Included all children admitted to the unit as well as those diagnosed with HB within the study period.

Missing medical records and children diagnosed outside the study period.

Ethics approval was granted by the Health Sciences Research Ethics Committee (HSREC) at the University of the Free State (Ethics number: UFS – HSD 2023/0217/2507), and permission to conduct the study was also obtained from the Free State Department of Health. Consent from participants was not required as it was a retrospective study. There was no contact between the researchers and the patients or their caregivers.

A data collection form was created on Redcap® with password protection and was used for data collection. Each participant was assigned a unique numerical identifier, thereby anonymity was maintained. No personal information was revealed to the public.

Twenty patients met the inclusion criteria. We audited the patient files and electronic records, and data were collected to include the following demographics: (1) age and sex, (2) constitutional genetic abnormalities, (3) malformations, (4) clinical presentation, (5) investigations done including full blood count, (6) platelet count, (7) AFP and liver function test, (8) histology and staging, and (9) management and outcomes namely survival and mortality. Management included medical treatment in the form of chemotherapy and surgical interventions which included surgical resection of the tumour. A pilot study was done with five participants to test the data collection tool, and some modifications were made like the exclusion of neonatal risk factors and maternal risk factors, such as smoking as there was scanty data on that. The results obtained from the pilot study were included in the final analysis. The collected data were anonymised before sending them to the biostatistician for statistical analysis.

Descriptive statistics, namely median and percentiles, as the distribution of numerical data was skewed, and frequencies and percentages for categorical data were calculated. The incidence of HB was calculated and described using 95% confidence interval (CI) for the incidence. Associations between prognostic indicators and mortality were calculated and described using Fisher’s exact test for sparse categorical data. The analysis was done by the Department of Biostatistics, University of the Free State (UFS). The data analysis for this study was generated using SAS software. Copyright, SAS Institute Inc. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. (Cary, North Carolina, United States [US]). A p-value of < 0.05 was defined as significant.

The number of children diagnosed with HB was 25 patients of the total 2076 children admitted with cancer within the study period, with an incidence of 1.2% and a 95% CI of 0.8% – 1.8%.

The majority of the patients, 60% (n = 12), were referred from within the Free State province, 30% (n = 6) were from the Northern Cape province and 10% (n = 2) from Lesotho.

The median time from onset of symptoms to presentation at UAH was 2.7 weeks with interquartile range (IQR) of 1.3 weeks – 8 weeks, and the longest time noted before presentation was after 10 months of symptoms. Other clinical characteristics, investigations and results are presented in Table 1.

In some children, an initial abdominal sonar was done to aid diagnosis. All the children had a computed tomography (CT) scan documented which was the most utilised radiological modality for staging. There was overlap with radiological investigations with most children having at least two modalities.

Biopsy was done in most of the patients and five did not have a documented biopsy. Rationale for non-performance of biopsy was not stated in the medical records (Table 1).

Histology showed the majority, 33% (n = 5), to be foetal and embryonal type, followed by mixed epithelial 27% (n = 4). Purely foetal was 13% (n = 2), and purely embryonal was 7% (n = 1). Thirteen per cent (n =2) showed foetal and mixed epithelial histology, and 7% (n = 1) showed macrotrabecular pattern.

Majority (n = 12; 60%) of the patients (with PRETEXT IV or metastasis) were classified as high risk at diagnosis and 35% (n = 7) as standard risk (PRETEXT I, II, III and no metastasis) (Table 2).

One child did not receive chemotherapy because the patient died 4 days after diagnosis before commencement of chemotherapy. For relapses and refractory HB, 5-fluorouracil and irinotecan were used. The number of chemotherapy courses given was a median of 6.00 with an IQR of 2.5 courses to 8 courses. Eight children had more than six courses of chemotherapy due to unfavourable response or disease progression. Surgical resection was possible in only 50% (n = 10) of the patients, and most patients were too unstable for surgery with advanced PRETEXT staging and high-risk stratification (Table 2).

Of the 20 children, 35% (n = 7) achieved remission, six (85%) of which were classified as standard risk. Unfortunately, 42% (n = 3) of them relapsed. Of the three patients who relapsed, the period to relapse was after 5 months in one patient and 12 months in the other two patients. In contrast, for the high-risk patients, which were 60% (n = 12), only one achieved remission and the majority, 91% (n = 11) had disease progression with poor or no response to chemotherapy (Figure 1).

At the end of the study period, 80% (n = 16) of the children had been documented to have died. After diagnosis, the longest survival duration was 3 years and 7 months and the shortest was 4 days (IQR of 0.22 months–1.02 years). Two patients (10%) had abandoned treatment. Both abandoned treatment after eight courses of chemotherapy at 13 months and 14 months post starting treatment, respectively. The remaining two patients (10%) were documented to be alive and still being reviewed.

For all the children, 1 year overall survival (OS) was 55% (n = 11), 2 year OS was 20% (n = 4), 2 year event free survival was 10% (n = 2), 5 year OS was 10% (n = 2) and 5 year event free survival was 10% (n = 2) (Table 2).

For the standard risk patients, 1 year OS was 85% as compared to the high-risk patients which was 33% (p = 0.147); 2 year OS was 42% for standard risk patients and 8% for high-risk patients (p = 0.117); and 5 year OS was 14% for standard risk patients and 8% for high-risk patients (p = 1.000) (Figure 2).

For those patients who received surgical resection, 1 year OS survival was 70% as compared to 40% for those who did not (p = 0.069); 2 year OS was 30% for surgical resection and 10% without surgical resection (p = 0.582) and 5 year OS was 20% for surgical resection and 0% without surgical resection (p = 0.473) (Figure 2).

Hepatoblastoma is a malignancy that has been documented to have a poorer prognosis in LMICs where presentation is often late with advanced disease and resources to effectively treat the disease are limited.¹⁷ In our study, the majority of the children presented with advanced stages of the disease with high-risk classification implying a delay in presentation and diagnosis.

The incidence of HB in this study was 1.2% (25 cases of HB out of 2076 admissions). This is lower than that observed in the literature. In a large cohort study published in Egypt in 2021, the incidence was noted to be 1.4% (208 children with HB out of 14,808 admissions with cancer).¹⁷

A higher median age of 24 months was noted in this study in contrast to other studies. Median age at presentation was 14 months in Turkey⁸ and 12 months in the Nashville US, India and Cairo studies.^17,18,19 In a study conducted in Red Cross War Memorial Children’s Hospital, in 2014, the median age was noted to be 12.8 months.²⁰ Age at presentation younger than 3 years old has been linked with better prognosis in the literature^21,22 which was like our findings in this study. A larger and more robust study will be needed to confirm this finding.

We also noted HB to predominate in male patients with a ratio of 1.9:1, which is consistent with the literature.^1,22,23 We, however, could not find evidence of higher mortality among male patients in the literature. In a study by Bassan J Allan et al in 2013, male patients were found to have higher survival rates.²⁴ More studies are needed to confirm this finding.

A minority of the children in our study, 20%, were born premature. This is not the case in high-income countries where prematurity and low birth weight have been found to increase the incidence of HB.^25,26 Of the four premature cases, one had a congenital brain malformation (absent corpus callosum) and presented with PRETEXT I and later developed acute lymphoblastic leukaemia (ALL); the other three children presented with advanced disease. Surveillance for intra-abdominal malignancies is important in premature babies.

Two children in this study had congenital abnormalities: absent corpus callosum described above and the other was diagnosed with Simpson–Golabi–Behmel syndrome which is known in the literature to increase the risk of developing HB.²⁷ The child with Simpson–Golabi–Behmel syndrome was male and aged 2 years and 8 months at diagnosis. The AFP levels were high at 582,050 and he presented with advanced stage, that is, PRETEXT IV + lung metastasis + invasion of the portal vein and inferior vena cava. He did not get surgical resection but had a tru-cut biopsy which showed macrotrabecular histological subtype. He received PLADO for 6 X cycles and demised after 1 year of diagnosis. This highlights the need for surveillance of HB and other malignancies in these subsets of patients.

Undernutrition is still a burden in LMICs¹⁶, and a third of our patients (n = 7) were noted to be under-nourished. Most of them (57%) presented with advanced stage of disease with PRETEXT IV and did not get surgical resection with 100% mortality.

We observed that all the children in our study presented with abdominal distension which is noted to be the most common clinical feature in the literature.⁷ This underscores the need to take abdominal distension in young children very seriously especially in the clinics and primary health care facilities.

Although AFP levels less than 100 ng/mL have been identified as a poor prognostic feature in the literature⁷, all the children in this study had an elevated AFP level of more than 100 ng/mL but OS was still lower than in the literature. A major confounding factor was likely the high-risk staging at presentation in the majority.

Most of the children had thrombocytosis (85%; n = 17) and presented with advanced stage of disease. In a study conducted by Archana et al in 2019 in India, thrombocytosis was also noted in the overwhelming majority, 90% of the patients.²⁸ Thrombocytosis has been associated with poor prognosis.²² However, in our study, two out of the three patients with platelets of less than 450 cells/mcl presented with PRETEXT IV + lung metastasis and had poor outcomes as well. More studies are needed to establish an association between thrombocytosis and poorer outcomes.

Distant metastasis at the time of diagnosis has been described as a negative prognostic indicator.²⁹ In our study, lung metastasis was the most common annotation factor, which is consistent with the literature as the commonest site of metastasis.^30,31 In a study by Tian Zhi et al, lung metastasis accounted for 80.3% of distant metastasis²² as compared to 100% in our study. One child also had brain metastasis and blood vessel invasion concurrently which signified late presentation and diagnosis when spread had already occurred.

In a study conducted in China in 2020, the most common PRETEXT stage at presentation was II (38.7%) and pretext IV was only 3.2%²⁹, while in our study, the most common PRETEXT stage was IV, 55% (n = 11), conferring high risk and negatively affecting outcomes. One child presented with spontaneous tumour rupture with accompanying haemorrhagic shock and demised within 5 days of presentation.

In the literature, the foetal histological subtype has been associated with the most favourable outcomes.³² However, when incompletely resected, the prognostic effect was not demonstrable.³³ We had two long-term survivors in our study: (1) the one had foetal + embryonal subtype and (2) the other had mixed epithelial subtype. Out of the children who demised, the longest lived for 3 years and 7 months had foetal + mixed epithelial subtype. This correlates with the study by Fan Li et al which showed that recurrence free rates were higher in children with mixed pathological subtypes.³⁴ Studies of larger cohorts are needed to correlate survival with the histological subtypes.

Super PLADO, 5-fluorouracil and irinotecan were used for relapses and refractory HB which is a similar practice in other centres.³⁵ Serial echocardiography (ECHO) and renal function tests were conducted to monitor for toxicity but there was no record of hearing tests in our study. High-frequency irreversible hearing loss has been noted in the literature to be the most common side effect of cisplatin chemotherapy.³⁶ In a study conducted in France in 2021, ototoxicity was noted to be the most concerning long-term treatment-related morbidity. They recommended hearing tests to be performed at least once within 5 years after diagnosis.³⁶

A study by Fan Li et al in 2021 showed that 71.1% of patients achieved initial complete remission, and they were standard risk patients (PRETEXT II and III) with a relapse rate of 30%.³⁴ Also, recurrence free survival rates were higher in children younger than 3 years and standard risk HB (PRETEXT II and III, no metastasis or vascular invasion).³⁴ In another study by T.Y Wang et al in China in 2008, 91% of the patients achieved remission.³⁷ In our study, only a third of patients, 35%, achieved initial complete remission which is lower than in the literature. Most of them (71%) had PRETEXT I staging, 5% PRETEXT II and 5% PRETEXT III, and only PLADO was used as chemotherapy. Three of these children (42%) relapsed, did not respond to second line and/or extended course of PLADO chemotherapy and subsequently demised; two abandoned treatments after 1 year and two were reportedly still alive at the time of the study (Table 3). Therefore, this study revealed that children with standard risk (PRETEXT stage I, II, III with no metastasis) were more likely to go into remission than high-risk patients. Our relapse rate is, however, higher than in the literature.

The lack of surgical resection of the tumour is a predictor of shorter OS.⁴ Most patients in our study could not be offered surgical resection because they were too unstable due to advanced disease signifying late presentation and diagnosis. More so, 20% (n = 4) of them died within 2 months of the diagnosis. This greatly impacted survival because all 10 (50%) of them who did not have surgical removal of the tumour had died by the third year of diagnosis. This correlates with a study conducted in Malawi in 2014 which showed 100% mortality in children who had incomplete or no surgical removal of the tumour.¹¹ In a study conducted in India in 2020, children who had surgical resection showed 1 year OS of 100% and 5 year OS of 83%.³⁸ In our study, the children who had surgical resection had 1 year OS of 70%. Survival will therefore improve if patients present with less advanced stages of the disease, and the whole tumour is successfully resected.

In our study, no patients received LT even though liver transplantation has been associated with higher survival outcomes in the literature.³⁹ Liver transplantation has been described as the preferred treatment modality for HR–HB³⁸ of which most of our patients were. In a study conducted in Turkey in 2022, it was observed that 33% of patients, in whom surgical resection was not possible, underwent LT and they achieved 75% survival.⁸ Also, in a study conducted in Seoul, South Korea in 2021, LT was observed to have a strong positive impact on the OS of patients with PRETEXT IV from 20% to 94%.⁴⁰ A study conducted at Red Cross War Memorial Hospital in 2004 showed that the development of liver transplant programme led to marked improvement in survival from 14% to 80%.^20,41 The 1 year OS for patients with PRETEXT IV in our study was noted to be 30%.

In South Africa, currently, there are only two liver transplant centres namely, Red Cross War Memorial Hospital and The Wits Donald Gordon Medical Centre. The latter has the largest dedicated paediatric liver transplant unit in sub-Saharan Africa. While LT in HB is a potentially curative option for some patients, it is not always the best choice. Factors such as chemotherapy response, tumour resectability, presence of metastases and the patient’s overall health play a role in the decision to proceed with transplantation or not. The above-mentioned factors played a considerable role in why it was mostly not considered as a treatment option. It emphasises the need to seek treatment in the early stages of the disease and complete resection of the tumour as the best chance of longer survival.

The median survival time in months in this study was 10 months which was significantly lower than that documented in literature: 17.86 months in Egypt in 2021¹⁷ and 45 months median follow-up time in Germany where children presented with less advanced disease.⁴

At the time of this study, 16 (80%) of the children had died even though HB has now been described as one of the curable malignancies in the paediatric population.⁴² Our study revealed a 1 year OS of 70% and a 5 year OS of 10% which is quite low compared to the findings of a study conducted in China in 2021 which showed 1 year OS of 95.3% and a 5 year OS of 79.8%.²² Also, the study conducted in Egypt by Soliman et al published in 2021 showed a 1 year OS of 73.8% and a 5 year OS of 51.3%.¹⁷

This study has the following limitations:

The study puts forward the following recommendations: