Selection of studies
A total of 948 studies were identified through database-searching, with a further 14 from hand-searching (Fig. 1). After removing duplicates, 808 titles and abstracts were screened. A total of 796 studies were excluded, and 12 full-text articles were assessed for eligibility. Of these, three were reviews [20,21,22], three did not specify the severity of liver disease [23,24,25], one had fewer than 10 patients that underwent appendicectomy  and one did not specify the number of patients who underwent appendicectomy . Four studies [16, 17, 28, 29] met the inclusion criteria.
Characteristics of included studies
The studies span an 18-year period from 2001 to 2019. Two studies were undertaken in the USA [17, 29], one in Denmark , and one in Japan . Three were nationwide database studies [16, 17, 29], and one, a retrospective observational study undertaken at two centres  (Table 1).
The total number of patients included was 168,134. There were 923 (0.5%) patients with cirrhosis and 167,211 (99.5%) without cirrhosis. From the data available, 56,654 (51.9%) patients were male and 52,417 (48.1%) were female. The treatment options in each study differed. Garci et al., included open appendicectomy (OA), laparoscopic appendicectomy (LA), and non-operative management (NOM). Tsugawa et al., included OA or LA, and Al-Azzawi et al., included LA only. Poulsen et al., did not specify the method of appendicectomy. In patients with cirrhosis, 630 underwent LA (76.0%), 55 OA (6.6%) and 144 (17.4%) NOM. In patients without cirrhosis, 83,851 underwent LA (77.4%), 12,610 OA (11.7%) and 11,768 (10.9%) NOM.
Mortality outcomes, post-operative complications, and length of stay
Following appendicectomy, three studies reported in-patient mortality [17, 28, 29] which ranged from 0 to 1.7% in patients with cirrhosis and 0.17–0.3% in patients without cirrhosis.
The USA nationwide database study by Garcia et al., reported mortality stratified by intervention (non-operative management, NOM; open appendicectomy, OA; laparoscopic appendicectomy, LA) and severity of cirrhosis (compensated cirrhosis; decompensated cirrhosis). The highest rate of mortality was in patients with decompensated cirrhosis managed with NOM (9.5%, 6/63). In the remaining decompensated cirrhosis cohort, the mortality rates were 4.3%, (2/47) following LA and 4.2% (1/24) following OA. However, the difference between management strategies in patients with decompensated cirrhosis was not statistically significant. In patients with compensated cirrhosis, mortality following LA (0.5%, 1/192) was significantly lower than both OA (3.2%, 1/31) and NOM (3.7%, 3/81) (p < 0.05). A similar trend was also found in patients without cirrhosis, with LA (0.1%, 87/83,473) having a lower mortality rate compared to both OA (0.6%, 81/12,610) and NOM (2.5% 290/11,768) (p < 0.05).
The Danish nationwide database study by Poulsen et al., reported 30-day mortality. In the group with cirrhosis, mortality was 9% (6/69), compared to 0.3% (413/58,982) in the group without cirrhosis (Table 2).
The three studies reporting hospital length of stay (LOS) [17, 28, 29] demonstrated an increased LOS in patients with cirrhosis following appendicectomy. Al-Azzawi reported LOS in patients with cirrhosis as 1.52 days compared to 1.1 in patients without cirrhosis. Garci et al., reported that LOS for both decompensated cirrhosis (9.5 ± 10.5) and compensated cirrhosis (7.1 ± 15.9) were significantly longer than patients without cirrhosis (3.4 ± 4.7) (p < 0.05). When comparing surgical techniques in patients with cirrhosis, Tsugawa et al. reported a significantly longer LOS (14.5 ± 4.3) following OA compared to LA (8.2 ± 2.6) (p < 0.05). Garcia et al. mirrored this trend in the decompensated cirrhosis cohort (OA 14.6 ± 11.7 vs. LA 6 ± 5.9, p < 0.05), compensated cirrhosis cohort (OA 11.8 ± 8.6 vs. LA 4.3 ± 4.1, p < 0.05), and in patients without cirrhosis (OA 4.8 ± 6.5 vs. LA 2.7 ± 3.1, p < 0.05).
Comparing patients with and without cirrhosis, Al-Azzawi et al., reported no significant difference in post-operative rates of pneumonia, urinary tract infections (UTI), surgical site infections (SSI), wound bleeding, pulmonary embolism (PE), C. difficile infections, or upper gastro-intestinal bleeding.
When comparing OA and LA in patients with cirrhosis, Tsugawa et al., showed a higher rate of SSI (OA 20.0%, 5/25 vs. LA 0%, 0/15) and wound bleeding (OA 20.0%, 5/25 vs. LA 0%, 0/15) in patients managed with OA (p < 0.05). Garcia et al., also report an increased incidence of SSI following OA in patients with compensated cirrhosis (OA 12.9%, 4/31 vs. LA 2.1%, 4/192), decompensated cirrhosis (OA 20.8%, 5/24 vs. LA 2.1%, 1/47), and patients without cirrhosis (OA 1.8%, 227/12,610 vs. LA 0.7%, 558/83,473) (p < 0.05).
All studies considered mortality as an outcome. The duration of follow-up in these studies ranged from in-hospital admission to 30 days post-operation. For the purpose of the meta-analysis, in-patient and 30-day mortality data were combined to calculate a pooled estimate. Patients undergoing NOM were excluded from the analysis. Due to the absence of a control group, results from Tsugawa et al. were excluded from the meta-analysis. Post-operative mortality for patients with cirrhosis was 1.76% (13/739) whereas for patients without cirrhosis it was 0.37% (582/155443). Compared with the control group, patients with cirrhosis had a significantly increased risk of mortality following appendicectomy (OR 9.92 (95% CI 4.67 to 21.06)), with a moderately-low heterogeneity between studies (I2 28%) (Fig. 2).
There was insufficient data to undertake a meta-analysis of LOS or complications.
Risk of bias
Three of the included studies were nationwide population-based reports and one was a multi-centre observational study. The level of evidence of these studies was level II/III. Reasons for high risk of bias included a lack of using secure records (e.g. medical records) to ascertain a diagnosis of liver cirrhosis, failure to control for confounding factors between the cirrhotic and non-cirrhotic group (e.g. age, ethnicity, co-morbidities), and short-follow-up time (Fig. 3).