Association of Leptin and Adiponectin with Risk and Prognosis of Hepatocellular Carcinoma: A Combination of Traditional, Survival and Dose-response Meta-analysis

Background: The association between leptin, adiponectin levels and the risk as well as prognosis of hepatocellular carcinoma has been investigated by an increasing number of studies, but the results were controversial. Methods: A meta-analysis was performed to assess the correlation between leptin, adiponectin levels and risk and prognosis of hepatocellular carcinoma (CRD42020195882). Through June 14, 2020, PubMed, Cochrane Library, Embase databases, Clinicaltrials, and Opengrey were searched, including references of qualifying articles. Titles, abstracts, and main texts were reviewed by at least 2 independent readers. Stata 16.0 was used to calculate statistical data. Results: Thirty studies were included in this meta-analysis and results showed that hepatocellular carcinoma group had signicantly higher leptin levels than the cancer-free control group (SMD = 1.83, 95% CI (1.09, 2.58), P = 0.000) , the healthy control group (SMD = 4.32, 95% CI (2.41, 6.24), P = 0.000) and the cirrhosis group (SMD = 1.85, 95% CI (0.70, 3.01), P = 0.002). Hepatocellular carcinoma group had signicantly higher adiponectin levels than the healthy control group (SMD = 1.57, 95% CI (0.37, 2.76), P = 0.010), but no statistical difference compared with the cancer-free control group (SMD = 0.24, 95% CI (-0.35, 0.82), P = 0.430) and the cirrhosis group (SMD = -0.51, 95% CI (-1.30, 0.29), P= 0.213). The leptin rs7799039 polymorphism was associated with increased risk of hepatocellular carcinoma (G vs A: OR = 1.28, 95% CI (1.10, 1.48), P = 0.002). There were linear relationships between adiponectin levels and the risk of hepatocellular carcinoma (OR = 1.066, 95% CI (1.03, 1.11), P = 0.001). In addition, the results showed that high/positive expression of adiponectin was signicantly related to lower overall survival in hepatocellular carcinoma patients (HR = 1.70, 95% CI (1.22, 2.37), P = 0.002); however, there was no signicantly association between the leptin levels and overall survival (HR = 0.92, 95% CI (0.53, 1.59), P = 0.766). Conclusion: The study shows that high leptin levels are associated with a higher risk of hepatocellular carcinoma. Adiponectin levels are proportional to hepatocellular carcinoma risk, and are related to the poor prognosis. The present study shows that high leptin levels are associated with a higher risk of HCC, which may represent a useful biomarker for early detection of HCC. AdipoQ levels are proportional to HCC risk, with a linear dose-response relationship, and are related to the poor prognosis, which may be a useful biomarker for estimating the prognosis of HCC.

review, editorials, letters to the editor, and experimental animal studies were excluded. When studies reported on the same or overlapping patient populations, only the study with the most complete data set and the most rigorous methodology was used.

Data extraction
Extracted the basic information from each study: the rst author, year of publication, country, study design, study period, number of subjects and sample source. In addition, Extracted the following information in Table 1: the source of the case group, the source of the CFC group, age, gender, body mass index, measured indicators and detected method. Summarized the following information in Table 2: the source of case/control, matching variables, single-nucleotide polymorphisms, genotyping methods, frequency of case and control genotype. Selected the following information in Table 3: follow-up, measured indicators, detected method, cut-off value, survival analysis, source of HR and analytic method. If there was inaccurate or missing information extracted from the original article, we tried to contact the corresponding authors of studies to guarantee data accuracy.

Quality assessment
The Newcastle-Ottawa Scale (NOS), a risk assessment tool recommended by the Cochrane collaboration, was applied to evaluate the quality of nonrandomized controlled studies. The following factors were taken into consideration: patient selection, comparability of the study groups, and assessment of outcome. The maximum score obtained by this scoring system was 9, and studies with scores ≥7 were defined as high quality [14]. The above steps were carried out by two researchers (Lilong Zhang, Qihang Yuan) independently and cross-checked, and all disagreements were dealt with by the senior authors (Weixing Wang).

Statistical analysis
All statistical analyses were performed using Stata 16.0. Continuous and dichotomous variables were compared by the standardized mean difference (SMD) and odds ratio (OR), respectively. Hazard ratio (HR) was calculated to assess correlations between AdipoQ or leptin expression and the prognosis of HCC. For studies that presented continuous data as median and range values/ interquartile, the means and standard deviations were calculated using statistical algorithms described by Luo et al. and Wan et al., respectively [16,17]. In the case of the studies only provided Kaplan-Meier survival curves, Engage Digitizer version 2.11 software was used to extract relevant numerical value from survival curves and calculate the HR (95% CI) [18,19]. All the effect quantities were represented by the 95% con dence interval (CI). P 0.05 was considered statistically signi cant. We used the chi-squared test to evaluate the statistical heterogeneity between different studies. P > 0.1 and I 2 < 50% indicated low heterogeneity where a xed-effect model was used; otherwise, the random-effect model was adopted.
Furthermore, a 2-stage dose-response meta-analysis was performed to explore the association between different categories of leptin, AdipoQ levels and HCC risk [20,21]. 1)The xed-effect nonlinear model was constructed based on the restrictive cubic spline function (Knot =3). 2)According to the results of the heterogeneity test and nonlinear correlation test, the corresponding model was adopted.
Finally, for the indicators with high heterogeneity, we carried out sensitivity analyses for identifying the source of heterogeneity and checking the robustness of the results. The leave-one-out method was employed, which allowed us to determine the implication of each study on the pooled effect size. Besides, meta-regression analysis was performed to explore the potential sources of heterogeneity. For indicators with over 10 included articles, we generated a funnel plot to inspect publication bias visually. Begg's and Egger's tests were also conducted for analyzing the publication bias quantitatively, where, P < 0.05 was regarded as statistically signi cant. We validated the results of publication bias by establishing trim and ll funnel plot if required.

Studies Retrieved and Characteristics
In this meta-analysis, we identi ed 1,068 potentially eligible records, and screened their titles and abstracts for inclusion. After detailed reading the full text of 78 records, 30 studies met our inclusion criteria (Fig. 1). Although Ebrahim's article met the research topics, it was excluded because the full text was not available [22]. Twenty-one articles  (10 case-control, 5 nested case-control, 4 cross-sectional and 2 cohort studies) evaluated the relationship between Leptin or AdipoQ levels and HCC risk, and main characteristics were summarized in Table 1. Three case-control studies [44][45][46] assessed the relationship between Leptin or AdipoQ gene polymorphism and HCC risk, and main characteristics were reported in Table 2. Six articles [47][48][49][50][51][52] analyzed the relationship between AdipoQ or leptin expression and the prognosis of HCC, and the main characteristics were summarized in Table 3. Quality assessment of the included studies using the Newcastle-Ottawa scale is shown in Table 1-3, and the scores obtained ranged from 5 to 8. Twenty articles were awarded 7 or 8 points, and considered as high-quality; Six studies were awarded 6 points and four studies were awarded 5 points, which were considered as moderate quality.
Meta-regression analysis showed that only the ethnicity (P = 0.004), not the source of control (P = 0.242) and case (P = 0.185), sample size (P = 0.735), mean age (P = 0.420), study design (P = 0.344), assay method (P = 0.606), ALT (P = 0.172) and albumin (P = 0.853), had signi cant impacts on the heterogeneity in the meta-analysis. To assess the impacts of each study on the overall meta-analysis, we carried out sensitivity analysis using the leaveone-out method. No substantial change of data on leptin levels was observed. Therefore, the results of our meta-analysis were relatively stable and credible (Fig. 4).
Funnel plot representing SMDs of the leptin levels in the HCC group compared to the CFC group was used to evaluate publication bias. Through the visual inspection of the funnel plot, there was obvious asymmetry that indicated a possibility of publication bias (Fig. 5), which were supported by Begg's tests (P = 0.034) and Egger's tests (P = 0.025). Therefore, further veri cation by trim and ll funnel plot was employed to adjust for the potential publication bias. However, the pooled data regarding leptin that had been signi cant before the adjustment with the "trim and ll" method remained signi cant after the adjustment (SMD = 3.486, 95% CI (0.937-6.035), P < 0.05), indicating that this publication bias did not affect the pooled estimates.
Khattab et al. [34] found that AdipoQ levels in HCC with the size of nodules≥5 cm were signi cantly greater than 5 cm (24.2 ± 2.1 vs 20.8 ± 3.8, P = 0.009), whereas, AdipoQ levels were not related to TNM stages, number of nodules and lymph node metastasis. Feder et al. [31] discovered that AdipoQ levels were no statistical difference in HCC and colorectal liver metastases patients, and negatively related to steatosis grade, but not correlate with in ammation or brosis score. Sadik et al. 40 reported that AdipoQ levels of cirrhotic HCC were signi cantly higher than the noncirrhotic HCC group, whereas leptin was not.
Meta-regression analysis showed that the source of control (P = 0.150) and case (P = 0.579), ethnicity (P = 0.338), sample size (P = 0.140), mean age (P = 0.540), study design (P = 0.283), assay method (P = 0.092), source of sample (P = 0.993), ALT(P = 0.544) and albumin (P = 0.575) had no signi cant effects on the heterogeneity in the meta-analysis. We also carried out sensitivity analysis using the leave-one-out method, and no substantial change of data on AdipoQ levels were observed, therefore, the results of our meta-analysis were relatively stable and credible (Fig. 9).
Funnel plot representing SMDs of the AdipoQ levels in the HCC group compared to the CFC group was used to assess publication bias. Through the visual inspection of the funnel plot, there was obvious asymmetry that indicated a possibility of publication bias (Fig. 10), which were not supported by the Begg's tests (P = 0.300) and Egger's tests (P = 0.142); therefore, further veri cation by trim and ll funnel plot was employed to adjust for the potential publication bias. The result of the "trim and ll" method revealed that no trimming was performed and the data were unchanged, suggesting that there was no signi cant publication bias.
As for AdipoQ, Cai et al. [45] found that the AdipoQ rs1501299 was associated with the increased susceptibility to HCC, and the additive model showed that the GT and GG genotypes were signi cantly associated with an increased HCC risk (GT vs TT: OR = 2.83, 95% CI (1.36, 5.89), P = 0.006; GG vs TT: OR = 4.52, 95% CI (2.25, 9.11), P = 0.001). In the dominant model analysis, the GG+GT genotypes were associated with a 3.8-fold elevated risk in HCC(GG+GT vs TT: OR = 3.795, 95% CI (1.92, 7.49), P = 0.001). However, the rs266729, rs822395, rs822396 and rs2241766 were not signi cantly associated with HCC. Unfortunately, we just retrieved one study that evaluated the association of AdipoQ gene polymorphism with HCC, so we failed to perform related meta-analysis.

Dose-response of circulating AdipoQ, leptin levels and HCC risk
Pooling data from 4 studies [24,28,35,38] with 1507 participants showed that there was a linear dose-response relationship between circulating AdipoQ levels and HCC risk ( Pnon-linearity = 0.233). We de ned the increment in 1μg/ml AdipoQ levels as a unit to show the trend more clearly. The trends were signi cant for increasing HCC risk per one unit increase of AdipoQ (OR = 1.066, 95% CI (1.03, 1.11), P = 0.001; Fig. 12), without signi cant heterogeneity (P heterogeneity = 0.338).
As for leptin, Aleksandrova et al. [24] and Chen et al. [28] both con rmed that circulating leptin levels were no signi cant dose-response trend in the development of HCC. Unfortunately, we only found these two studies, and we unable to perform a meta-analysis.
Pooling data of 3 studies [47,50,52] with 241 participants measured the association between leptin expression and prognosis for HCC. Heterogeneity analysis showed that signi cant heterogeneity was observed among the studies (I 2 = 66.0%, P = 0.053), the random-effect model was applied. The results showed that high/positive expression of leptin was not signi cantly associated with prognosis in HCC patients (HR = 0.92, 95% CI (0.53, 1.59), P = 0.766; Fig. 13).

Page 6/17
Over the past two decades, many studies have explored the correlation between abnormal expression of leptin and AdipoQ and various obesity-related cancers. In 2016, Wei et al. [53] carried out a meta-analysis of 107 articles to investigate AdipoQ levels in various malignancies and found that AdipoQ levels in some cancer cases (including acute leukemia, multiple myeloma, breast cancer, colorectal cancers, endometrial cancer, prostate cancer, thyroid cancer, tongue cancer, gastroesophageal cancer) were signi cantly lower, and in HCC was signi cantly higher than in the CFC group. However, only 7 articles regarding HCC were involved in the meta-analysis. Song et al. [54] analyzed 9 Chinese and English studies and concluded the same results. Our results indicated that HCC patients showed signi cantly higher AdipoQ levels than the healthy control group, but no signi cant difference about AdipoQ levels than CFC group. To date, no meta-analysis on leptin and HCC risk has been carried out, and our results indicated that HCC patients showed signi cantly higher leptin levels than the CFC group, healthy control group and cirrhosis group. Besides, comparing the HCC group and the different sources of the CFC group, the results were different. Thus, we can conclude that AdipoQ and leptin levels are altered in patients with chronic hepatitis and cirrhosis compared to healthy controls, which is consistent with Buechler's conclusion [13].
However, since patients of included studies were recruited from the different contexts of gender, race, other demographical parameters as well as overall health status of an individual, the high heterogeneity was observed in this pooled analysis and the ndings should be explained with caution. We conducted meta-regression, subgroup analyses, and carried out sensitivity analyses to determine the source of heterogeneity. In the pooled analysis of leptin, the results of meta-regression indicated that the heterogeneity derived from ethnicity, and subgroup analyses showed that heterogeneity was related to the source of the control group, study design, assay method and baseline levels of albumin. In the pooled analysis of AdipoQ, the source of heterogeneity was not found by the meta-regression, and heterogeneity was linked to the source of the control group, study design, mean age and sample source by the subgroup analyses.
Many single-nucleotide polymorphisms were found in the leptin gene, and the earliest one was the LEP rs7799039 polymorphism, an SNP identi ed in the 50-untranslated region of the leptin gene [55], which has been studied in various malignant tumors and was suggested it could affect the transcriptional level and leptin expression [56]. Some previous meta-analyses indicated that LEP rs7799039 polymorphism conferred the risk of cancer [57][58][59]. However, no meta-analysis was conducted to explore the correlation between LEP rs7799039 polymorphism and HCC risk. In this study, we found that LEP rs7799039 polymorphism was involved in the susceptibility to HCC. Unfortunately, the included studies are too few, and the conclusion ought to be further veri ed by more high-quality studies.
By dose-response, we are able to more clearly explore the association between AdipoQ and leptin and the risk of HCC. In 2019, Yoon et al [60] found that AdipoQ levels were signi cantly associated with decreased risk of cancer, such as breast, colorectal, and endometrial cancer, and leptin was signi cantly associated with increased risk of cancer, such as endometrial and kidney cancer. However, our results showed that the increase of AdipoQ levels were signi cant for increasing HCC risk. There were just 2 dose-response studies between leptin and HCC risk, so meta-analysis was given up.
Leptin and AdipoQ may have closely associated with not only the occurrence of cancer, but the prognosis of cancer. Our ndings that high/positive expression of AdipoQ was signi cantly correlated with lower OS in HCC patients are similar to the meta-analysis of 10 studies, which revealed that elevated AdipoQ expression was related to poor prognosis in cancer patients (including HCC patients) [61]. It is worth noting that high/positive expression of leptin was not signi cantly correlated with prognosis in HCC patients in this meta-analysis.
High levels of leptin may play an important role in the promotion of cancer cell migration, proliferation, survival and angiogenesis [62,63]. This is achieved by the activation of Janus kinase/signal transducer and activator of transcription, phosphatidylinositol 3-kinase, mitogen-activated protein kinase, and extracellular signal-regulated kinase signaling pathways [64,65], which were thought to be related to oncogenes [66,67]. Leptin can also promote the development of liver brosis, steatosis and proin ammation [4,68]. Besides, Mittenbuhler et al. revealed leptin signaling as a promoter of HCC in obesity [69].
Many studies have found that AdipoQ have signi cant anti-proliferative, anti-carcinogenic activity and anti-in ammatory [70]. Nazmy et al. found that AdipoQ's tumoricidal activity could provide more protection for the body against the HCC by hindering reduction in p53 expression, reactivation of TNFrelated apoptosis-inducing ligand signaling and induction of apoptotic pathway [71]. Al-Gayyar et al. revealed that AdipoQ could completely block the increase of sulfatase 2 induced by HCC, ameliorate the expression of tumor invasion markers, matrix metalloproteinase-9, syndecan-1 and broblast growth Factor-2 induced by HCC, decrease the expression of NF-κB and tumor necrosis factor α (TNF-α) induced by HCC, thus achieve the hepatoprotective [72]. Manieri's research showed that adipoQ can activate two proteins in liver cells, p38α and AMP-activated protein kinase, which can prevent cell proliferation and impair tumor growth [73]. Our nding that elevated AdipoQ was linked to higher risk and poor prognosis of HCC, which seems paradoxical with the above view. Some mechanisms have been suggested to account for this contradiction. 1) AdipoQ resistance: even if AdipoQ is plentifully expressed, it may not be able to prevent the poor prognosis due to the down-regulated of AdipoQ receptor or the dysfunctional of AdipoQ signaling pathway. Many HCC patients have liver cirrhosis or brosis, both of which are related to the down-regulation of AdipoQ receptors in the liver and reduction of AdipoQ clearance, leading to AdipoQ resistance status [74]. Also, the expression of AdipoQ originally enhances to compensate for the progression of HCC, but the higher levels of AdipoQ are ineffective due to the overall worsening of the patient's physical condition [51]. 2) AdipoQ stimulates AKT-mediated activation of cancer cells, which is a signi cant predictor of poor survival [51,75].
There are some limitations in this meta-analysis that should be considered. First, the partial results should be interpreted with caution due to the high level of heterogeneity. Second, few studies were conducted to explore the correlation between leptin, AdipoQ gene polymorphism and HCC risk. The results of LEP rs7799039 polymorphism and HCC risk should require more studies to con rm. Third, we extracted partly HR from the survival curve of the original article, which may introduce some small errors. Finally, almost all studies included in the meta-analysis had leptin and AdipoQ levels measured only one time and did not show its long-term changes during the course of HCC.

Conclusions
The present study shows that high leptin levels are associated with a higher risk of HCC, which may represent a useful biomarker for early detection of HCC. AdipoQ levels are proportional to HCC risk, with a linear dose-response relationship, and are related to the poor prognosis, which may be a useful biomarker for estimating the prognosis of HCC.    TT=12; TG=60; GG=128; G  allele=316; T allele =84   TT=39; TG=69; GG=92; G  allele=253; T allele =147 8 C = Case-control, SNP = single-nucleotide polymorphisms, PCR-RFLP = polymerase chain reaction-restriction fragment length polymorphism, NOS = Newcastle-Ottawa Scale         Sensitivity analysis of comparing circulating leptin levels between the HCC and cancer-free control group.        Sensitivity analysis of comparing circulating adiponectin levels between the HCC and cancer-free control group.

Figure 10
Funnel plot of comparing circulating adiponectin levels between the HCC and cancer-free control group.

Figure 11
Forest plot for the association between leptin rs7799039 and HCC risk Figure 12 Dose-response associations of circulating adiponectin levels and HCC risk.

Figure 13
Forest plot of the relationship between adiponectin, leptin expression and survival in HCC.