2.1 Patients

We undertook a retrospective folder review of all children from birth up to and including 16 years of age with histologically confirmed mature teratoma (MT) and IT from nine paediatric oncology units (POUs) across South Africa in the public and private sector. Our ability to collect detailed data was limited by the retrospective nature of the study over an extended period of time. Histologically MTs⁹ were considered to be those with a mixture of mature/differentiated tissues from all three germ layers, whilst ITs¹⁰ were defined as those with variable amounts of immature/embryonal tissue, and were graded 1, 2 or 3 depending on the amount of immature neural tissue present. In the earlier part of the time period, grade was almost never assigned, leaving only a small minority of patients assigned with the relevant grade toward the latter end of the time period. Patients with teratomas containing malignant non-germ cell tumour elements were excluded. None of the remaining patients with ITs were reported to have malignant germ cell tumour elements present. Tumour marker elevations were checked for diagnostic significance using standard nomograms.² All patients with ITs with elevated levels of beta-human chorionic gonadotropin (β-HCG) and alpha-fetoprotein (AFP) for age were excluded on suspicion that they may represent mixed germ cell tumours. Patients with metastatic disease were excluded under the same presumption. Deidentified demographic and presentation data included age, sex, nutritional status assessed by anthropometry, socio-economic status (SES) measured by annual household income, symptom profiles, co-morbidities, tumour stage, site and histology and tumour markers. Staging was carried out pre-operatively with the use of computed tomography or magnetic resonance imaging in conjunction with elevated tumour markers and/or a tissue biopsy in patients where a safe upfront resection was not possible. In patients with resectable tumours, post-operative notes in conjunction with pathology reports were used to assess microscopic or macroscopic residual disease, as it impacted stage in the context of a multidisciplinary discussion forum. Treatment-related and outcome data included timing and extent of tumour resections, chemotherapy regimens (neo-adjuvant and adjuvant) where relevant, the number of courses of chemotherapy, the use of radiotherapy, medical and surgical treatment-related sequelae and interventions, the nature of follow-up and disease outcomes. Lost to follow-up was defined as patients who did not return for consultations prior to 5 years from the end of treatment, and treatment abandonment was defined as failure to complete potentially curative therapy. Nutritional parameters were defined according to the WHO Nutritional Landscape Information System (NLIS) Interpretation Guide, 2010.¹¹ SES was assigned using the Western Cape Government's Federal Acquisition Regulation No. 9 of 2017. Given the retrospective nature of the study, consent was waivered.

2.2 Statistical analysis

Stata Statistical Software v. 14 (StataCorp) was used for consolidation, uniform recording and descriptive statistics. Survival analyses and output graphics were conducted with R Statistical Environment v. 4.1.2 (R Core Team). The Kaplan–Meier procedure was used to estimate overall survival (OS) curves and survival at specific time points (1, 2 and 5 years), with associated 95% confidence intervals.^{12, 13} The log-rank test was applied to assess the existence of significant differences between survival curves in different subgroups. Multivariate Cox regression models were used to determine the association of the OS and event-free survival (EFS) relative to a series of potential prognostic factors.¹⁴ The proportional hazards assumption was computed using the scaled Schoenfeld Residuals test.¹⁵ A p-value less than 5% was used as a cut-off to define statistical significance.

2.3 Ethical approval

Ethical approval was obtained from the Human Research Ethics Committee of the University of Cape Town (HREC 002/2018) with reciprocal approval obtained from all participating centres nationally.

3.1 Demographics

There were 60 patients with MT and 40 with IT eligible for analysis. The median age in the MT group was 7 months (interquartile range [IQR]: 0–74.5 months). Male patients (n = 14) presented at a younger median age of 2 months (IQR: 0−8.75 months) compared to female patients (n = 46) at 9 months (IQR: 0–88.5 months). Similarly, male patients (n = 10) with ITs presented at a younger median age of less than 1 month (IQR: 0–1.75 months) compared to female patients (n = 30) at 4.5 months (IQR: 0–162months).

3.2 Tumour site

In the combined group, 31 patients presented with ovarian primaries and the remaining 69 were all extragonadal: 30 sacrococcygeal, 19 in the head and neck, 11 retroperitoneal (abdomen and pelvis) and nine in the mediastinum. Table 1 shows the disaggregation by histological subgroup.

3.3 Stage

Stage 1 patients comprised the majority of MTs (n = 55; 91.6%) and 75% (n = 30) of the ITs. Table 2 shows the distribution of patients by stage in each group (Table 2).

3.4 Tumour markers

AFP was the most commonly elevated documented tumour marker at diagnosis in 55 (27 MTs and 28 ITs) patients. All patients had levels that plotted within normal ranges according to the standard age nomograms.

3.5 Treatment interventions and outcomes

3.6 Follow-up

3.7 Statistical analysis

Age, sex, nutritional status, SES and tumour site did not impact outcome in a significant way. Stage was highly predictive of outcome, both on univariate and multivariate analysis for both groups (MT p < .001 and IT p = .013), particularly for patients with Stage 3 disease (p < .001). Patients with Stage 1 disease had a significantly higher chance of survival compared to those with Stage 2 and 3 disease (HR 19.32; p = .0031) (Table S1). The EFS by stage for MTs and ITs is shown in Figure 1 (p < .001). The EFS for MTs (p < .001) and ITs (p = .013) by stage disaggregated by group are shown in Figures S1 and S2.

3.8 Outcomes

The 5-year estimated OS ratio for the whole group was 85.4% (95% CI: 79.1%–92.2%), and the 5-year EFS ratio for the whole group was 78.7% (95% CI: 71.1%–87.0%). The 5-year OS for patients with MTs was 80.8% (95% CI:71.9%–90.9%), while the EFS for patients with MTs was 74.0% (95% CI: 64.1%–85.5%). The 5-year OS for those with ITs was 92.8% (95% CI: 86.1%–100%) (p = .151), while the EFS for those with ITs was 86.7% (95% CI: 76.9%–97.9%) (p = .115). There was no significant difference in OS (p = .337) (Figure 2) or EFS (p = .512) (Figure 3) between MTs and ITs.

4.1 Limitations

We acknowledge the limitations of this retrospective review inherent with the idiosyncrasies of local and often disparate historical, institutional decision-making in terms of pathology reporting, access to specialised surgery and the use of chemotherapy. Additionally, central histology review in our setting was not possible or available historically. Interpretation of data was made more challenging by differences in laboratory assessments of disease markers, differing timing and modalities of disease response assessments and the lack of clear details with respect to fertility-sparing surgery in those with ovarian tumours. The length of the review period is also complicated by the changes in definitions of disease and the evolving approaches to management over time, which may have generated inconsistencies in practice. We recognise that divergent practice is a reality of historical cohorts, but this only emphasises the necessity for a cohesive new national treatment approach as an imperative to improve survival going forward.

4.2 Implications for practice

This retrospective study is part of a larger project to harmonise treatment guidelines for children with malignant and benign EGCT in South Africa with the aim of improving survival rates by standardising risk stratification, surgical approaches and indications for chemotherapy. The guideline also provides an opportunity to identify and correct systemic inconsistencies in practice. To that end, central pathology review has been built in as a critical fall back to ensure accurate diagnosis and as a mechanism to standardise pathology reporting between centres. It also seeks to build increasingly collaborative networks whose components adhere to best practice surgical guidelines, and limit the administration of potentially harmful chemotherapy, often as a reflexive strategy in an environment where systemic delays in services and resource limitations cause worrying hiatuses in care. Appropriate treatment is built on accurate diagnosis, and multidisciplinary team meetings continue to play a critical role in the diagnostic and therapeutic evaluation of patients. Collaborative, well-communicated working plans, which include clinical reviews by both oncology and surgical staff, cast the foundations of integrated inter-disciplinary care, which ultimately serves the patient. The early identification of pre- and post-operative complications is the desirable endpoint of a system that blends continuous in-service training and heightened surveillance with shared responsibility. These inter-disciplinary assessments should happen as a matter of course, whether electively scheduled or done post-operatively in the case of children who require emergency surgical intervention. The ambulatory setting post discharge provides clinicians (oncologists and surgeons) with another opportunity to consolidate clinical reviews, contextualise pathology results for the patient and caregivers and to craft a plan for follow-up.

Consequently, the guideline also seeks to create more parity in structured follow-up, particularly for the children from low-income households where we found the attrition rates to be higher, evidenced by the fact that 84.6% of patients who were lost to follow-up were from low socio-economic households. This is aggravated by the ongoing lack of consensus in defining adult versus paediatric age cut-offs for care, inconsistencies in or the absence of supported transitional care for children into the adult service and the absence or minimal availability of combined adult-paediatric clinics. All of these factors conjointly increase the chances that patients are lost (especially the most socio-economically vulnerable), and prevent us from collecting reliable long-term survival and late effects fertility data. Instituting a national guideline creates an opportunity to bring allied disciplines like registry services, social workers and gynaecology and endocrine services into the fold and to develop new spaces that have historically been in deficit.