In this population-based study, we assessed temporal trends in pediatric hepatoblastoma incidence and constructed a nomogram to provide an individualized prediction of OS. The current analysis showed a significant annual increase (2.2%) in hepatoblastoma incidence between 2000 and 2015. Although management of this disease has evolved over the past 30 years, there is no model for predicting OS of hepatoblastoma patients. The constructed nomogram was found to be accurate in prognostic prediction for hepatoblastoma patients, with a C-index of 0.79.
Owing to the rarity of hepatoblastoma, single-institution data were insufficient for assessing its incidence trends. Until now, only a few population-based studies have reported hepatoblastoma incidence trends in the past few decades. Darbari et al. [1] reported the incidence of primary liver malignancies among U.S. children aged less than 20 years from 1973 to 1997, using the SEER database. They found that hepatoblastoma incidence had increased, with age-adjusted incidences of 1.09 per 1,000,000 for the whole cohort, 1.22 per 1,000,000 for males, and 0.96 per 1,000,000 for females. However, clinical outcomes and prognostic factors were not reported. In another SEER study of 606 patients diagnosed with hepatoblastoma between 1973 and 2009, Allan et al. [8] found a significant increase in the overall age-adjusted incidence of hepatoblastoma with an APC of 2.18%. However, the reported incidence was not stratified by sex and age. Although outcomes and prognostic factors were reported, the study did not provide a new prognostic model. A recent population-based study analyzing the Taiwan Cancer Registry (TCR) database reported that, from 1995 to 2012, the overall incidence of hepatoblastoma in children increased by 7.4% per year and, specifically, by 6.5% among males [16]. In that study, the incidence of hepatoblastoma was analyzed according to sex and age. However, no outcomes were reported. In the present study, we have provided updated information on hepatoblastoma incidence trends. We found that its incidence increased by 2.2% annually from 2000 to 2015 and, notably, that the incidence increased most significantly among males and in 2- to 4-year-old children.
In the past decades, the management of hepatoblastoma has evolved to include not only surgical treatments such as liver resection and transplantation but also non-surgical treatments such as chemotherapy. Progress in treatment has contributed to prolonged OS in patients with hepatoblastoma. The 10-year OS rate was 81.0% in our analysis, which was better than 61% reported in a SEER-based analysis from 1973 to 2009 [8]. In a single-center retrospective study involving 30 patients younger than 18 years who underwent liver transplantation for hepatoblastoma between 1997 and 2014 at Stanford University School of Medicine, Pham et al. [11] reported 10-year disease-free survival and OS rates of 82% and 84%, respectively. In another SEER study including 318 hepatoblastoma patients undergoing surgery between 1998 and 2009, McAteer et al. [27] found that the overall 5-year disease-specific survival rate was 85.7%, and the rates for patients undergoing resection and transplantation were essentially equivalent (85.6% vs. 86.5%, P = 0.66). In the present study, patients who underwent liver resection (n = 341) and liver transplantation (n = 84) had comparable 10-year OS rates (89.3% vs. 90.1%, P = 0.891).
Surgery, ethnicity, and age were identified as independent prognostic factors in the present study. Surgery was a strong predictor of OS. Moreover, the need for liver resection has increased over time: the number of patients receiving liver resection for hepatoblastoma increased by 1.4 folds between the time frames of 2004–2007 and 2012–2015. However, this increase was not observed in patients that received liver transplantation. The number of patients who underwent liver transplantation remained relatively constant between 2004–2007 and 2012–2015. These data emphasize that more children now have the opportunity for a possible cure by surgical resection. With advances in surgical management and chemotherapy, the 5-year OS rate for children with hepatoblastoma has increased to more than 80%. In the current study, we included patients diagnosed between 2004 and 2015 that had been reported in SEER. There is no doubt that these patients mainly benefited from efficient surgical management and the use of novel chemotherapy regimens in the modern era.
It has been reported that age of < 5 years old was a favorable factor for survival [8]. In the present analysis, the 0- to 1-year-old group had the longest OS, whereas the 2- to 4-year-old group had the shortest OS. African–Americans also had shorter OS compared with Caucasians and other ethnicities. This finding was consistent with previous studies [14, 28]. It can be speculated that patients of African–American ethnicity may have had less access to surgical care. Further investigation is necessary to determine if a variation in tumor biology or another etiology is driving this trend. Further surrogate measures, such as poverty/household income, geographic origin of the patient, proximity to surgical centers/tertiary hospitals, or perhaps similarities with the epidemiology of other types of embryonal cancers, can be analyzed in other longitudinal databases with a wider range of epidemiological data as well.
In the present study, distant disease at presentation was not found to be an independent factor for survival. This finding differed from the results of previous studies [8, 14], where it was shown to be a significant prognostic factor. The reason remains unclear, but this may be related to the increased use of chemotherapy, to which hepatoblastoma is highly sensitive [11, 29]. The fact that AFP was not an independent prognostic factor stands out when taking into account existing literature, but this can be attributed to the database effect and also the different use of AFP or the sporadic testing. Further longitudinal studies are needed to determine whether AFP can be used in predicting the prognosis of hepatoblastoma.
In the present study, we constructed a nomogram based on the results of multivariate analysis and included common clinical factors in the model. Our nomogram performed well in predicting OS, and its prediction was supported by the bootstrap-corrected C-index and calibration curve. However, this study contained some limitations. First, information on chemotherapy regimen used was not included, due to incomplete data and biases associated with unstated reasons for receiving or not receiving chemotherapy in the SEER database. Second, although our data were collected from 18 cancer registries of the SEER database, our sample size was still relatively small. The SEER program is a high-quality, population-based cancer registry. Case finding audits of high-volume facilities are routinely performed for data accuracy, completeness, and reporting timeliness. Each SEER registry is given an annual Data Quality Profile, which assesses the extent to which each registry provides data meeting contractual standards. Thus, the SEER registry robustly captures the first course of any cancer-directed surgery. However, given that chemotherapy is routinely used in clinical work, our results represent current treatment outcomes for hepatoblastoma. Third, we could only internally validate the nomogram [30, 31] because of the rarity of hepatoblastoma. Although the nomogram showed good discriminatory ability and the calibration curves were consistent, an external validation of the model is still necessary. Lastly, as a retrospective review, our study is inherently prone to selection bias.