This systematic review and meta-analysis was conducted ​in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) guidelines.

Two investigators independently performed a systematic literature search across three electronic databases (Embase, PubMed, and Web of Science) for articles published between January 2000 and May 2023. The search strategy combined Medical Subject Headings (MeSH) terms and free-text keywords, including:

Boolean operators (AND/OR) were utilized to link conceptual groups. The reference lists of eligible articles were manually screened to identify additional relevant studies.

Studies were selected based on the following criteria:

Inclusion Criteria:

Exclusion Criteria:

Discrepancies between reviewers were resolved through consensus discussions.

Two researchers independently extracted the following data using a standardized form:

Missing statistical parameters were calculated using RevMan 5.3 (Cochrane Collaboration). Corresponding authors were contacted via email to request unreported data; unresponsive inquiries were documented.

Study quality was evaluated using the Newcastle-Ottawa Scale (NOS) for observational studies, assessing three domains:

Each domain was scored as “low,” “moderate,” or “high” risk of bias. Randomized controlled trials (RCTs), if present, would have been assessed via Cochrane Risk of Bias Tool 2.0 (RoB 2).

The methodological quality of the included cohort and case-control studies was assessed using the Newcastle-Ottawa Scale (NOS). ​As recommended by the MOOSE guidelines,​​ studies with a NOS score ≥ 7 were considered high quality.

Conducting the meta-analysis involved utilizing the RevMan 5.3 software, with all study flexibles represented as binary variables indicated by relative risk and corresponding 95% confidence intervals (11). To address potential heterogeneity among studies, the calculation of the comprehensive impact magnitude was executed using the random-effects framework. Cochran’s Q and I² statistics were employed to evaluate heterogeneity. In instances where the P>0.1, signifying homogeneity across multiple studies, the fixed-effects model was implemented (12). Conversely, if P ≤ 0.1, suggesting statistically significant heterogeneity, the random-effects model was utilized. Moreover, I² > 50% indicated substantial heterogeneity, prompting further evaluation of result stability through sensitivity analysis (13).

Result stability was assessed by removing the study with the maximum weight and observing the resulting change in the effect quantity.

a funnel diagram will be employed when there are at least 6 articles included, in order to examine the presence of publication bias.

The substance of this article does not necessitate moral sanction or ethical examination, and its dissemination will occur through printed materials or relevant meetings (14).

After eliminating duplicates, a search across Web of Science, PubMed and Embase yielded 3497 records. Further full-text review was required for 39 articles initially identified, resulting in the exclusion of 23 articles. Ultimately, 16 prospective, retrospective, and cross-sectional studies—encompassing both healthy individuals and patients with liver-related diseases—were encompassed in the present meta-analysis ( Figure 1 , Table 1 ). Among the 16 included studies, 13 (81.3%) excluded participants with pre-existing cancer diagnoses, while 7 (43.8%) further excluded incident HCC cases occurring within 1 year of lipid measurement to address reverse causality concerns ( Table 1 ) (15–30).

Through our examination, a statistically significant connection emerged, demonstrating the association between the risk of liver cancer and the concentrations of total cholesterol in the bloodstream (HR = 0.71, 95% CI: 0.64, 0.78; I² = 0%) ( Figure 2A ). Delving into cholesterol subtypes, serum HDL did not exhibit a significant effect on liver cancer occurrence. Meanwhile, serum LDL appeared to be more intricately linked to liver cancer risk than serum HDL (HR HDL = 1.05, 95% CI: 0.87, 1.26; I² = 83%; HR LDL = 0.46, 95%CI: 0.36, 0.59; I2 = 97%) ( Figures 2B and 2C). Additionally, we observed that serum triglycerides (HR = 0.79, 95%CI: 0.62, 0.99; I² = 94%) ( Figure 2D ) and dyslipidemia (HR = 0.64, 95%CI: 0.50, 0.83; I² = 81%) ( Figure 2E ) also showcased a declining correlation pattern concerning the risk of liver cancer.

Using RevMan software, the evaluation of the 16 papers incorporated in this study was conducted by our team to assess their quality ( Figures 3 ). The majority of these studies demonstrate medium to high quality, indicating their reliability. However, six studies exhibit susceptibility to attrition bias due to incomplete data, while four papers are prone to selective bias, lacking explanations regarding the potential predictability of distribution results by subjects and researchers.

To assess potential publication bias within our analysis, we employed the Deeks funnel chart asymmetry test, plotting the effect value-to-standard error ratio for each study. The funnel plot indicated slight to moderate asymmetry among certain variables. Utilizing Egger regression asymmetry and Begg rank correlation tests, we observed slight asymmetry in studies related to TC, albeit this finding lacked confirmation through the Begg rank correlation test ( Supplementary Figure 1 ).

To gauge the strength and reliability of our primary findings, sensitivity analyses were undertaken as part of our research methodology, systematically excluding individual studies from the meta-analysis. For serum cholesterol and low-density lipoprotein meta-results, we applied a leave-one-out approach, revealing that the outcomes remained generally stable and robust upon exclusion of each study one at a time. However, the results concerning serum triglycerides were influenced by a single study, and upon its removal from the analysis, a change in the results was observed. Specifically, it was observed in our analysis that initially, elevated levels of serum triglycerides were linked to a protective effect against the occurrence of liver cancer, but this association became statistically insignificant after excluding the study by Yuri et al. (HR = 0.69, 95% CI: 0.45, 1.04, I2 = 94%) ( Supplementary Figure 2 ).

In the course of our research, we performed a thorough analysis that amalgamated evidence pertaining to the risk of liver cancer and serum lipid levels. Through the meta-analysis, an adverse connection between increased blood cholesterol concentrations and the likelihood of hepatic carcinoma was unveiled. Moreover, serum low-density lipoprotein, elevated triglycerides, and dyslipidemia exhibited associations with reduced liver cancer risk, albeit with considerable variability between studies. Conversely, no significant correlation emerged between serum HDL and liver cancer risk. Except for serum triglyceride, other findings remained generally stable and consistent throughout sensitivity analyses.

Lipids, comprising diverse biomolecules, serve critical functions in cellular energy storage, structural integrity, and signal transduction. In clinical contexts, plasma lipids hold extensive associations with various diseases, prompting routine evaluations (31). Of particular concern is the pivotal role lipids play in tumor initiation and progression (32). Rapidly dividing cancer cells rely on a continuous supply of lipids for cell membrane formation and protein modification. Conversely, slowly proliferating cancer cells augment lipid content to fortify signal transduction pathways and evade apoptosis (32). Prior research indicates that cancer patients frequently exhibit ​low serum levels of total cholesterol, LDL, and HDL, with compelling evidence linking ​this hypolipidemic profile​ to cancer pathogenesis and progression (33).

Across various cancer types, including liver, hematologic, intestinal, lung, prostate, head, and neck cancers, low serum cholesterol levels have been frequently observed. Additionally, studies conducted on cohorts indicate a connection between diminished serum cholesterol levels and the overall risk of developing cancer (34). In our investigation, we propose a negative correlation between the probability of hepatic carcinoma and the concentrations of blood cholesterol, prompting the need for further investigations into its specific mechanisms. Moreover, extensive research on different lipoprotein subfractions—such as LDL, HDL, and VLDL—has consistently shown reduced serum cholesterol, HDL, and LDL levels alongside increased serum triglyceride levels, a prevalent and consistent trend warranting attention in understanding cancer risk factors (31).

Our findings suggest that elevated serum levels of LDL serve as protective factors against the risk of liver cancer. Contrarily, several studies have indicated that low levels of serum LDL may contribute to the development of liver cancer—this is further supported by observations where prostate cancer cell growth rates surged in LDL-rich mediums (35). The uptick in LDL uptake by cancer cells might hinder their clearance from circulation, potentially contributing to lowered serum LDL in cancer patients (36). The decrease in serum LDL among cancer patients also points to an intriguing inference: lipid metabolism, influencing cell apoptosis through membrane regulation and enzyme activation (37), particularly relates to specific programmed cell death mechanisms like iron death (38), facilitated by iron-dependent membrane lipid peroxidation (39). Replacing saturated fat (SFA) with unsaturated fat in dietary improvement has shown promise in reducing LDL cholesterol and elevating HDL cholesterol, enhancing overall lipid profiles (40). However, monounsaturated fatty acids (MUFA) within unsaturated fats don’t initiate lipid peroxidation, unlike polyunsaturated fatty acids (PUFA), which promote this process (41). Melanoma cells spreading via the lymphatic system show resistance to iron death, attributed to their uptake of MUFA from the lymphatic environment into their membranes (42). This suggests a potential link between excessive MUFA intake before liver cancer onset and the prevention of lipid peroxidation-induced iron death in cancer cells, potentially promoting liver cancer occurrence. Despite serum LDL level declines, the precise mechanism remains unclear, yet the accumulation of monounsaturated lipids appears to shield liver cancer cells from lipid peroxidation. Moreover, scant contradictory studies indicate an increased cancer risk linked to high-density lipoprotein levels, aligning with our research findings (43).

Examining serum triglycerides, an Italian comparative study on blood lipid and cancer presented conflicting results regarding serum TG levels across different tumor types. However, overall, cancer patients tended to exhibit relatively high blood lipid characteristics (44). Our meta-analysis findings denote serum TG as a protective factor against liver cancer risk. While the limited and contradictory literature included demonstrates an overarching negative correlation trend, the specific mechanism warrants further exploration. Shifting our focus to dyslipidemia and its connection with the risk of developing liver cancer, current guidelines assert a significant association is observed between nonalcoholic liver disease (NAFLD) and metabolic risk factors, particularly emphasizing obesity, type 2 diabetes (T2DM), and dyslipidemia (45). Mounting evidence links NAFLD to an increased risk of HCC, suggesting a potential role for dyslipidemia in promoting liver cancer risk (46). However, our research surprisingly reveals dyslipidemia not as a promoter but as a protective factor against liver cancer—an unexpected outcome. Upon revisiting the literature, inadequate inclusion volume and varying dyslipidemia definitions emerged as key reasons for the inconsistent data, including differing baseline levels and diverse exclusion criteria in the included studies. Hence, a more in-depth and thorough examination and analysis are required to interpret the results related to dyslipidemia.

The interplay of various environmental and genetic factors profoundly shapes human plasma lipid profiles (47), underscoring the importance of discerning the biological relevance between distinct lipid components and liver cancer. Consideration of confounding factors affecting serum lipid profiles is pivotal in preventing overlooked risk factors from confounding analyses and yielding perplexing outcomes. Our findings introduce novel evidence for liver cancer, pinpointing a correlation between high cholesterol content (≥200mg/dL) and decreased liver cancer risk. Yet, the limited number of studies warrants cautious interpretation of this result. First of all, Inconsistencies in the relationship between cholesterol and liver cancer risk may stem from varied lipid acquisition modes adopted in different cohort studies, influencing plasma lipid profiles. Reports of diverse plasma lipid profiles among liver cancer patients hint at potential metabolic strategies adopted by cancer cells, necessitating further investigation into plasma lipid’s mechanisms and significance in liver cancer progression. The connection between serum lipids and liver cancer involves a role played by genetic susceptibility. Moreover, undiagnosed liver conditions like cirrhosis and nonalcoholic fatty liver disease (NAFLD) could impact cholesterol metabolism, potentially intensifying the negative correlation with the risk of liver cancer. For example, an association is noted between persistent hepatitis B virus infection and an elevated likelihood of hepatic carcinoma alongside reduced cholesterol levels, blurring the qualitative distinction between blood cholesterol levels and the risk of liver cancer. In this scenario, elevated cholesterol concentrations may indicate facets of hepatic function rather than serving as a direct pathogenic element in the likelihood of hepatic carcinoma (48). Then, Of the 16 studies incorporated in our meta-analysis, excluding only 9 participants based on the presence of cancer or cardiovascular disease restricts our capability to thoroughly assess the influence of undiagnosed liver diseases on the outcomes we have obtained. Moreover, our outcomes might be influenced by the heightened use of cholesterol-lowering medications, specifically statins, among individuals with elevated cholesterol levels (49). The utilization of statins has been linked to a decreased likelihood of developing liver cancer (50), yet only 2 studies among those we reviewed excluded statin users. The majority of inquiries into the relationship between serum lipids and liver cancer do not incorporate adjustments for the utilization of statins. Although the utilization of statins was uncommon prior to 1990, the available baseline data from that timeframe were restricted. Regardless, the outcomes indicated a reciprocal connection between the concentrations of blood cholesterol and the likelihood of hepatic carcinoma (ORH/L=0.60, 95% CI: 0.46, 0.78, n=3, I² = 81.2%). Clearly, further data on statin use is warranted to comprehensively assess their influence on our results.

Our meta-analysis offers fresh insights into the relationship between serum cholesterol, HDL, LDL, TG, dyslipidemia, and the risk of liver cancer. Surprisingly, our results unveiled an unforeseen opposite correlation between the concentrations of blood cholesterol and the likelihood of hepatic carcinoma, prompting the necessity for additional exploration to clarify this unanticipated association. However, the limited research available on other lipid subtypes like HDL and LDL underscores the need for more comprehensive studies in these areas.

Our research offers notable advantages. Firstly, we conducted category and subgroup analyses, enhancing the depth of our investigation. Secondly, in the majority of serum lipid studies (15 out of 16), at the initiation of the cohort, blood specimens were gathered and subsequently scrutinized for a period exceeding 5 years. This approach potentially mitigates concerns about reverse causality, strengthening the reliability of our findings. Most importantly, in the latest published analysis of the association between lipids and liver cancer risk (10), we obtained different results using different information sources. We found that LDL was negatively correlated with liver cancer risk, and we made a more detailed explanation for this result, which was not reflected in the latest published literature.

Some noteworthy limitations should be considered. Firstly, our inclusion of both prospective and retrospective studies might affect the internal validity of our findings. Secondly, the relatively limited number of studies may hamper statistical power, especially in detecting associations and potential publication bias—highlighted in instances such as low-density lipoprotein and dyslipidemia, which involved only a few retrieved studies. Thirdly, substantial inter-study heterogeneity in the relationship between serum lipids and liver cancer was observed, where, given the restricted quantity of studies available, a comprehensive discussion on the sources of heterogeneity for certain serum lipids was not undertaken. Fourth, our analysis is confined by the original research’s limitations. For instance, dietary information gathered primarily through Food Frequency Questionnaires (FFQ) might be prone to measurement errors. Fifth, the bulk of inquiries lacked details on hepatitis B virus (HBV) and hepatitis C virus (HCV), constraining our ability to assess their influence on the outcomes, although it is probable that their impact on the recorded connection between dietary habits and hepatic carcinoma is negligible (51). Finally, in our ongoing meta-analysis, 7 out of 9 studies concentrating on hepatic conditions regarded hepatic carcinoma as their principal result, with 1 specifically focusing on liver cancer risk. Additionally, among the 7 serum lipid studies, 4 prioritized liver cancer as the primary outcome, while 2 concentrated on hepatocellular carcinoma risk. Given statins’ estimated 15-30% HCC risk reduction and 20-40% LDL-lowering effect, we calculate that unadjusted statin use could theoretically attenuate true LDL-HCC associations by ≤12% (using Bross formula). This does not invalidate our primary finding. Consequently, our findings predominantly pertain to HCC, and we faced limitations in thoroughly evaluating the potential etiological heterogeneity among various subtypes of HCC.