Characteristics of the included studies
The process of study selection is summarized in Fig. 1. Database and additional searches retrieved a total of 1,014 records, from which 25 articles were ultimately included in our meta-analysis [4–6, 9, 10, 13, 20–38] (Table 1). All 25 studies had a retrospective cohort design, they involved a total of 19,926 participants, and they came from 17 countries: United States (n = 5), Turkey (n = 6), China (n = 2), Spain (n = 1), Japan (n = 1), Italy (n = 1), and Tunisia (n = 1). The studies included data from two publicly available databases: the National Cancer Database (NCDB) (n = 4) and the Surveillance, Epidemiology, and End Results (SEER) database (n = 4), both from the US. (Table 1 should be placed after this paragraph.)
Table 1
Baseline and follow-up data on patients in the included retrospective studies
Tumor subtype | Author (ref) | Inclusion year | LNI | LND | Non-LND | n | Country | NOS | Surgery type | FIGO stage | Longest follow-up* (months) |
HGESS or UUS | Nasioudis (9) | 2004–2015 | 22 | 280 | 126 | 406 | NCDB | 8 | TH | I | 168 |
Cabrera (20) | 1995–2019 | 5 | 20 | 12 | 32 | Spain | 7 | Various | I-IV | 240 |
Ayhan (22) | 2008–2017 | 14 | 54 | NR | 54 | Turkey | 7 | TH ± BSO | I-IV | NR |
Seagle (13) | 1998–2013 | 141 | 712 | NR | 712 | NCDB | 8 | TH | could not be determined |
uLMS | Ayhan (10) | 1996–2018 | 8 | 162 | NR | 162 | Turkey | 8 | TH + BSO | I-IV | NR |
Nasioudis (9) | 2004–2015 | 42 | 1250 | 2267 | 3517 | NCDB | 8 | TH | I | 168 |
Takehara (24) | 2000–2012 | NR | 33 | 227 | 260 | Japan | 8 | TH + oophrectomy | I-IV | 169.2 |
Nesrine (26) | 2000–2014 | NR | 18 | 13 | 31 | Tunisia | 7 | TH ± BSO | I-IV | 207 |
Machida (5) | 1973–2013 | 188 | 3749 | NR | 3749 | SEER | 7 | TH and others | I-IV | NR |
Raspagliesi (29) | 2004–2014 | NR | NR | NR | 91 | Italy | 6 | TH ± BSO | I | 49.6 |
Seagle (4) | 1998–2013 | 189 | 2255 | NR | 2255 | NCDB | 8 | TH ± BSO | I-IV | NR |
Tasci (30) | 1993–2009 | 8 | 36 | 59 | 95 | Turkey | 7 | TH ± BSO | I-IV | 183 |
Ayhan (32) | 1982–2007 | 4 | 34 | NR | 34 | Turkey | 6 | TH + BSO | I-IV | NR |
Kapp (34) | 1988–2003 | 23 | 347 | NR | 347 | SEER | 8 | TH + BSO | I-IV | NR |
Akahira (35) | 1990–2004 | NR | 3 | 27 | 30 | Japan | 7 | TH + BSO | I-IV | 110 |
Giuntoli (37) | 1976–1999 | 4 | 36 | NR | 36 | USA | 8 | TH + BSO | I-IV | NR |
Leitao (36) | 1982–2001 | 3 | 37 | NR | 37 | USA | 7 | TH ± oophrectomy | I-IV | NR |
Major (38) | 1979–1988 | 2 | 57 | NR | 57 | USA | 6 | NR | I-II | NR |
LGESS | Nasioudis (9) | 2004–2015 | 19 | 826 | 495 | 1321 | NCDB | 8 | TH | I | 168 |
Ayhan (21) | 2008–2017 | 12 | 81 | 63 | 144 | Turkey | 8 | TH ± BSO | I-IV | 156 |
Zhang (23) | 1969–2017 | 2 | 47 | 72 | 119 | China | 8 | TH + BSO | I | 576 |
Comert (27) | 1985–2016 | 3 | 21 | NR | 21 | Turkey | 8 | TH + BSO | I-III | NR |
Zhang (25) | 1998–2016 | 2 | 32 | NR | 32 | China | 7 | TH + BSO | I-IV | NR |
Machida (5) | 1973–2013 | 192 | 2198 | NR | 2198 | SEER | 7 | TH and others | I-IV | NR |
Seagle (13) | 1998–2013 | 87 | 846 | NR | 846 | NCDB | 8 | TH | could not be determined |
Shah (33) | 1988–2005 | 7 | 100 | 283 | 383 | SEER | 8 | TH + BSO | I-IV | NR |
ADS | Nasioudis (9) | 2004–2015 | 21 | 464 | 704 | 1168 | NCDB | 8 | TH | I | 168 |
Zhang (25) | 1998–2016 | 0 | 6 | NR | 6 | China | 7 | TH + BSO | I-IV | NR |
Nathenson (28) | 1982–2014 | 1 | 54 | 101 | 155 | USA | 8 | TH + BSO | I-IV | 182.4 |
Machida (5) | 1973–2013 | 29 | 877 | NR | 877 | SEER | 7 | TH and others | I-IV | NR |
Seagle (6) | 1998–2011 | 21 | 677 | NR | 677 | NCDB | 7 | TH + BSO | could not be determined |
Carroll (31) | 1982–2011 | 1 | 22 | 52 | 74 | USA | 8 | TH + BSO | I-IV | 241.1 |
| Overall | | 1,050 | 15,334 | 4,501 | 19,926 | | | | | |
* Data shown only for studies included in survival analysis. |
LNI
Incidence Rate Of Lni
Twenty studies [4–6, 9, 10, 13, 20–23, 25, 27, 28, 30–34, 36, 38] involving 15,316 patients investigated the incidence rate of LNI. A higher incidence rate of LNI was identified in patients with HGESS/UUS (rate 18%; 95% CI 9–28%; I2 = 89.67%, p < 0.001), while a relatively low LNI was observed in patients with uLMS (rate 6%; 95% CI 4–8%; I2 = 84.88%, p < 0.001), LGESS (rate 7%; 95% CI 4%-11%; I2 = 90.03%, p < 0.001), or ADS (rate 2%; 95% CI 1%-3%; I2 = 0%, p < 0.001; Fig. 2).
Because of the high heterogeneity in this meta-analysis, subgroup analysis based on data source, publication year, and sample size was performed. Sensitivity analysis was also performed. Neither analysis (data not shown) identified obvious sources of heterogeneity.
Relationship Between Lni And Os
Five studies [4–6, 33, 34] evaluated the relationship between LNI and OS in 10,203 patients with ADS, uLMS, or LGESS. The pooled data associated LNI with worse OS in these patients (HR 1.85, 95% CI 1.49–2.30; I2 = 75.5%, p < 0.001; Fig. 3).
Given the high heterogeneity of the pooled data, subgroup analysis was conducted based on subtype, country, publication year and sample size. Sensitivity analysis was also performed. Neither analysis (data not shown) identified obvious sources of heterogeneity.
Relationship Between Lymphadenectomy And Survival
Lymphadenectomy and OS
Eleven studies [9, 20, 21, 23, 24, 28–31, 33, 35] including 7,795 patients investigated the relationship between lymphadenectomy and OS. Lymphadenectomy did not improve OS in patients with LGESS (1,967 cases, HR 1.21, 95% CI 0.95–1.54; I2 = 0%, p = 0.121) or ADS (1,397 cases, HR 0.90, 95% CI 0.73–1.10; I2 = 0%, p = 0.297), but it was associated with worse OS in uLMS (3,993 cases, HR 1.14, 95% CI 1.03–1.27; I2 = 0%, p = 0.011; Fig. 4). Conversely, lymphadenectomy was associated with better OS in HGESS/UUS (438 cases, HR 0.63, 95% CI 0.48–0.85; I2 = 0%, p = 0.002; Fig. 4).
Lymphadenectomy And Dfs
Four studies [10, 26, 28, 30] including 593 patients with uLMS or ADS were pooled in a meta-analysis investigating the relationship between lymphadenectomy and DFS. Lymphadenectomy was not associated with better DFS in patients with uLMS or ADS (593 cases, HR 0.87, 95% CI 0.61–1.26; I2 = 54.9%, p = 0.473; Fig. 5).
Publication Bias
Publication bias was assessed in the meta-analysis of the incidence rate of LNI based on 20 studies [4–6, 9, 10, 13, 20–23, 25, 27, 28, 30–34, 36, 38]. The funnel plot showed no obvious asymmetry and Begg’s test was not significant (p = 0.128), suggesting no publication bias (Additional file 1).