Study selection

A total of 240 articles were identified in the initial step of the systematic literature search. After removing 44 duplicated and 186 articles by screening the titles and abstracts, the full-text review was performed for the remaining 10 potentially eligible studies. Finally, we included six articles in this review [12,13,14,15,16,17]. Figure 1 shows a flow diagram explaining the processes involved in identifying and including/excluding studies. The literature search strategy is summarized in Additional file 1: eTable 1.

Fig. 1

Flow diagram of the literature search

Demographics

Two randomized studies and four nonrandomized studies are presented in Table 1, which provides a general overview of patient demographics in the six articles. Among the six studies included, two types of ND were observed in the reported data. Radical ND referred to surgical clearance of all five lymph node (LN) groups (I–V), and selective ND consisted of the clearance of less than five LN groups of the ipsilateral neck. The pooled prevalence of the type of ND did not show a significant intergroup difference (P = 0.97). Additional file 1: eTable 2 and eTable 3 describe the results of the quality assessment. Two types of incisions, hidden and minimal, were used for END among the six studies. The hidden incisions, such as retroauricular or facelift incisions, were placed mainly behind the auricle and on the hairline to improve the cosmesis after ND. Minimal incisions, such as small submandibular or small suprasternal incisions, were used to perform LN dissection through a small cutaneous incision to reduce scarring after surgery.

Table 1 Basic characteristics of the included studiesOutcomesLN yield

Five of the six included studies reported the number of LNs retrieved from ND [12,13,14,15,16,17]. The pooled analysis of the overall study group did not demonstrate a significant difference between the two groups regarding the number of LNs yielded (MD, 0.43; 95% confidence interval [CI], − 0.44 to 1.29) (Fig. 2A). Meta-analysis of the three studies using hidden incisions revealed a similar number of LNs yielded between the two ND groups (MD, − 0.46; 95% CI − 2.88 to 1.96) (Fig. 2B) [13, 16, 17]. Meta-analysis of the two studies using minimal incisions also showed no intergroup differences (MD, 0.66; 95% CI − 0.43 to 1.74) (Fig. 2C) [12, 15].

Fig. 2

Forest plot of the LN yield. A Overall study group. B Studies using hidden incisions in END. C Studies using minimal incisions in END. CI, confidence interval; CND, conventional neck dissection; END, endoscope-assisted neck dissection

Operative time

Five of the six studies recorded the operative time needed to complete ND [12,13,14,15, 17]. The pooled results of the overall study groups showed that the operative time was longer in the END group than the CND group (MD, 30.72; 95% CI 12.27 to 49.17). An approximately 30-minute difference in the operative time was observed (Fig. 3A). Meta-analysis of the two hidden-incision studies showed that the operative time was longer in the END group (MD, 18.81; 95% CI 13.48 to 24.15) (Fig. 3B) [13, 17]. Meta-analysis of the three minimal-incision studies also revealed that the operative time was longer in the END group (MD, 35.18; 95% CI 9.65 to 60.72) (Fig. 3C) [12, 14, 15]. Further subgroup analysis was not performed due to the limited number of eligible studies.

Fig. 3

Forest plot of the operative time. A Overall study group. B Studies using hidden incisions in END. C Studies using minimal incisions in END. CI, confidence interval; CND, conventional neck dissection; END, endoscope-assisted neck dissection

Intraoperative blood loss

Four of the six studies reported intraoperative blood loss in ND [12, 14, 15, 17]. The pooled analysis showed comparable amounts of intraoperative blood loss between the two groups (MD, 3.12; 95% CI − 18.59 to 24.83) (Fig. 4A). Meta-analysis of hidden-incision studies could not be performed due to limited study numbers. A meta-analysis including three studies using minimal incisions demonstrated that intraoperative blood loss was lower in the END group (MD, − 10.26; 95% CI − 19.49 to − 1.03) (Fig. 4B) [12, 14, 15]. Subgroup analysis according to types of incision showed that the subgroup difference between hidden- and minimal-incision studies was significant, suggesting that it was a potential source of heterogeneity (P for subgroup difference < 0.001).

Fig. 4

Forest plot of intraoperative blood loss. A Overall study group. B Studies using minimal incisions in END. CI, confidence interval; CND, conventional neck dissection; END, endoscope-assisted neck dissection

Length of hospital stay

The three minimal-incision studies recorded the length of hospital stay after ND [12, 14, 15]. The pooled results showed that the length of hospital stay was shorter in the END group (MD, − 1.13; 95% CI − 1.86 to − 0.41) (Fig. 5A).

Fig. 5

Forest plot of the length of hospital stay and recurrence. A Forest plot of the length of hospital stay. B Forest plot of local recurrence in the ipsilateral neck. C Forest plot of regional recurrence in the ipsilateral neck. CI, confidence interval; CND, conventional neck dissection; END, endoscope-assisted neck dissection

Ipsilateral nodal recurrence

Two of the studies recorded the local and regional nodal recurrence of the ipsilateral neck in the follow-up period [12, 17]. The pooled results showed that the incidence of local nodal recurrence was similar between the two groups (RD, − 0.02; 95% CI − 0.08 to 0.03) (Fig. 5B). The pooled results also demonstrated no significant intergroup difference regarding the incidence of regional nodal recurrence (RD, − 0.01; 95% CI − 0.05 to 0.04) (Fig. 5C).

Postoperative marginal mandibular nerve injury

The incidence of marginal mandibular nerve injury was reported by four of the included studies [12, 14, 15, 17]. Temporary injury was observed in two studies [12, 14]. Two other studies did not specify whether the injury was temporary or permanent [15, 17]. The data, regardless of the type of nerve injury, were pooled for analysis, and the results demonstrated comparable incidence between the two groups (RD, 0.00; 95% CI − 0.05 to 0.06) (Fig. 6A). Pooled analysis of the two studies reporting temporary marginal mandibular nerve injury also revealed a similar incidence between the two groups (Additional file 1: eFigure 1).

Fig. 6

Forest plot of postoperative complications. A Marginal mandibular nerve injury. B Skin-edge necrosis. C Hematoma or bleeding. D Seroma. CI, confidence interval; CND, conventional neck dissection; END, endoscope-assisted neck dissection

Postoperative skin edge necrosis

Three of the six studies reported the incidence of skin edge necrosis [12, 16, 17]. The pooled analysis did not demonstrate a significant intergroup difference (RD, 0.09; 95% CI − 0.03 to 0.20) (Fig. 6B).

Postoperative hematoma or bleeding

Four of the six studies reported the incidence of hematoma or bleeding after ND [12, 15,16,17]. The pooled results demonstrated a comparable incidence of postoperative hematoma or bleeding (RD, − 0.00; 95% CI − 0.05 to 0.04) (Fig. 6C).

Postoperative seroma

Two of the six studies reported the incidence of seroma formation after ND [12, 17]. The pooled results showed that the incidence was similar between the two groups (RD, 0.09; 95% CI − 0.11 to 0.29) (Fig. 6D).

Publication bias

Additional file 1: eTable 4 shows the funnel plots as well as the results of Egger's and heterogeneity tests. No evident publication bias was noted based on Egger's intercept test.