Second primary cancers in people who had cancer as children: an Australian Childhood Cancer Registry population‐based study

Recent therapeutic advances for children with cancer — typically, intensive chemotherapy, often in combination with radiation therapy — have improved survival for many with childhood malignancies.1,2 However, a large proportion of patients experience adverse health effects throughout life that are attributable to either their cancer or its treatment, including organ dysfunction (eg, cardiopulmonary, renal, gastrointestinal dysfunction), impaired physical development, obesity, problems with vision or hearing, reduced fertility, and neurocognitive deficits.3,4

One of the most serious potential consequences is a second cancer. Although it is recognised that survivors of childhood cancer have an increased risk of being diagnosed with further cancers, the available information has largely been derived from studies of patients diagnosed before modern treatment protocols were introduced, or from studies limited to second cancers diagnosed at least 5 years after the first diagnosis.5,6 The only published Australian data derive from a single institution study in New South Wales that was restricted to children with a first diagnosis of cancer during 1972–1999 and who were still living at least 5 years after this diagnosis.7

In this article, we report our investigation of second primary cancers in a more contemporary national, population‐based cohort of people diagnosed with cancer during childhood. It is also the first study in Australia to examine the risk of a second cancer in the years immediately following the diagnosis of the first cancer.

Methods

Data source

We analysed de‐identified unit record data from the Australian Childhood Cancer Registry (ACCR), which includes data on all cancers diagnosed in children under 15 years of age in Australia since 1983. The ACCR ascertains cases from each state and territory cancer registry annually.

Information on second primary cancers is obtained by regular matching of ACCR data with the comprehensive Australian Cancer Database, which holds data on all primary malignant cancers (except basal and squamous cell carcinomas of the skin) diagnosed in Australian residents of any age. The rules for defining second primary cancers are complex: in our study, a new cancer was generally deemed a second primary if its morphology (Berg group)8 was distinct from that of the first cancer; if the second tumour had the same or a related morphology, it was classed as a recurrence or relapse of the original primary tumour.

We included data for patients with a first diagnosis of cancer between 1 January 1983 and 31 December 2013, with follow‐up on second cancers available to 31 December 2015 (ie, 2–33 years' follow‐up); the maximum age attained by an included person was 47 years (a 14‐year‐old patient diagnosed with cancer in 1983 and still living at the end of 2015). Second primary cancers diagnosed within two months of the first cancer were excluded to minimise detection bias. To allow comparison with the results of other studies, we also estimated the incidence of second cancer in people whose second cancer was diagnosed at least 5 years after their first. Patients with recurrence of the original cancer or progressive disease were not included in this analysis, nor were cancer diagnoses subsequent to the diagnosis of the second primary cancer.

Statistical analysis

The estimated risk of a child developing a second primary cancer, compared with the incidence of cancer in the general population, was expressed as a standardised incidence ratio (SIR). For cancers in patients under 15 years of age, population incidence rates were based on ACCR data; for cancers in patients aged 15 years or more, incidence was calculated from the Australian Cancer Incidence and Mortality (ACIM) workbooks published by the Australian Institute of Health and Welfare for all cancers combined.9 Time at risk accumulated from two months after the first diagnosis until 31 December 2015, the date of diagnosis of a second primary cancer, or the date of death, whichever occurred first. The number of second primary cancers expected in the study cohort was obtained by applying the cancer incidence rate for the general population to the person‐years at risk, matched by sex, 5‐year age group, and calendar year. SIRs were then calculated by dividing the actual number of second primary cancers by the expected number. Differences between SIRs by characteristics of interest were assessed in unadjusted negative binomial models, with the actual number of cancers as the dependent variable, offset by the log of the expected number; results were expressed as incidence rate ratios (IRRs).

Cumulative 30‐year incidence was calculated with the Stata command stcompet.10 SIRs were calculated with the Stata command strate; 95% confidence intervals (CIs) were calculated with the quadratic approximation to the Poisson log likelihood for the log‐rate parameter.11 Data were analysed by sex, age group at first diagnosis, and type of first childhood cancer, defined in diagnostic groups or subgroups according to the International Classification of Childhood Cancer, third edition (ICCC‐3).12 SIRs by mode of treatment (chemotherapy only, radiotherapy only, both modes, neither mode) and time from first diagnosis (2 months to less than 2 years, 2 to less than 5 years, 5 to less than 10 years, 10 to less than 20 years, 20 years or more) were estimated. Time from first diagnosis was also modelled as an ordinal variable in a negative binomial model to examine the trend by time from first diagnosis.

Ethics approval

The ACCR operates with ethics approval from the University of Queensland Behavioural and Social Sciences Ethical Review Committee (reference, 2004000302), as well as from the human research ethics committees at each of the state and territory cancer registries and major paediatric treating hospitals. Separate ethics approval was not required for this study.

Results

Of the 18 230 children under 15 years of age diagnosed with cancer during 1983–2013 and still living at least two months after their diagnosis, 388 (2%) were diagnosed with a second primary cancer during 228 411 person‐years of follow‐up (median follow‐up, 10.9 years; interquartile range [IQR], 3.9–20.0 years). The median time between first and second diagnosis was 9.9 years (IQR, 4.0–16.4 years).

The overall cumulative incidence of second primary cancers 30 years after first diagnosis was estimated to be 4.4% (95% CI, 3.8–5.0%). The risk of being diagnosed with a new primary cancer was five times as high for people diagnosed with a first cancer during childhood as for the general population (SIR, 5.13; 95% CI, 4.65–5.67) (Box 1); the absolute excess risk was 13.9 (95% CI, 13.4–14.2) cancers per 10 000 person‐years at risk.

For the 12 699 people diagnosed with cancer during childhood and alive 5 years later (69.7% of the cohort), the median follow‐up time was 16.2 years (IQR, 3.9–20.0 years). For this subgroup, which included 271 of the 388 second primary cancers (70%), the SIR was 4.29 (95% CI, 3.81–4.83). Thirty‐year cumulative incidence was similar to that of the entire cohort (4.6%; 95% CI, 4.0–5.3%).

The relative risks of second primary cancer at least two months after the first diagnosis were similar for both sexes (females v males: IRR, 1.14; 95% CI, 0.94–1.39). Although risk was elevated for all diagnosis age groups, the SIR for children aged 0–4 years at first diagnosis was significantly higher than for those aged 10–14 years (IRR, 1.52; 95% CI, 1.21–1.92). SIRs by diagnostic group of childhood cancer ranged between 1.90 (95% CI, 1.10–3.27) for other malignant epithelial neoplasms and melanoma to 9.23 (95% CI, 6.84–12.4) for soft tissue sarcomas, which includes rhabdomyosarcomas (SIR, 19.9; 95% CI, 14.3–27.6), with 36 cases of second primary cancer diagnosed in 567 people who had rhabdomyosarcoma during childhood (6.3%). Relative risk of second cancers were also large for the diagnostic subgroups intracranial and intraspinal embryonal tumours (SIR, 9.79; 95% CI, 6.38–15.0) and Ewing tumours and related bone sarcomas (SIR, 11.4; 95% CI, 7.25–17.8) (Box 1).

Patients who underwent neither chemotherapy nor radiotherapy had the lowest relative risk of second cancers, although their risk was still significantly higher than for the general population (SIR, 2.14; 95% CI, 1.64–2.78); children who had undergone both chemotherapy and radiotherapy had the highest risk of a second primary tumour (SIR, 9.80; 95% CI, 8.35–11.5; chemotherapy and radiotherapy v neither chemotherapy nor radiotherapy: IRR, 4.59; 95% CI, 3.37–6.23) (Box 2).

Relative risk was highest in the years immediately following the first diagnosis, the SIR peaking 2–5 years after diagnosis at 10.3 (95% CI, 8.20–13.0) and declining to 2.58 (95% CI, 2.02–3.30) for 20–33 years (for trend: P < 0.001) (Box 3).

The most common types of second primary cancer were thyroid carcinomas (65, 17%; median time between first and second diagnoses, 14 years) and acute myeloid leukaemias (57, 15%; median time between diagnoses, 2.3 years). There were also 57 second primary cancers in the diagnostic group “other and unspecified carcinomas”, including 15 women with breast cancers. Among people alive 5 years after their first diagnosis, almost one‐quarter of the second primary cancers diagnosed were thyroid carcinomas (63 of 271, 23%), but only 11 (4%) were cases of therapy‐related acute myeloid (Box 4).

Discussion

We report the first national data on second primary cancers in people diagnosed with cancer during childhood in Australia, including second cancer diagnoses within 5 years of the initial diagnosis. The only previous Australian investigation of second primary cancers following childhood cancer was much smaller: the cohort comprised 896 children who were under 14 years of age at diagnosis, had been treated at the Sydney Children's Hospital in Sydney between 1972 and 1999, and were alive 5 years after their initial diagnosis. The SIR in this study (4.98), based on 16 second primary cancers,7 was similar to our figures of 5.13 for children alive at least two months and 4.29 for those alive at least 5 years after their first diagnosis.

In other countries, reported SIRs for second primary cancer in people who have had childhood cancer range between 3 and 20;5,6,13,14,15,16,17,18 our findings are at the lower end of this range. The variance in risk estimates can be attributed to differences in an array of factors that make direct comparisons difficult, including length of follow‐up, the time period of the initial diagnoses, the cancer types included, forms of treatment of the first cancer, criteria regarding age at diagnosis, local cancer incidence rates, and type of study. For example, five of the eight cited investigations included only patients diagnosed before 2000 (and as early as the 1940s),5,6,13,14,15 four included only people alive 5 years after the initial diagnosis,5,6,14,15 and five included patients first diagnosed after the age of 15 years;5,14,15,16,18 five were population‐based,6,13,15,16,18 three were institutional studies.5,14,17

Overall cumulative 30‐year incidence of cancer for 5‐year survivors in our study (4.6%) was substantially lower than for 5‐year survivors in the United States Childhood Cancer Survivor Study (CCSS; 7.9%).5 This may be explained by differences in the time period of the first diagnoses (our study, 1983–2013; CCSS, 1970–1986), in age at first diagnosis (our study, under 15 years; CCSS, under 21 years), and median follow‐up time (our study, 16 years; CCSS, 23 years).

The greatest relative risk of second primary cancers in our study was for people diagnosed with rhabdomyosarcoma during childhood; the SIR of 19.9 was much higher than the 5.7 estimated by a United States study that included people diagnosed with rhabdomyosarcomas before the age of 19 years during 1973–2010.19 However, the authors of the American study noted that most of the second cancers identified were in people with first primary rhabdomyosarcoma diagnoses during 2000–2010; restricting analysis to these cases resulted in a SIR of 21, similar to our figure. Without explaining the recent sharp rise in relative risk for people diagnosed with childhood rhabdomyosarcoma, the authors suggested that they may be genetically predisposed to cancer, adding to the effects of exposure to chemotherapy and radiotherapy while young.19

Endocrine system‐related second primary cancers were relatively frequent among 5‐year survivors in our study, particularly thyroid and breast cancers. Radiotherapy is the only risk factor for subsequent development of these two solid cancer types supported by strong clinical evidence.20 Chemotherapy of childhood cancer increases the risk of a second cancer diagnosis, particularly of haematological malignancies,21 the most common form being therapy‐related acute myeloid leukaemia.22 About 80% of therapy‐related acute myeloid leukaemias and 30% of all second primary cancers in our cohort were diagnosed within 5 years of the initial cancer diagnosis, which indicates the importance of estimating second primary cancer risk from close to the first diagnosis. While it is possible that lifestyle and environmental factors also contribute to the increased risk of second cancers, the health behaviour of people who had childhood cancer is similar to that of their peers, apart from their generally being less physically active.23

Complementing our findings, other studies have found that survivors of childhood cancer remain at increased risk of cancer many decades after their initial diagnosis,16,24 showing the importance of ongoing surveillance. However, a recent Australian survey of long term survivors of childhood cancer found that more than 60% believed they were not at greater risk of new neoplasms than other people,25 which suggests that educating people about this topic is difficult.

Strengths and limitations

The population‐based cohort and national coverage for second primary cancer diagnoses were major strengths of our study. All analysed data were independently recorded for administrative purposes at the time of diagnosis, removing the potential for information or recall bias. To reduce the possibility of increased medical surveillance of newly diagnosed cancer patients leading to detection bias, we excluded second primary cancers identified within two months of the first cancer.

We acknowledge that some second primary cancers diagnosed soon after the establishment of the ACCR in 1983 may have been incorrectly classified as first primary cancers, but this would have had only a small impact on our results. We could not analyse second primary cancers following the administration of specific chemotherapeutic agents because this level of treatment detail is not currently available in the ACCR. Finally, our results were not adjusted for potential confounders; for example, information about risk factors is not collected by the ACCR, and cancer stage at diagnosis was only recorded for more recent cases (2006–2014).

Conclusion

Second primary cancers are a lifelong possibility for people who survive childhood cancer, and they are a leading cause of treatment‐related mortality among survivors.26,27 Whereas the estimated size of the effect depends on study design, our finding that survivors of childhood cancer are at significantly increased risk of subsequent cancers is consistent with findings from other studies; the increase in risk gradually diminishes over time from first diagnosis, but the level of risk never returns to that of the general population. Advances in treatment for childhood cancer in recent decades have significantly improved survival for many people with childhood malignancies; the challenge is to reduce treatment‐associated morbidity without reducing survival.

Access to study data

We have full access to all data (including statistical reports and tables) analysed in this study.

Box 3 – Standardised incidence ratios for second primary cancers diagnosed two months to 33 years after diagnosis of a primary cancer before the age of 15 years, Australia, 1983–2013, by time from first diagnosis

---

CI = confidence interval. ^* Compared with incidence rate for general Australian population; follow‐up to 31 December 2015.