Para-Aortic Lymph Node Dissection for Patients with Node-Negative Cervical Squamous Cell Carcinoma: A Retrospective Cohort Study

,


Background
Cervical cancer is a malignant tumor in the female reproductive system with the highest incidence worldwide [1]. Te widespread application of cervical cytology screening in recent decades has enabled the early detection and treatment of cervical cancer, and the mortality of cervical cancer has decreased signifcantly [2]. For patients with early-stage cervical cancer, radical surgery is still the primary treatment [3].
In the current National Comprehensive Cancer Network (NCCN) guidelines, the recommendation for early-stage operable cervical cancer patients is to perform radical hysterectomy (RH) and pelvic lymph node dissection (PLND) with or without para-aortic lymph node dissection (PALND) [4]. Whether the para-aortic lymph nodes (PALNs) need to be removed has not been confrmed. In this situation, how can we decide whether to perform PALND?
It is well accepted that patients with PALN metastasis show poorer overall survival (OS) and progression-free survival (PFS) following surgical resection than patients with the same International Federation of Gynecology and Obstetrics (FIGO) stages (2009 version) who do not have PALN metastasis [5], and PALN involvement has become an essential part of cervical cancer staging in the new FIGO staging system (2018 version) [6]. Additionally, extended-feld radiation is indicated in cases of PALN dissemination [7]. Terefore, for patients with suspected PALN metastasis, surgeons should remove the nodes without hesitation. However, PALN metastases are found in less than 10% of patients with early-stage cervical cancer [8], and the anatomy surrounding PALNs is complex. Te routine addition of PALND might add considerable operation time and increase surgical complications, such as vascular, ureteral, and nerve injury, as well as the risk of infection, lymphocyst formation, lymphedema, and thrombophlebitis, even by an expert surgeon [9,10]. For patients with clinically negative PALN, should we perform PALND routinely? To answer this question, we must know whether additional PALND translates to a survival beneft in patients with negative PALNs. To the author's knowledge, there is currently no large-scale study on patients with negative PALN only.
In contrast to other gynecological tumors, the lymph node metastasis pattern of cervical cancer is relatively clear and has been studied and reviewed in detail, from regional lymph nodes to distant lymph nodes [11]. In general, cancer cells from the cervical tumor frst spread to the nodes of the obturator and external iliac, which are the sentinel lymph nodes of the cervix, and then proceed to the common iliac and PALN, similar to a "ladder," where the common iliac lymph node is a key site in the middle of the ladder [12]. It is very rare that PALNs are positive and pelvic lymph nodes are negative, occurring in less than 1.5% of patients with earlystage cervical cancer [13]. Teoretically, the probability of positive PALNs and negative common iliac lymph nodes should be even lower. Terefore, for patients who have undergone common iliac lymph node resection and are confrmed as negative by pathology, there is a small likelihood of overlooking PALN metastasis.
Cervical squamous cell carcinoma (CSCC) accounts for more than 80% of cervical cancers, and the lymph node metastasis rate is lower than that of other pathological types of cervical cancer [14]. Terefore, we planned this large retrospective cohort study on PALND for patients with early-stage CSCC who were PALN-negative. Patient data from our center were collected and analyzed. Propensity score matching (PSM) [15] was used for protocol analysis.

Patients.
After Institutional Review Board approval, the medical records of all CSCC patients with FIGO stage IB1-IIa2 (2009 version) who underwent abdominal RH ± bilateral salpingo-oophorectomy and PLND ± PALND between 2006 and 2014 at Fudan University Shanghai Cancer Center (FUSCC) were reviewed using our institutional cancer registry. Te study was approved by the ethics committee at our center and the work has been reported in line with the STROCSS criteria [16].

Treatment.
All the enrolled patients had undergone standard surgery by experienced gynecological oncologists in our center. PALND is usually conducted along the fow of the abdominal lymphatics up to the level of the inferior mesenteric artery (IMA). When infra-IMA node metastasis was suspected by preoperative evaluation or intraoperative exploration, PALNs were also removed up to the level of the left renal vein. Adjuvant treatment was provided to patients with risk factors according to the NCCN guidelines. Individuals who had two or more intermediate-risk factors (tumor diameter ≥4 cm, ≥1/2 depth of stromal invasion, or lymphovascular space invasion (LVSI) positive) and those who had a minimum of one high-risk factor (parametrium positive, lymph node metastasis, or surgical margins involved) received adjuvant radiotherapy or concurrent chemoradiotherapy.

Follow-Up.
After the treatment was completed, the patients were followed up regularly: every 3 months for the frst two years, every 6 months from the second year to the ffth year, and once a year thereafter. Te follow-up items included symptom consultation, pelvic physical examination, squamous cell carcinoma antigen (SCC-Ag), ultrasonography, computed tomography (CT) scan, magnetic resonance imaging (MRI) scan, and vaginal cytology test. Te site of recurrence was documented as locoregional or distant, and it was recorded as a distant recurrence when both were present. Distant recurrences were identifed as metastases beyond the pelvis. PFS was defned as the time (months) from diagnosis to disease recurrence. OS was defned as the time (in months) from initial diagnosis to death from any cause. Data on patients with no evidence of disease recurrence or death were censored at the date of the last follow-up.

Study Design.
Te fow chart of the study is shown in Figure 1. A total of 3,995 patients with stage IB1-IIA2 CSCC underwent radical surgery in our center. After excluding patients with incomplete information, 3,085 patients remained. Ten, we excluded patients with positive PALNs, less than 3 months of follow-up, positive or unknown common iliac lymph nodes, and uncertain recurrence sites. Finally, 2,427 patients were included in this study. According to the aforementioned theory of lymph node metastasis patterns in cervical cancer, for patients who had undergone common iliac lymph node resection and had negative postoperative pathology, the PALN was also defaulted to negative. In this set of data, only 2 patients (0.47%) had negative common iliac lymph nodes, but postoperative pathology showed positive PALNs, indicating that common iliac lymph nodes can indeed be used as an indicator of the status of PALNs. Te number of remaining patients with "RH + PLND" and negative common iliac lymph nodes was 2,095, while 332 patients who underwent "RH + PLND + PALND" and PALN removal were negative. Ten, we used PSM to keep the baselines of the two groups ("−PALND" and "+PALND") of patients consistent (ratio � 1, caliper � 0.01), thereby reducing possible confounding factors in the subsequent prognostic analysis. PSM was performed through a multiple logistic regression model in which PALND was the dependent variable and the covariates were potential confounding factors, including patient age, menopausal status, FIGO stage, tumor diameter, depth of stromal invasion, parametrial invasion, LVSI, vaginal margin invasion, pelvic lymph node metastasis, adjuvant therapy, and preoperative serum SCC-Ag. Patients who were found to be outside the caliper were excluded. Unmatched patients were also excluded.

Statistical
Analysis. Continuous variables, including patient age, number of PALNs resected, and preoperative SCC-Ag, were divided into two groups by their respective medians. Pearson's chi-square test and Fisher's exact test were used to compare clinicopathological factors between groups. Survival curves between diferent groups of patients were estimated by Kaplan-Meier survival curves and the logrank test, with stratifcation according to the factors. Calculations of hazard ratios (HRs) with 95% confdence intervals (CIs) were made through multivariate analyses using Cox proportional hazards models to evaluate the prognostic factors related to survival. All statistical analyses were performed on all matched/paired patients. P < 0.05 was considered to indicate a statistically signifcant diference. Te statistical analyses of the observed data were performed by SPSS (Version 22.0, Chicago, USA) and R (Version 3.5.1, Vienna, Austria). Figures 1 and S1, the patients were matched by PSM: 313 were allocated to the −PALND group (cohort 1), and 313 were allocated to the +PALND group (cohort 2). Te median patient age at the time of surgery was 47 years (range, 22-73 years), the median preoperative SCC-Ag value was 6.5 ng/ml (range, 0.1-70 ng/ml), and the median follow-up time was 83.2 months (range, 4-161 months). Table 1 summarizes the patient demographics and tumor characteristics before and after matching. Before matching, statistically signifcant diferences were noted between the two groups with respect Patients with stage IB1 -IIA2 CSCC following radical surgery in FUSCC (n=3995) Propensity Score Matching Analysis

OS and PFS of the Matched Cohorts: −PALND
vs.+PALND. At the last follow-up, 26 patients died, and 66 patients experienced recurrence. When we compared the prognosis of the two cohorts, we found a remarkable result: Kaplan-Meier analysis and the log-rank test revealed that patients in cohort 2 with PALND showed poorer prognosis than patients in cohort 1 without PALND in terms of OS (P � 0.014, Figure 2(a)); however, there was no signifcant diference in PFS between the two groups (P � 0.521, Figure 2(b)). To eliminate bias from other prognostic factors, such as FIGO stage, tumor diameter, depth of stromal invasion, LVSI, parametrial invasion, vaginal margin invasion, pelvic lymph node metastasis, adjuvant therapy, and SCC-Ag, we examined the impact of these prognostic factors on survival by multivariate analysis. Te results showed that PALND was an independent prognostic factor for OS (P � 0.021, Figure 3).

OS of the Matched PALND Cohort
According to the Number of PALNs Removed. Next, we analyzed whether patient OS was related to the number of PALNs removed. According to the median value, we divided the +PALND group into the PALND < 5 and PALND ≥ 5 groups, and Kaplan-Meier analysis showed that there was no signifcant diference in OS between the two groups (P � 0.709, Figure 4).

Recurrence Patterns between Groups.
In total, 28 patients in the −PALND group and 38 patients in the +PALND group experienced relapse. In terms of the site of recurrence, there were no diferences in recurrence patterns between the two groups, but there was a higher but not signifcant distant recurrence rate in patients with PALND (69.4% vs. 65.2%, P � 0.774, Table S1).

Subgroup Analysis.
In subgroup analysis, subgroups were stratifed by the variables known at the time of diagnosis, which included patient age, menopausal status, FIGO stage (2009 version), tumor size, and preoperative SCC-Ag value. As shown in Figure 5, the Kaplan-Meier method and log-rank test results showed that the OS of patients in cohort 2 was worse than that of patients in cohort 1 when they were aged ≤47 years (P � 0.033), were premenopausal (P � 0.032), were in stage IB (P � 0.009), or had SCC-Ag <6.5 (P � 0.009). A Cox regression analysis was conducted with various prognostic and clinicopathological factors in each subgroup, as shown in Figure S2. Te multivariate analysis results showed that PALND was an independent risk factor for OS in patients aged ≤47 (P � 0.045), those who were premenopausal (P � 0.045), those in stage IB (P � 0.031), and those with SCC-Ag <6.5 (P � 0.020).

Discussion
Te survival rate of patients with early-stage cervical cancer is very high, and some patients can even be cured [17]. Tis study had an average of more than 80 months of long-term follow-up, and the patients' survival rate was still higher than 90%. Te pursuit of these patients is not only to live but also to live with quality. At present, a series of studies at our center have confrmed that patients with early cervical Journal of Oncology 5 cancer can safely and efectively undergo radical trachelectomy, which preserves fertility [18][19][20]. Tis shows the great progress brought by precision medicine. Extensive lymph node resection, especially PALND, may cause serious complications. To determine whether the lymph nodes need to be removed, more research is needed. Lymphadenectomy has always been regarded as a very important step during surgery for many kinds of malignant tumors [21,22]. However, as medical technology becomes increasingly accurate, there is an increasing discussion about the necessity and extent of lymphadenectomy. A randomized controlled study conducted by Sasako et al. showed that treatment with D2 lymphadenectomy plus PALND does not improve the survival rate in curable gastric cancer compared with D2 lymphadenectomy alone [23]. Similar conclusions are gradually being reported in pancreatic neuroendocrine tumors [24], bladder cancer [25], and papillary thyroid cancer [26]. In gynecologic tumors, the famous LION study showed that for patients with advanced ovarian cancer who reach R0 and have no abnormal lymph nodes before and during the operation, compared with patients without lymph node dissection, lymph node dissection has no signifcant impact on OS or PFS but increases the incidence of postoperative complications [27]. In endometrial cancer, there is also a very authoritative randomized controlled study that found no evidence of a beneft in terms of OS or PFS for pelvic lymphadenectomy in women with early endometrial cancer [28]. However, no similar high-level studies have been reported for cervical cancer. Our previous study showed that the number of positive lymph nodes truly afected the prognosis of patients with early-stage CSCC [29,30]. Te resection of too many "useless" lymph nodes may not have a signifcant efect on the prognosis of patients. Te present study found that for patients with negative PALN, PALND may be "rubbing salt on the wound." Taking the risk of complications may even further afect the patient's prognosis.
Actually, a small number of retrospective analyses have attempted to evaluate the therapeutic role of PALND in cervical cancer. Tsuruga et al. showed that indications for PALND in the surgical treatment of stage IB2, IIA2, or IIB cervical cancer need to be individualized. Patients with common iliac lymph node metastasis are possible candidates [10]. Gonzalez-Benitez et al. reported that the survival beneft of PALND in patients with advanced cervical cancer is lacking [31]. Del Carmen et al. emphasized that in patients undergoing RH and PLND for stage IA2-IB2 cervical cancer, PALND is not warranted based on the low risk of isolated metastatic disease and lack of survival beneft associated with the procedure [13]. Te results of these studies indicate that patients with cervical cancer should be treated individually, and surgeons may choose not to remove the PALN in some special cases. However, in this study, we focused on patients with stage IB1-IIA2 CSCC who can undergo surgery and used the "ladder" rule of cervical cancer lymph node metastasis to accurately screen patients with negative PALNs.  Journal of Oncology Teoretically, this part of the population should include patients least able to beneft from PALND. Depending on whether PALND was performed, we divided the patients into two groups, and patient background was ideally balanced in the groups by matching; therefore, the comparison of −PALND and +PALND was considered reliable. Interestingly, the results showed that although there was no signifcant diference in PFS between the two groups, the OS of patients with PALND was worse. Te results of the multivariate analysis also showed that PALND was an independent predictor of OS. At the same time, to verify whether the prognosis of patients is diferent due to an insufcient number of PALNs resected, we grouped the patients by the number of lymph nodes removed, and the results showed that there was no diference in the prognosis between the two groups. Tis result challenges conventional cognition and reminds surgeons that when treating patients with clinical PALN negativity, the options of "yes or no" may no longer be relevant since "absolutely not" may be a preferred option.
One of the reasons for this discrepancy may be that local immune function is damaged after lymph node removal. Te lymphatic system is the frst line of immune defense [32]. A recent study clearly showed that regional lymph nodes play an important role in antitumor immunity, particularly in microsatellite instability-high (MSI-H)/defcient mismatch repair (dMMR) colorectal cancers with a high tumor mutational burden (TMB), requiring careful consideration of excessive nonmetastatic lymph node dissection in MSI-H/ dMMR colorectal cancer patients [33]. Terefore, for patients with negative lymph nodes, maintaining the integrity of the lymphatic system as much as possible and protecting the patient's immune function may improve the patient's survival rate and quality of life, while blindly clearing the entire lymphatic system may be counterproductive.
Due to the uncertainty of the current guidelines, when the clinical evaluation of the PALN is negative, the surgeon will inevitably have subjective factors when selecting patients who need to undergo PALND. Sometimes patients who are young, are premenopausal, have a low BMI, and are not prone to postoperative complications may be more suitable. To help better select preoperative patients, our subjects were then stratifed according to the variables known at the time of diagnosis, and the data were further analyzed. Te results showed that for patients younger than 47 years, those who were premenopausal, those in FIGO stage IB, and those with SCC-Ag lower than 6.5, patients with PALND had worse survival. Furthermore, when multivariate analysis was performed, PALND was an independent factor. Tis reminds us that when treating these patients, if the clinical evaluation of the PALN is negative, we need to be very cautious when choosing PALND.
Tere are several limitations to this study, the most obvious being its retrospective nature at a single institution. Because of the nonrandomized nature of the study design and potential allocation biases arising from the retrospective comparison between groups, PSM was utilized to keep the patient characteristics of the two cohorts well-balanced, but unknown confounders still represent a source of bias.
Another shortcoming is the lack of statistical data on the occurrence of postoperative complications in this study. Whether PALND will increase the incidence of complications cannot be answered by this study, and it is doubtful whether the poor survival of patients with PALND is related to serious adverse events. In addition, the patients included in this study were only CSCC patients, and patients with cervical cancer of other pathological types were not included in this study. More studies are needed to determine the relationship between PALND and patient survival in these pathological types.
In summary, in contrast to other studies, our data showed that PALND had a negative impact on the OS of patients with early-stage CSCC. Te pattern of recurrence was similar with and without PALND, and omitting PALND did not increase the rate of distant recurrence. For patients with cervical cancer who are clinically evaluated as PALNnegative, especially those who are younger, who are premenopausal, whose tumors are confned to the cervix (stage IB), and who have relatively low SCC, PALND needs to be selected with special care. Te convention of feeling free to decide whether to perform PALND needs to be changed. It is recommended that we strive to fnd more accurate, convenient, and economical diagnostic methods for PALN metastasis. Nevertheless, whether PALND truly afects the OS of patients with cervical cancer and whether the guidelines need to clearly indicate which patients are not suitable for PALND require a higher level of evidence, that is, a prospective randomized controlled study, for verifcation.

Data Availability
Te data used to support the fndings of this study will be made available on request to the corresponding author.

Conflicts of Interest
Te authors declare that they have no conficts of interest.
Shanghai Key Clinical Specialty Project (grant no. shslczdzk06301). Table S1: site of frst tumor recurrence. Figure S1: propensity scores of diferent groups (A). Distribution of the propensity scores (B). Figure S2: multivariate analysis of clinicopathological factors for OS in patients aged ≤47 years (A), patients who were premenopausal (B), patients with FIGO stage IB (C), and patients with SCC-Ag<6.5 ng/mL (D). (Supplementary Materials)