This is a real-world data study; our primary objective was to evaluate the risk of HCC in patients with indeterminate liver nodules. Our findings revealed that 16% of the patients developed HCC during the study period, with cumulative incidence rates at 1, 3, 5, and 7 years of 3.7%, 10%, 20%, and 30%, respectively. Comparing our results with previous studies that evaluated the cumulative incidence of HCC in cirrhotic patients, we observed that our cohort exhibited a higher cumulative incidence of HCC over the years. For instance, a Swedish nationwide retrospective study found a cumulative incidence of HCC in cirrhosis of 8.3% at 5 years [17]. Ioannou et al. found a 4.7% HCC incidence over 3.6 years [9], while Paranaguá Vezozzo DC et al. reported a 14.3% cumulative incidence over 5 years [18]. We also compared our results with studies that assessed the risk of HCC transformation from preexisting indeterminate nodules. Cococcia et al., in a retrospective study of 109 patients with indeterminate nodules, found that HCC developed in one-fifth of the cases over 4.6 years, which closely aligns with our own [14]. Another study indicated that the 3-year rate of nodule transformation to HCC was 15.7% [15]. Kobayashi et al. reported cumulative HCC development rates of 3.3%, 9.7%, and 12.4% in the first, third, and fifth years, respectively, for regenerative nodules [19]. Beal et al. followed patients with indeterminate lesions and reported an HCC incidence of 21% in 4 years, consistent with our findings [20]. Although there is some variation in HCC incidence among these studies, the overall trend indicates a progressive increase in HCC incidence over time, particularly during the five-year period, with a more significant rise in patients with these nodules. Several theories have been proposed to explain the rising incidence of hepatocellular carcinoma (HCC) in the context of liver cirrhosis over time. These include the potential influence of chronic hepatic inflammation, promoted hepatocyte DNA synthesis, and the accrual of genomic alterations over time [21].

In our cohort, nearly 88% of HCC cases were attributed to HCV-related cirrhosis, consistent with findings from other studies [14, 15, 22]. The increased incidence of HCC in individuals with HCV may stem from the direct carcinogenic effects of specific HCV viral proteins involved in various oncogenic processes. Additionally, differences in the pathogenesis of necroinflammation between HCV and other etiologies, such as MASLD and alcohol-related liver disease, may lead to varying risks of HCC development [14, 22]. Our study revealed a higher incidence of HCC in males compared to females, likely influenced by demographic, behavioral, and environmental factors, consistent with findings from other epidemiological studies [14, 23]. In our univariate analysis, we found associations between smoking, alcohol intake, and male gender as predictors of HCC, consistent with previous studies [14, 23, 24]. However, these associations did not remain significant in the multi-regression model. Additionally, our multivariate regression analysis revealed that nodule size, especially nodules larger than 1-2 cm and LI-RADS-3 nodules, were independently associated with HCC, the same findings seen in previous studies. For instance, Khalili et al. reported a malignancy rate of 14%-23% for indeterminate 1–2-cm nodules[16].Similarly, Cococcia et al. found a higher incidence of HCC in patients with lesions ≥ 1 cm. However, the author did not find a significant association between lesion size and HCC in regression analysis, possibly due to the small sample size [14].

It is believed that the higher risk of hepatocellular carcinoma (HCC) associated with larger nodular size (especially > 1 cm) may be attributed to a potentially higher growth rate and a shorter tumor volume-doubling time, which are linked to the progression of nodules to hypervascular HCC [25]. In a meta-analysis of 12 studies evaluating LI-RADS-3 lesions, the risk of developing HCC ranged from 15.8% to 44.4% [26]. The presence of hyperintense T2 lesions may potentially indicate the development of HCC. However, our multi-regression model analysis did not reveal any significant correlations. These results are consistent with those reported by Rimola et al. [27].

Our research found that Child–Pugh grade, MELD score, and platelet count level were not predictors for HCC transformation despite their previously established correlation in studies [14, 15]. Our analysis indicated that the majority of cases at baseline and at HCC time were classified as Child–Pugh grade A or B rather than C, highlighting the limitations of relying on Child–Pugh grade for HCC prediction. These findings align with a retrospective study of HCC patients, where 78.7% of the patients were identified as Child–Pugh A or B [28]. During the follow-up period, two cases exhibited nodule disappearance. Additionally, there were five mortality cases during the study period, all of which were in the non-HCC group, with three attributed to complications related to liver disease. In the subgroup exploratory analysis for the HCC group, the study compared the characteristics between patients with denovo HCC and those with nodule transformation HCC. It's important to note that the study was purely descriptive and did not conduct any inferential analysis. Therefore, the findings may not be generalizable. However, the primary aim was to identify distinct features between both groups, as detailed in the results section. It's worth mentioning that denovo HCC occurred in 32% of all HCC cases, aligning with findings from Arvind et al. These findings indicate the possibility of inflammatory and covertly carcinogenic conditions that could increase the vulnerability of the entire liver to the development of HCC [29].

In previous studies on indeterminate nodules, denovo cases were typically excluded as the focus was solely on the nodules' outcome. In our study, however, we sought to assess the overall incidence of HCC in these populations. This approach makes our results more applicable, as patients and clinicians are concerned with the overall incidence of HCC, irrespective of whether the tumor originated from previous nodules. Our study has several strengths. These include a relatively large sample size, precise criteria for nodule diagnoses, the use of multiple MRI keywords during patient search to reduce selection bias, a lengthy follow-up period (with the majority of cases being followed up for 3–5 years and almost 25% observed for up to 7.5 years), a diverse patient population in terms of ethnicity at our center, and comprehensive radiological descriptions of the nodules provided by a specialized radiologist. However, the study has some limitations, mainly due to its retrospective design, which introduces potential biases such as observer bias and hidden unmeasured confounders. The data collection process relied on records, occasionally leading to incomplete information. As the field of artificial intelligence (AI) progresses in the realm of medicine, it is imperative that further research be conducted using deep learning AI and incorporating multiomics, such as radiomics and genomics. This will provide a more comprehensive understanding of the disease processes and aid in predicting the risk of HCC transformation from indeterminate liver nodules. Controlled prospective studies should be conducted over an extended period to establish the risk of HCC in patients with indeterminate liver nodules. Additionally, these studies can explore the potential of a carcinogenic microenvironment that may elevate the risk of metachronous HCC outside the nodules. By embracing a comprehensive approach, a more accurate risk assessment and an improved understanding of potential relationships can be achieved.