Cancer incidence and spectrum among children with genetically confirmed Beckwith-Wiedemann spectrum in Germany: a retrospective cohort study

Beckwith-Wiedemann syndrome (BWS) is a cancer predisposition syndrome caused by defects on chromosome 11p15.5. The quantitative cancer risks in BWS patients depend on the underlying (epi)genotype but have not yet been assessed in a population-based manner. We identified a group of 321 individuals with a molecularly confirmed diagnosis of BWS and analysed the cancer incidence up to age 15 years and cancer spectrum by matching their data with the German Childhood Cancer Registry. We observed 13 cases of cancer in the entire BWS cohort vs 0.4 expected. This corresponds to a 33-fold increased risk (standardised incidence ratio (SIR) = 32.6; 95% confidence interval = 17.3-55.7). The specific cancers included hepatoblastoma (n = 6); nephroblastoma (n = 4); astrocytoma (n = 1); neuroblastoma (n = 1) and adrenocortical carcinoma (n = 1). The cancer SIR was highest in patients with a paternal uniparental disomy of 11p15.5 (UPDpat). A high cancer risk remained when cases of cancer diagnosed prior to the BWS diagnosis were excluded. This study confirms an increased cancer risk in children with BWS. Our findings suggest that the highest cancer risk is associated with UPDpat. We were unable to confirm an excessive cancer risk in patients with IC1 gain of methylation (IC1-GOM) and this finding requires further investigation.


METHODS
We performed a systematic survey among molecular genetic laboratories in Germany that offer BWSp testing. Nineteen different institutions were identified, of which 17 provided their data. One communicated that they had no molecularly confirmed cases. Another one did not respond. Generally, genetic testing was performed by methylation-specific multiplex ligation-dependent probe amplification (MS MLPA; Kit ME030, MRC Holland, Amsterdam, The Netherlands), and this assay includes both copy number and methylation analysis. Thus, duplications and UPD can be discriminated. A total of 321 individuals with molecularly confirmed BWSp were identified. We investigated the occurrence of childhood cancer in this cohort by matching their data with the GCCR. 18 We also included an analysis of both benign and malignant tumours of the central nervous system (the GCCR only collects benign tumours from the CNS, but not from other regions). The study was approved by the ethical review committee at the Hannover Medical School (#3281-2016). The matching procedure has been described elsewhere. 19 Briefly, molecularly defined cases of BWSp confirmed between 1 January 2006 and 31 December 2018 were identified at 17 main German laboratories that offer molecular genetic testing for BWSp. Using names and dates of birth, we matched laboratory-diagnosed cases of BWSp with the database of the GCCR (59,205 childhood cancer patients at cut-off date). As described previously, the GCCR registers~97% of all German childhood malignancies diagnosed at an age of up to 15 years in Germany since 1980. 18 All diagnoses defined in the International Classification of Childhood Cancer, Third edition are registered systematically. 20 The personal identifiers from individuals confirmed as having BWSp by the laboratories were encrypted using the same asymmetric key that is employed by GCCR. A matching procedure identified individuals with BWSp and childhood cancer, as described previously. 19,21 We accumulated person-years of observation from birth to date of last follow-up, as described previously. 19,21 Person-years of observation were left censored before 1 January 1980 and were observed through 31 December 2018, or right censored at cancer diagnosis, or their 15th birthday, or their death, if recorded, whichever occurred first. Vital status information in the laboratory data was incomplete; individuals without a specified status were assumed to be alive at the cut-off by default. All cancer cases younger than 15 years from the registry with an encrypted name between 1980 and 2018 were included in the matching procedure. Comparisons are presented as standardised incidence ratios (SIRs) with an exact 95% confidence interval (CI), as described previously. 19,21 Expected values were derived from the same subset of the GCCR data, which was used for the matching procedure. 18 SIRs were compared as described elsewhere. 22

RESULTS
We identified 321 individuals with a molecularly confirmed diagnosis of BWSp whose childhood period overlapped with the GCCR activity. Laboratory tests were conducted between 2006 and 2018. The following epigenetic and genetic defects were identified in peripheral blood: IC2-LOM (n = 208), UPDpat (n = 64), IC1-GOM (n = 31), duplication/deletion (n = 12), pathogenetic variant in CDKN1C (n = 6). This distribution is similar to the distribution known from other BWSp cohorts. [5][6][7] The 321 BWSp patients included in the final analytic data file contributed 2306.6 person-years of observation. Birth years ranged from 1978 to 2018. Age at genetic testing ranged from 0 to 38 years. The male-to-female ratio was 0.84. Thirteen patients with cancer, presenting between 1989 and 2018 and diagnosed with a BWSp-causing molecular defect between the years 2006 and 2018 were identified in the BWSp cohort (Table 1).
On the basis of all person-years and the age distribution of the studied population, 0.4 cases of childhood cancer, all sites combined, would be expected vs 13 observed, a 33-fold increase (SIR = 32.6; 95% CI = 17.3-55.7) ( Table 2). Age at diagnosis ranged from 0 to 3.8 years. The childhood cancer risk in BWSp patients due to UPDpat was 153-fold increased (SIR = 152.5, 95% CI = 73.1-280.5), whereas patients with IC2-LOM had an 11-fold (SIR = 11.4, 95% CI = 2.3-33.2) increased risk. There were no cancers  Table 3 To reduce selection bias, we conducted a sensitivity analysis by excluding the nine cases in whom the cancer diagnosis was made prior to the molecular BWSp (excluded cases are marked in Table 1). After exclusion of these cases, we calculated an unbiased 12-fold increased cancer risk (SIR = 12.3, 95% CI = 3.4-31.5) for all BWSp patients combined.

DISCUSSION
This is, to our knowledge, the first population-based study to quantify the cancer risk in children with BWSp with IC1-LOM, IC2-GOM, UPDpat, deletions/duplications and pathogenetic variants in CDKN1C. We observed a significant excess risk for all childhood cancers combined compared with the general population. The elevated overall cancer risk was primarily due to significant excesses of hepatoblastoma and nephroblastoma. Notably, we also observed rare cancers, such as astrocytoma and ganglioneuroblastoma. We found one previous case of ganglioneuroblastoma 9 and another case of brain glioma 23 in the literature.
The chromosome 11p15.5 region that is altered in individuals with BWSp harbours several imprinted genes, among them CDKN1C, IGF2 and H19, which are regulators of growth. 24 Therefore, the observed high cancer risk in individuals with BWSp who carry a constitutional or mosaic lesion in the 11p15.5 region is biologically plausible. Although a number of studies have shown a significant association between cancer risk and BWSp molecular subtype, 6 a population-based epidemiologic study has not yet been conducted to address this topic quantitatively. This approach has the advantage of being less prone to ascertainment bias and allows for the calculation of SIRs. Thus, methodological differences in study design, including different ascertainment and follow-up criteria, prevent a direct comparison of these studies with ours. Brioude 17 In contrast to these published studies, we have collected molecularly confirmed cases of BWSp in a population-based manner and have calculated the relative cancer risks compared to population-based incidence rates in an effort to analyse the cancer risks employing a less biased method.
The overall cumulative childhood cancer incidence observed in this cohort are in agreement with the cancer incidences described in previous studies. [5][6][7] We were able to demonstrate the elevated childhood cancer risk in individuals with BWSp compared to the population average risk, even when all patients with a cancer diagnosis prior to the BWSp diagnosis were excluded. In agreement with previous studies, we found a high cancer risk to be associated with UPDpat. 1 The risk in patients with UPDpat was significantly higher than the risk in IC2-LOM (Table 2). To our surprise, we observed no cases of cancer in BWSp children with IC1-GOM; however, this finding was not statistically significant. This molecular subgroup is believed to be associated with the highest cancer risk. 1 Larger cohorts of patients with IC1-GOM may be required to calculate the cancer risk more precisely. In agreement with previous reports, we observed a cumulative childhood cancer risk of <3% in BWSp children with IC2-LOM. 1,[5][6][7] Notably, even with small percentage risk this was still an 11-fold increased cancer risk with a wide CI. Although some molecular subgroups have a lower risk of cancer, parents who decline surveillance 1 should be aware that the absolute of cancer is still increased in these groups compared with the general population.
Our study has several limitations. (1) The study is not powered to detect any increase (or decrease) in incidence of the more common childhood cancers. Combining data with other cohorts may be a strategy to reach this goal. (2) We were unable to ascertain cancers in patients after their 15th birthday, as the   (4) Vital status information in the laboratory data was incomplete; individuals without a specified status were assumed to be alive at the cut-off by default potentially leading to an underestimation of the cancer risk. (5) As a consequence of identifying susceptible individuals through genetic diagnostic labs, our analytic cohort excludes BWSp patients who have never undergone genetic testing (i.e. who were diagnosed clinically without molecular confirmation) as well as BWSp patients who had a clinical diagnosis but negative molecular testing. There is no way to evaluate the impact of this subgroup's absence on our analysis. Given the fact that the percentage of tested and untested individuals with BWSp is unknown, we are unable to calculate the BWSp incidence based on our data. (6) The lack of replicating previous results on an increased cancer risk for IC1-GOM might reflect the smaller number of Person-Years analysed in this group (5-fold less than for IC2-LOM). (7) One German laboratory with molecularly confirmed cases of BWSp did not participate in this research and it cannot be ruled out that some additional patients received their molecular tests in laboratories that we did not identify, because their diagnostic portfolio is not publicly available, or in laboratories from other countries. However, outsourcing of genetic testing is unusual in Germany, as health insurance companies routinely cover genetic testing only if done in a national lab.
In conclusion, this population-based study demonstrated an increased childhood cancer risk, especially hepatoblastoma and nephroblastoma, in children with BWSp. Our findings suggest that in children with BWSp the highest cancer risk is associated with UPDpat, whereas the risk in IC1-GOM requires further investigation. Our data are consistent with current surveillance guidelines. 1 The patient had neuroblastoma and ganglioneuroblastoma.
Cancer incidence and spectrum among children with genetically confirmed. . . S Cöktü et al.