Population-based survival

Chapter on population-based survival as part of the children, teenagers and young adults  UK cancer statistics report 2021.

Survival estimates are the percentages of patients who are still alive at a specified time after their diagnosis of cancer, in this report we focus on five-year survival. Survival is a useful measure of the effectiveness of efforts to detect and diagnose cancers at an early stage and to treat them effectively.

Our analyses are for children and TYA who were residents of the UK and had a diagnosis of any malignant neoplasm or non-malignant CNS tumour included in ICCC-3 during the 20-year period of 1997-2016. The data set was the same as for the incidence analyses, except that cases ascertained by death certificate only were excluded (135 cases) and cases removed if lost to follow-up.

For people with multiple primary tumours, when calculating survival for all cancers combined, only the first primary tumours within the study period were included in the cohort count (1,077 multiple cases were removed). ¹⁰

The study censoring date for follow-up was 31 December 2018. Observed survival was estimated actuarially by Kaplan-Meier analysis using the ‘complete’ approach, in which all patients are included, not just those who have at least five years of follow-up. Trend in survival by single year of diagnosis was analysed by Cox regression and tested by the χ2 test with 1 degree of freedom.

A trend was defined as statistically significant if the p-value was less than 0.05. Given the large number of cancer groups for which a test for trend is reported, it should be remembered that some significant results would be expected to occur by chance.

Five-year survival rates for all cancers combined and for ICCC-3 main groups are shown in Tables 6, 7 and 8 in Appendix B for children, TYA and the whole 0-24 age group respectively for the UK and the constituent countries.

Detailed results for cancer subgroups and divisions of the diagnostic classification and for certain other subsets of particular interest are tabulated in Tables 9, 10 and 11 in Appendix B, for the UK only. In addition to results for persons diagnosed during the single 20-year period 1997-2016, results are given for 1997-2001, 2002-2006, 2007-2011 and 2012-2016 for diagnostic categories with at least 30 registrations in each of the four five-year periods.

Figures 4, 5 and 6 below show survival for all cancers combined for 0-14 year olds, 15- 24 year olds and 0-24 year olds in the UK. Appendix C shows the UK survival charts for subgroups and divisions of the diagnostic classification and for certain other subsets of particular interest.

Note:

¹⁰ This follows the principle used for survival analysis published by the Office for National Statistics and Public Health England. The number of cases included in this survival analysis will differ to National Statistics released by each UK nation, such as the “Childhood cancer survival in England: Children diagnosed from 2001 to 2017 and followed up to 2018)” published by the Office for National Statistics and Public Health England. The cancer cases for the analysis in this report have been classified using ICD-O-3 and with different definitions from those used for the National Statistics. For example, cancers of the skin other than melanoma and secondary and unspecified malignant neoplasms were included in our study but were excluded in the National Statistics. The data cleaning process for the analysis in this report also differs from that applied for the National Statistics.

Five-year survival for children under 15 years of age who were diagnosed with cancer in 1997-2016 was 81%, slightly lower than for TYA aged 15-24 years, 84% (Tables 6 and 7). This pattern was not, however, replicated across all diagnostic categories; with some instances where children aged under 15 had the better survival.

Despite relatively small numbers, the survival estimate was higher in Wales for childhood and TYA CNS tumours and for TYA leukaemias than in England and the UK as a whole. Possible explanations for this observation include: differences in coding or recording of cases between registries (for instance, increased recording of certain populations of asymptomatic or low grade tumours such as children with neurofibromatosis type I-related optic pathway gliomas); a chance finding (made more likely by multiple statistical testing of a defined population); and a real difference in survival. The differences were most marked in the early period and are a focus for investigation by the registries.

Overall, five-year survival from diagnosis with leukaemia was higher for children than for TYA, 85% compared with 68% in the UK. Children also had markedly higher survival than TYA for lymphoid leukaemia (89% vs. 65%) and acute myeloid leukaemia (67% vs. 57%), but lower survival than TYA for chronic myeloproliferative diseases (85% vs. 92%) and myelodysplastic syndrome and other myeloproliferative diseases (68% vs. 81%).

Survival from all lymphomas combined was 90% for children and for TYA. Survival from Hodgkin lymphoma was very high both for children (96%) and TYA (95%). For non Hodgkin lymphoma (including Burkitt lymphoma) children had higher survival (86%) than TYA (80%), and the difference between the age groups was most marked for precursor cell lymphomas (82% vs. 63%) and for Burkitt lymphoma (88% vs. 79%). Treatment strategies vary between the two age groups and it is unclear to what extent the differences in survival between children and TYA, albeit modest, reflect different biology versus different treatment approaches.

Survival from CNS tumours was lower for children (74%) than for TYA (78%). The survival gap in favour of TYA was wider for ependymoma (children 71%, TYA 88%), medulloblastoma (children 65%, TYA 75%) and for mixed and unspecified gliomas (children 44%, TYA 64%). Survival exceeded 90% in both children and TYA for pilocytic astrocytoma, craniopharyngioma and neuronal and mixed neuronal glial tumours. For astrocytomas other than pilocytic astrocytoma, children had higher survival (66%) than TYA (50%).

The spectrum of CNS tumours is different between young children, older children, adolescents and young adults, and a simple comparison of survival between CNS tumours in children and TYA misses much relevant detail. Even within a single tumour type, there may be marked morphological and biological differences with age. For example, the four major medulloblastoma subgroups have characteristic and distinct age peaks and outcomes that vary between infancy, mid-childhood and young adulthood, while the molecularly defined grade II-III ependymomas characteristic of young children are uncommon in TYA and older adults, who are more likely to have WHO grade I tumours or molecularly distinct grade II-III tumours.

Survival from neuroblastoma and other peripheral nervous cell tumours was similar for children (66%) and TYA (69%). Within that category, however, survival from neuroblastoma (including ganglioneuroblastoma) was higher for children (66%) than for TYA (53%).

Five-year survival of children with retinoblastoma was 99% while no cases were registered among TYA during the study period.

Children with renal tumours had higher survival (88%) than TYA (71%). The great majority of renal tumours in children were nephroblastoma (Wilms tumour), with survival of 91%, whereas most renal tumours in TYA were carcinomas, with somewhat lower survival of 71%. The survival gap in favour of children was much greater for hepatic tumours (children 74%, TYA 30%). Most hepatic tumours in children were hepatoblastoma, with five-year survival of 79%, but nearly all those in TYA were carcinomas for which survival was only 29%.

Children had higher survival than TYA from bone tumours (66% vs. 62%) and soft-tissue sarcomas (71% vs. 65%). The largest difference in survival within these categories was for rhabdomyosarcoma (children 69%, TYA 37%). Rhabdomyosarcoma still has one of the worst outcomes of any cancers in the TYA age group. The difference between childhood and TYA rhabdomyosarcoma survival is likely to be due, at least in part, to the relatively higher prevalence of fusion-negative tumours in children and fusion-positive or pleomorphic tumours in TYAs. Children also had higher survival than TYA with Ewing sarcoma family tumours, arising in bone and soft tissues (children 67%, TYA 52%). Survival from osteosarcoma, the most frequent bone tumour in people under 25 years old, was 60% for ages 0-24 years with little difference between children and TYA.

Survival of children and TYA with CNS germ-cell tumours was 88%, with little difference between the two age groups. For germ-cell tumours of other extragonadal and extracranial sites, children had higher survival (87%) than TYA (71%). Survival from gonadal germ-cell tumours was over 96% for children and for TYA. Gonadal carcinomas were rare in children but had survival of 77% in TYA.

Among children, five-year survival exceeded 95% for carcinomas in each of the two most frequent sites, thyroid and skin. Among TYA, in whom carcinomas are more frequent, survival varied widely by subgroup. Within the head and neck, five-year survival ranged from over 95% for thyroid and salivary glands to 73% for nasopharynx and 79% for other head and neck sites. Survival from carcinomas of trachea, bronchus and lung was 78% overall, but within this group of sites it was 97% for neuroendocrine tumours and 44% for other histological types. Five-year survival of TYA with carcinomas in the other most frequent sites was 99% for skin, 84% for cervix, 80% for breast, 57% for colorectal (except appendix), and 41% for stomach and other upper gastrointestinal tract. Survival from malignant melanoma was lower for children (86%) than for TYA (94%).

Overall, five-year survival was 77% for children under 15 years of age who were diagnosed in 1997-2001, 79% for those diagnosed in 2002-2006, 82% for those diagnosed in 2007-2011, and 84% for those diagnosed in 2012-2016 (Table 6).

Five-year survival was 79% for teenagers and young adults aged 15-24 years who were diagnosed in 1997- 2001, 83% for those diagnosed in 2002-2006, 86% for those diagnosed in 2007-2011, and 87% for those diagnosed in 2012-2016 (Table 7).

For both age groups, the trends in survival by period of diagnosis were highly significant (p<0.001).

There were also significant increasing trends in survival during 1997-2016 for many diagnostic categories, including lymphoid leukaemia, acute myeloid leukaemia, chronic myeloid leukaemia, chronic myeloproliferative diseases, myelodysplastic syndrome (in children), Hodgkin lymphoma, non-Hodgkin lymphomas, astrocytoma, other and mixed gliomas, brain stem glioma (in children), meningioma (in TYA), neuroblastoma (in children), osteosarcoma, Ewing sarcoma family of tumours (in TYA), rhabdomyosarcoma, CNS germinoma (in TYA), breast carcinoma (in TYA) and malignant melanoma (in TYA).

Some categories which already had an excellent prognosis with five-year survival exceeding 90% during the study period, with little room for further improvement; include pilocytic astrocytoma, pituitary adenoma and carcinoma, craniopharyngioma, retinoblastoma, nephroblastoma (Wilms tumour), gonadal germ-cell tumours, thyroid carcinoma and skin carcinoma.

The trend of survival increase was not uniform across the entire period, and there are several cancers with poorer prognosis which there has been little evidence of improvement in the most recent periods, notably ependymoma, medulloblastoma, hepatoblastoma, Ewing sarcoma in children and colorectal carcinomas in TYAs.

TYAs continued to have worse survival than children for osteosarcoma, Ewing sarcoma and rhabdomyosarcoma, without any recent improvement. However, while TYA survival remains worse than childhood survival for acute lymphoid and myeloid leukaemia and non-Hodgkin lymphomas, the gap has reduced over time.

The improvements in survival witnessed in acute myeloid leukaemia and acute lymphoblastic leukaemia during the period, which were relatively greater in TYA than in children, were likely to be due to successive improvements in treatment related to clinical trials in the two diseases, combined with a change in overall treatment strategy in acute lymphoblastic leukaemia in TYA patients from a typical 'adult' strategy to a more 'paediatric' strategy following the recognition that TYA patients had higher survival when treated according to paediatric protocols.

The introduction of novel targeted treatments for both Hodgkin and non-Hodgkin lymphomas is likely to have contributed to improvements in survival in both age groups seen over the study period seen over the study period.

The gradual improvement in survival seen across several CNS tumours throughout the period in children and TYA is encouraging but cannot be ascribed to a single factor.

Increasing specialisation of diagnostic imaging, molecular pathology and histopathology, and technical refinements in neurosurgery and radiotherapy are likely to have played a part.

For the malignant brain tumours ependymoma and medulloblastoma, there was little evidence of improvement. It was not possible to distinguish between patients with high risk and standard risk medulloblastoma, defined by a combination of age, stage, completeness of resection and histological and molecular features. Therefore, it is possible that there have been improvements in survival for some groups that were not detectable in this analysis. Future analyses should be able to better distinguish between patients with standard and high risk disease.

For ependymoma, the SIOP Ependymoma II clinical trial and the national Ependymoma Multidisciplinary Advisory Group (EMAG) collaborative began subsequent to the incident period examined in this report and their influence on the survival of childhood or TYA ependymoma have not yet been measured in national data.

The progressive improvements in the survival of osteosarcoma are welcome and somewhat difficult to explain. The last major clinical trial in osteosarcoma failed to show a difference between standard and escalated treatments, and the chemotherapy regimen in use has been standard for many years. It is likely, therefore, that the improvements in survival throughout the period are related to centralisation and increasing specialisation of surgical services, for a minority of patients the availability of high dose proton therapy for inoperable tumours, and greater standardisation of the established chemotherapy regimen.

One barrier impacting on the recent improvement in survival for rhabdomyosarcoma in TYA may be that successive European rhabdomyosarcoma trials have historically had an arbitrary upper age threshold for inclusion of 21 years. While these trials are routinely opened in paediatric centres, they are not routinely available to TYA patients. The abolition of an arbitrary age threshold for trial recruitment is a cause for hope for the future for these patients.

For Ewing sarcoma, there was an improvement in survival in TYA but not in children over the 20 year study period. The improvement in TYA is encouraging in a disease that is quoted in clinical trial literature as having no survival improvement for several decades. While TYA survival remains lower than that in childhood, the gap is now less than at the beginning of the study period.

The improvement over time is not fully explained: most of the trials undertaken during this period did not report improvements in survival other than the most recent trial, EuroEwing 2012, which took place during the most recent period. The improvements may reflect greater centralisation of care, more treatment according to standardised chemotherapy protocols and more centralised decision-making around local disease control via a national UK Ewing sarcoma multi-disciplinary team (MDT).

Figure 4 - All cancer combined. aged 0-14 years

Figure 5: All cancer combined, aged 15-24 years

Figure 6: All cancer combined, aged 0-24 years