent of PLND, but only the removed lymph node count. Therefore, as the previous study10,13, we used the RLNC as an indicator for the extent of PLND.
We established a prediction model for LN metastasis based on the preoperative information from patients with RLNC over 6. The area under receiver operating characteristic curve of our prediction model for training sample is 0.829. The predictive LNMR in each LNMR group was also similar to the proportion of LN metastasis. By using the subgroup analysis, we found more RLNC (RLNC > 11) could significantly improve CSS and OS in the subgroup of LNMR ≥ 30% but not significant in the subgroup of LNMR < 30%. And the interaction analysis showed statistical significance between LNMR and RLNC with regard to CSS. One previous study analyzed the SEER database and found that PLND at RP is associated with low cancer specific mortality and overall mortality in the setting of metastatic PCa 8. In our study, we found the survival benefit of extended PLND was more significant in patients with higher risk of lymph node metastasis (over 30%). In these patients, we speculate that there might also be potential micro-metastasis that is undetected. Another possible explanation is that all metastatic PCa is at very high risk of LN metastasis. The consistence of these results suggested that PLND might provide survival benefit in very advanced prostate cancer with more LN metastasis or distant metastasis. However, this inference requires further validation.
Taken together, our findings have important clinical implications. We identified a subgroup of D’Amico high risk PCa patients that benefit significantly from more RLNC in PLND. Such an indication will be valuable for future design of randomized clinical trial for high risk PCa. The present prospective clinical trials recruited all patients with intermediate risk and high risk patients, and the subgroup analysis only included one category, such as PSA value, Gleason score or clinical T stage14. Future prospective studies might use LNMR as an indicator for subgroup analysis.
Our study is not devoid of limitations. First, we arbitrarily used 6 and 11 as the cutoff value of analysis. Although the LN count is almost consistent with the standard and extended PLND in a prospective study, we couldn’t obtain the specific PLND range from the SEER database. Therefore, our results revealed the prognostic significance of PLNC, but not the extent of PLND. Second, we only included patients not receiving adjuvant radiotherapy, so the conclusion couldn’t be applied to patients receiving adjuvant radiotherapy. On the other hand, the impact of extended PLND could be evaluated more objectively in our study. Third, the data from SEER database is retrospectively obtained and has a high potential for bias due to selection of patients undergoing surgery. Fourth, the prediction model for LNMR was established based on patients with RLNC over 6, but the results might still underestimate the metastatic LN. Fifth, the LNMR subgroup was also arbitrarily selected, and the results showed more RLNC provided survival benefit when LNMR ≥ 30%, which required further validation. Last but not least, several important variables were unavailable in the SEER database, including American Society of Anesthesiologists score, comorbidity, adjuvant hormonal therapy and baseline hematological and biochemical blood values that might potentially predict survival in PCa.
The present retrospective study showed that more RLNC in PLND is associated with better CSS and OS when the LNMR is over 30%. Prospective randomized trials are necessary to determine the value of extended PLND in PCa patients with high LNMR.
From the SEER Plus database between 2004 and 2019, a total of 250020 patients were retrieved with histologically confirmed adenocarcinoma of the prostate (International Classification of Disease for Oncology [ICDO-3] code 8140, site code C61.9) receiving radical prostatectomy (surgery site codes 50 or 70) and aged between 29 and 74 years. Only D’Amico high risk patients (≥ cT2c, PSA > 20 ng/mL, and/or Gleason score ≥ 8) treated with RP and PLND were included. The exclusion criteria included metastatic disease, unknown PSA level, T stage or Gleason score, unknown RLNC information about PLND, brachytherapy and adjuvant external beam radiation therapy. All patients had available data on follow-up. These selection criteria yielded 96875 patients. The flowchart of selection process was shown in Fig. 1.
Descriptive statistics included frequencies and proportions for categorical variables. Means ± standard deviation, or median (25th and 75th ranges) were reported for continuously coded variables. The Chi-square tested the statistical significance in proportions differences. The variance analysis was used to examine the statistical significance of means differences. RLNC was defined according to the SEER database and patients were stratified according to the median RLNC and 75th percentile. Multivariable COX regression analysis was used to assess the prognostic value of RLNC and CSS, OS.
First, a multivariate logistic regression model was established to test the association between LN metastasis and covariates (age of diagnosis, T stage, PSA level, Gleason score) in patients with RLNC over 6. From this model, the predicted LNMR was calculated for every patient based on the prediction model. Second, all the high risk patients were divided into four subgroups according to the LNMR (LNMR < 0.1, 0.1 ≤ LNMR < 0.2, 0.2 ≤ LNMR < 0.3 and LNMR ≥ 0.3). The 10-year cancer-specific mortality (CSM) rates were plotted against RLNC according to different LNMR subgroups. The 10-year CSM rates were also plotted against LNMR according to PLND range (the cutoff of RLNC was 6 and 11). Third, we applied the multivariable COX regression analysis in each LNMR subgroup to assess the impact of RLNC on CSS and OS. The covariates in multivariable COX model included preoperative PSA level, age at diagnosis, T stage, gleason score, race and marital status. The interaction analyses were performed to investigate the association between RLNC and LNMR (LNMR ≥ 0.3 and < 0.3).
A 2-tailded p < 0.05 was considered to be statistically significant in all analysis. Data were analyzed with the use of the statistical packages R (The R Foundation; http://www.r-project.org; version 4.2.0) and EmpowerStats (www.empowerstats.net, X&Y solutions, Inc. Boston, Massachusetts).
Data availability statement
The data that support the findings of the current study are available from the public Surveillance, Epidemiology, and End Results datebase (https://seer.cancer.gov/).
Acknowledgements
None.
Author Contributions
Concept: Xintao Li, Xu Zhang, Minghui Yang; data acquisition: Yu Gao, Yanzhong Liu, Shaoxi Niu; statistical analysis: Yu Gao; manuscript drafting: Xintao Li, Yu Gao, Yanzhong Liu; manuscript supervising: Jianye Li, Xu Zhang, Minghui Yang. All authors have read and agreed to the published version of the manuscript.
Conflict of Interest
The authors declare no competing interests.
Informed Consent
The data of this research were obtained from the public database and no informed consent was required.
Ethical approval statement
The data of this research were obtained from the public database and no ethical approval was required.