Cancer in Children: Clinical Management, 5th Edition

Chapter 1. The epidemiology of cancer in children

C. A. Stiller

G. J. Draper

Introduction

Although childhood cancer is rare, accounting for less than 1 per cent of all cancer in industrialized countries, it is of great scientific interest for a number of reasons. Several types of cancer are virtually unique to childhood, whereas the carcinomas most frequently seen in adults, those of lung, female breast, stomach, large bowel, and prostate, are extremely rare among children. Some of the most striking progress in cancer treatment has been made in paediatric oncology. Investigation of childhood tumours has led to major advances in the understanding of the genetic aetiology of cancer. In this chapter we consider the classification of childhood cancer, the principles of cancer registration, incidence and survival rates, and aetiology.

Classification

The great majority of malignant neoplasms occurring in adults are carcinomas, and so the International Classification of Diseases, by which cancers other than leukaemias, lymphomas, mesothelioma, Kaposi sarcoma and cutaneous melanomas are classified purely by site of origin, is reasonably satisfactory for the presentation of their incidence rates. In contrast, childhood cancers exhibit great histologic diversity and often arise in embryonal precursor cells. Some types of tumour can arise in many different primary sites. Therefore it is more appropriate for childhood cancers to be classified according to histology and site. The current standard classification is the International Classification of Childhood Cancer,1 with groups defined by the codes for morphology in addition to topography in the second edition of the International Classification of Diseases for Oncology (ICD-O). A new version, based on codes in the third edition of ICD-O, is in preparation. The classification contains 12 main diagnostic groups, most of which are divided into subgroups. The main groups are as follows: I leukaemia; II lymphomas and reticuloendothelial neoplasms; III central nervous system and miscellaneous intracranial and intraspinal neoplasms; IV sympathetic nervous system tumours; V retinoblastoma; VI renal tumours; VII hepatic tumours; VIII malignant bone tumours; IX soft tissue sarcomas; X germ cell, trophoblastic, and other gonadal neoplasms; XI carcinomas and other malignant epithelial neoplasms; XII other and unspecified malignant neoplasms. Most of these groups are limited to malignant neoplasms, but there are two exceptions. Benign and unspecified intracranial and intraspinal tumours, including choroid plexus papilloma, ganglioglioma, non-malignant gliomas, craniopharyngioma, pituitary adenoma, pinealoma, meningioma, and tumours of unspecified type are included in group III because they are recorded by many cancer registries. For the same reason, non-malignant intracranial and intraspinal germ cell tumours are also included (in group X).

Childhood cancer registration

The aim of a cancer registry is to collect information on all cases of cancer occurring within a geographically defined population. Sometimes there are insufficient resources to maintain a population-based registry, but nevertheless it may be possible to collect similar data on all cases seen in one or more hospitals or pathology departments. In the absence of population-based data, these series can still yield much useful information, and several population-based cancer registries trace their origins to hospital or pathology series. The amount of data collected on each case varies enormously between registries, but the minimum consists of the patient's name (or other unique personal identifier), sex, date of diagnosis and age at diagnosis, and sufficient information on the primary site and histologic type of the neoplasm for it to be coded according to whatever system is used, preferably ICD-O. Most registries collect the patient's date of birth where known, as this is particularly useful for linking data on the same patient from different sources and for detecting duplicates. The basis of diagnosis is also usually recorded, at least to the extent of whether there was histologic verification. There can be many different sources of information, of which the most frequently used are clinical records, pathology records, death certificates, and other cancer registries. The registration data should be checked for internal consistency. Computer programs which do this for age, sex, primary site, and histologic type are included in the IARC Technical Report International Classification of Childhood Cancer.1 The IARC monograph, Cancer Registration: Principles and Methods,2 gives detailed information on many aspects of setting up and maintaining a cancer registry. The operating methods of many individual registries are described in International Incidence of Childhood Cancer, Volume II.3 Recommendations on registry practices and guidelines on confidentiality are given in a recent publication by the European Network of Cancer Registries.4

Population-based general cancer registries, including patients of all ages, now cover the whole or part of about 100 countries. As childhood cancer accounts for a very small proportion of all cancer, registries which restrict their coverage to children are able to collect more detailed information on each case. There are population-based childhood cancer registries in several countries or parts of countries in Europe, the Americas and Oceania. These registries are often operated in close collaboration with paediatric oncologists; in Germany, for example, there is a common follow-up system for the childhood cancer registry and for clinical trials. In many countries where there is no national population-based cancer registration, including the USA, Japan, and Spain, registers of patients are maintained by national organizations of paediatric oncologists or by clinical trial groups.

Incidence

Incidence in Great Britain

Table 1.1 gives numbers of cases and incidence rates for the 12 main groups and principal subgroups of the International Classification of Childhood Cancer in Great Britain during the period 1987–1996. The pattern of incidence is typical of that found among the mainly White populations of industrialized countries in Europe, North America, and Oceania, although some of the rates are towards the lower end of the usual range for these populations. The total age-standardized annual incidence was 134 per million children, giving a cumulative risk of 1 in 514 of developing cancer during the first 15 years of life. About a third of all childhood cancers are leukaemias, predominantly acute lymphoblastic leukaemia (ALL). Brain and spinal tumours are the second most common diagnostic group, accounting for about a quarter of registrations, with astrocytomas being the most frequent histologic type. Lymphomas account for 9–10 per cent, and non-Hodgkin lymphoma (NHL) has a somewhat higher incidence than Hodgkin disease. Neuroblastoma and Wilms tumour, the two most frequent embryonal tumours of childhood, each account for 6–7 per cent of registrations, as do soft tissue sarcomas, while retinoblastoma accounts for 3 per cent. Nearly all the remaining cases are bone sarcomas, germ cell tumours, and epithelial tumours. Of this last group, malignant melanoma, skin carcinoma, and thyroid carcinoma are the most frequent, but none of them accounts for more than 1.5 per cent of all childhood cancer.

Within childhood, the total incidence of cancer is highest (180 per million) in the first 5 years of life, compared with about 100 per million for the age group 5–14 years. The age–incidence distribution varies considerably between diagnostic groups. There is a marked peak in the incidence of ALL at age 2–3 years. Early age peaks are also found for all the distinctive embryonal tumours. The highest incidence of neuroblastoma, retinoblastoma, and hepatoblastoma is in the first year of life, but the peak for Wilms tumour occurs slightly later. In contrast, Hodgkin disease and bone sarcomas are virtually never seen before the age of 2 years, and their incidence increases steeply throughout childhood and adolescence. Among boys, the incidence of testicular germ cell tumours is highest in early childhood, and the start of the sharp increase in incidence during adolescence and early adulthood is barely noticeable before the age of 15; among girls, ovarian germ cell tumours are rare until the postpubertal increase, which begins at an earlier age than among boys.

International variations

There is considerable systematic variation in the many types of childhood cancer between different regions of the world and between ethnic groups in the same country.3,5

Leukaemia

In the USA, there is a substantially lower incidence of ALL in the Black population and the early childhood peak is very much attenuated. In contrast, there is little evidence of ethnic variation in the incidence of childhood leukaemia in the UK; in particular the pattern of occurrence of ALL among both Black children and children of South Asian ethnic origin is very similar to that among White children, with a marked peak in early childhood. In many developing countries of Asia and Latin America the early childhood peak is also less marked and the total incidence is again lower. A similar pattern has been found in the former socialist countries of central and eastern Europe, although there are indications that a more marked early childhood peak is evolving.6 There is little international variation in the incidence of acute non-lymphoblastic leukaemia (ANLL).

Lymphomas

Childhood Hodgkin disease has a relatively high incidence, particularly among younger children, in developing countries of Latin America and the Middle East. Lymphomas, both Hodgkin disease and NHL, are more common among South Asian children in the UK than among White children, and the excess of Hodgkin disease is again greatest among younger children. The highest incidence of Burkitt lymphoma, sometimes as large as 80 per million, is found in a broad geographic band of tropical Africa where malaria is endemic, and the incidence is also high in Papua New Guinea; in both of these regions, Burkitt lymphoma is the most common childhood cancer. Elsewhere it is harder to identify patterns of incidence for Burkitt lymphoma since many cases have been registered simply as NHL. The incidence of all NHL, including Burkitt lymphoma, is relatively high in Mediterranean countries and the Middle East and in some Latin American countries.

Table 1.1. Childhood cancer in Great Britain, 1987-1996: numbers of registrations; age-specific, agestandardized (World Standard Population), and cumulative incidence rates, and sex ratio

Diagnostic group

Totalregistrations

Annual rates per million by age group

Age-standardized rate per million

Cumulative rate per million

Sex ratio(M/F)

0-4 years

5-9 years

10-14 years

All cancers

13904

180.2

103.2

105.9

133.8

1947

1.2

I Leukaemia

4413

67.1

31.9

23.8

43.1

614

1.3

   Acute lymphoblastic

3580

56.2

26.9

16.3

35.2

497

1.3

  Acute non-lymphocytic

674

8.8

4.2

6.0

6.5

95

1.1

  Chronic myeloid

87

1.3

0.4

0.7

0.8

12

1.8

  Other specified

13

0.1

0.1

0.2

0.1

2

0.6

  Unspecified

59

0.8

0.3

0.6

0.6

8

1.3

II Lymphomas

1298

6.2

12.5

18.4

11.8

185

2.3

  Hodgkin disease

508

0.9

4.3

9.5

4.5

73

2.1

  NHL including Burkitt lymphoma

756

5.0

7.8

8.6

7.0

107

2.3

  Miscellaneous reticuloendothelial

10

0.1

0.1

0.1

0.1

1

9.0

  Unspecified

24

0.2

0.3

0.3

0.2

3

3.0

III Brain and spinal tumours

3309

33.2

33.4

26.4

31.3

465

1.1

  Ependymoma and choroid plexus

328

5.2

2.1

1.8

3.2

46

1.3

  Astrocytoma

1383

13.3

13.9

11.7

13.0

195

1.0

  Primitive neuroectodermal

665

7.5

7.1

4.0

6.4

93

1.5

  Other gliomas

410

3.0

5.2

3.4

3.8

58

1.0

  Other specified

337

2.3

3.5

3.8

3.1

48

1.2

  Unspecified

186

2.0

1.5

1.8

1.8

26

1.0

IV Sympathetic nervous system tumours

938

21.6

3.4

0.7

9.6

128

1.2

  Neuroblastoma

925

21.4

3.3

0.6

9.5

127

1.2

  Other

13

0.2

0.1

0.1

0.1

2

0.4

V Retinoblastoma

439

11.3

0.6

0.0

4.6

60

1.0

VI Renal tumours

793

16.6

4.0

1.1

8.1

109

1.0

  Wilms tumour etc

771

16.6

3.9

0.6

7.9

106

1.0

  Renal carcinoma

20

0.0

0.1

0.5

0.2

3

1.5

  Other

2

0.1

-

-

0.0

0

-

VII Hepatic tumours

128

2.5

0.5

0.5

1.3

18

2.0

  Hepatoblastoma

101

2.4

0.2

0.2

1.0

14

2.1

  Hepatic carcinoma

27

0.1

0.3

0.3

0.2

4

2.0

VIII Bone tumours

568

0.8

4.0

11.6

5.0

82

1.0

  Osteosarcoma

304

0.2

2.3

6.4

2.7

44

1.0

  Chondrosarcoma

13

0.0

0.1

0.3

0.1

2

2.3

  Ewing sarcoma

227

0.5

1.6

4.5

2.0

32

1.1

  Other specified

13

0.0

0.0

0.3

0.1

2

0.6

  Unspecified

11

0.1

0.1

0.1

0.1

2

0.6

IX Soft tissue sarcomas

1012

12.5

7.6

8.3

9.7

142

1.2

  Rhabdomyosarcoma

574

8.6

4.6

2.7

5.6

80

1.4

  Fibrosarcoma etc.

118

0.9

0.8

1.6

1.1

17

1.1

  Kaposi sarcoma

3

-

0.1

-

0.0

0

0.5

  Other specified

252

2.3

1.7

3.2

2.4

36

1.1

  Unspecified

65

0.7

0.4

0.8

0.6

9

1.3

X Germ cell and gonadal tumours

455

5.9

1.9

5.0

4.4

64

0.8

  CNS germ cell

133

0.9

1.0

1.9

1.2

19

1.5

  Other non-gonadal germ cell

113

2.8

0.1

0.2

1.2

15

0.3

  Gonadal germ cell

190

2.2

0.7

2.4

1.8

27

1.1

  Gonadal carcinoma

16

-

0.1

0.4

0.1

2

0.1

  Other gonadal

3

0.1

0.0

-

0.0

0

0.5

XI Epithelial tumours

458

1.2

2.8

9.3

4.1

66

0.8

  Adrenocortical carcinoma

15

0.2

0.1

0.1

0.1

2

0.2

  Thyroid carcinoma

55

0.1

0.3

1.3

0.5

8

0.4

  Nasopharynx carcinoma

25

-

0.1

0.6

0.2

4

2.1

  Melanoma

157

0.6

1.2

2.7

1.4

23

0.7

  Skin carcinoma

74

0.2

0.5

1.5

0.7

11

1.1

  Other carcinoma

132

0.2

0.6

3.1

1.2

19

0.9

XII Other and unspecified malignant neoplasms

93

1.2

0.6

0.8

0.9

13

0.6

  Other specified

15

0.2

0.1

0.1

0.1

2

0.5

  Other unspecified

78

1.0

0.5

0.7

0.7

11

0.7

NHL, non-Hodgkin lymphoma; CNS, central nervous system.
Source: National Registry of Childhood Tumours.

Brain and spinal tumours

In the USA, the incidence of brain and spinal tumours is lower among Black than among White children, while in Britain, children of South Asian, and perhaps especially Indian, ethnic origin also have a lower incidence. In developing countries the recorded incidence of brain and spinal tumours is often low, sometimes as little as 5 per million. It is unclear to what extent this reflects under-ascertainment, particularly in areas without neurological services, rather than a reduced underlying risk.

Neuroblastoma

The recorded incidence of neuroblastoma in several countries is much higher than in the UK, particularly in the first year of life, possibly because of increased detection of otherwise silent tumours during routine health checks. In Japan, and parts of some other countries, mass biochemical screening for neuroblastoma has led to particularly high incidence rates in infancy. In the USA, neuroblastoma has a lower incidence among Black than among White infants, but the rates are similar for the two ethnic groups at the age of 1 year and above. Recorded incidence in developing countries is often very low, but in some African registries it is similar to that among Black children in the USA. Thus it seems likely that there is little variation in underlying risk, and that recorded incidence reflects the proportions of tumours that are diagnosed and registered.

Retinoblastoma

Retinoblastoma occurs in two distinct forms, heritable and non-heritable. Heritable retinoblastoma includes all cases of bilateral retinoblastoma and a few children with unilateral tumours, and the incidence is relatively constant throughout the world. Non-heritable retinoblastoma is always unilateral and there are large variations in incidence, with substantially higher rates in many developing countries, particularly in sub-Saharan Africa.

Wilms tumour

Variations in the incidence of Wilms tumour depend largely on ethnic group rather than geographic area. Black children in the USA, the UK, and Africa have a higher incidence than White children, although their age distributions are similar. Children of East Asian ethnic origin in the USA and Asia have a lower incidence than White children, and the deficit is more marked after the first year of life.

Liver tumours

Hepatoblastoma has apparently constant incidence throughout the world. The incidence of hepatic carcinoma in children is highest in regions of the world where the disease is also common among adults, namely East and Southeast Asia, Melanesia, and sub-Saharan Africa. Nearly all childhood cases in these high-risk regions occur in chronic carriers of hepatitis B.

Sarcomas

Of the two principal types of childhood bone sarcoma, osteosarcoma appears to have a similar incidence in most populations, although it may be lower in Asia. In contrast, there are striking variations in the incidence of Ewing's sarcoma. This tumour has very low incidence in East and Southeast Asia, and among Black populations in Africa, the USA, and the UK.

Rhabdomyosarcoma is the most common soft tissue sarcoma of childhood in most populations. In much of South and East Asia, and also among Asian children in the UK, its incidence is rather less than among White children. Kaposi sarcoma is extremely rare in children in most regions of the world, but in East and Central Africa, where it is endemic among adults, the incidence among children was about 2 per million in about 1970. Since then, there have been very large increases, and rates of over 50 per million were recorded in the 1990s. The great majority of the increase is clearly related to the AIDS epidemic, which has been particularly severe in the region.

Other cancers

Germ cell tumours are somewhat more common in East Asia than in other regions of the world.

By far the highest incidence of childhood thyroid carcinoma, around 80 per million, has been recorded in areas of Belarus contaminated by radiation from the Chernobyl nuclear reactor explosion. Some extra cases may have been detected particularly early by intensive screening, and it is well known that thyroid cancer in other populations can be undetected for many years after onset. However, although the true size of the excess risk is hard to determine, its existence is not in doubt, both because the rates were 50 times greater than those seen elsewhere and because of the aggressive histologic type in many cases. Further evidence for an effect of shortlived radioactive fallout is provided by the return of incidence to normal levels among children conceived since the accident.

The highest incidence of nasopharyngeal carcinoma among children is in North Africa, a region of intermediate risk for adults, where it can account for up to 10 per cent of all childhood cancers. In East Asian countries, where incidence is highest among adults, children have only a moderately elevated incidence, up to 0.8 per million, while in the USA the incidence among Black children is about 1 per million, nine times that among Whites in the same country.

There is an exceptionally high incidence of skin carcinoma, both squamous cell and basal cell, among children in Tunisia, where the great majority of cases are associated with xeroderma pigmentosum. Malignant melanoma has a very high incidence in Australia and New Zealand, where it is also a very common cancer among adults, resulting from high levels of sun exposure.

Trends in incidence

The most striking increases in incidence of any childhood cancers, those relating to Kaposi sarcoma in East and Central Africa and to thyroid carcinoma in areas close to Chernobyl, have been described above. Otherwise, any changes have been much more modest. (Of course, the very large increases in recorded incidence of neuroblastoma in Japan and some other areas with population screening represent increased detection rather than a change in underlying risk.) The best documented trend concerns ALL, the most common childhood cancer in all developed countries. The early childhood peak started to emerge in mortality data in England and Wales in the 1920s, and it was certainly well established among White children in the USA by the early 1940s. A further modest rise, particularly in early childhood, took place in some Western countries until the late 1970s. Meanwhile, increasing incidence in early childhood among the Black population in the USA led to the emergence of a moderate peak in the age-incidence curve for this ethnic group. Small increases have also been observed in some Western populations for a number of other childhood cancers, notably brain tumours, neuroblastoma, and soft tissue sarcomas. However, it is not yet clear how much of these increases is attributable to changes in diagnostic practice rather than in underlying risk.

Clusters

A cluster of cases of a disease can be defined as the occurrence of a substantially larger number of cases than expected in a small geographically defined population, usually during a short period of time. There have been many reports of clusters of childhood leukaemia and other cancers. Among the most well known are those in the vicinity of some nuclear installations, although it is now generally accepted that they are unlikely to result from radiation exposure. Reported clusters need to be investigated carefully, the first stage being to check that the details of diagnosis, dates, and location are correct and the incidence rate is indeed high. Cluster investigations may yield clues to the aetiology of the particular cases or of the disease in general, although they are usually disappointing in this respect. It is also important to address the concerns of the local population, often with respect to the role of possible environmental pollution. In order to systematize the investigation of clusters, guidelines have been published in several countries, including the UK.7

Survival rates

In contrast with the very small improvements in prognosis that have occurred for many of the common adult cancers, the past 40 years have seen dramatic increases in the survival rates for most types of childhood cancer. Table 1.2 shows the 5-year survival rates for children in the UK diagnosed in successive quinquennia between 1962 and 1996 separately for the principal diagnostic groups. During this period, survival improved substantially for virtually all diagnostic groups, although the main advances for different types of cancer have occurred at different times. In the 1960s, there were particularly marked improvements in the prognosis for children with Hodgkin disease and Wilms tumour. The 1970s saw large improvements for a wide range of diagnoses, most notably ALL and NHL. In the 1980s, survival rates increased for children with ANLL, bone sarcomas, and germ cell tumours. By 1992–1996, 5-year survival exceeded 50 per cent for all the diagnostic groups shown. Analyses of data from the EUROCARE collaboration have shown substantial variation across Europe in survival rates for many childhood cancers.8 Survival tended to be highest in the Nordic countries and lowest in Eastern Europe. Unlike some adult cancers, survival from childhood cancer in Western Europe is similar to that in the USA.9

Table 1.2. Actuarial 5-year survival rates (per cent) for children diagnosed with the principal diagnostic groups in the UK in the period 1962–1996

 

1962–1966

1967–1971

1972–1976

1977–1981

1982–1986

1987–1991

1992–1996

ALL

4

17

44

56

70

75

81

ANLL

2

2

7

17

30

47

54

Hodgkin disease

39

68

81

89

89

93

94

NHL (including Burkitt lymphoma)

17

21

26

45

67

76

78

CNS

37

37

43

48

54

57

68

Neuroblastoma

18

17

19

31

43

41

53

Retinoblastoma

88

86

86

88

90

95

96

Wilms tumour

29

43

62

76

80

82

80

Osteosarcoma

17

18

22

25

47

51

57

Ewing sarcoma

25

23

40

34

45

68

61

Rhabdomyosarcoma

25

23

33

44

58

59

66

Gonadal germ cell

55

52

56

74

90

94

96

All cancers

24

29

42

51

62

67

73

Source: National Registry of Childhood Tumours.

The spectacular increases in survival rates described here are undoubtedly largely due to advances in treatment and supportive care. However, at the same time as these developments in therapy, there was also a substantial movement towards concentration of treatment at relatively few specialist centers, at some of which very large numbers of children are treated. Also, whereas at one time there were national clinical trials only for acute leukaemia and a handful of other diagnostic groups, in many countries there are now national or international trials and studies open for entry of children with all but the rarest types of cancer, and the proportions of eligible children that are entered in these trials have also increased. For several diagnostic groups, survival rates have been found to be higher among children who were treated at specialist centres or who were entered in clinical trials regardless of the treatment arm.10,11

Follow-up

As survival rates have improved for nearly all childhood cancers, the number of long-term survivors has greatly increased and these survivors already include a substantial number of adults. Several large studies of survivors from childhood cancer are now in progress, addressing such topics as the criteria for cure, quality of life, risk of second primary neoplasms, fertility, and the health of the survivors' offspring. Late effects and long-term follow-up are discussed from a clinical point of view in Chapter 11. Here, we summarize the epidemiologic data. Although late relapses can occur, the great majority of 5-year survivors can be regarded as cured; only one-tenth of them die of recurrent tumour or a treatment-related cause during the ensuing 10 years.12,13,14 The risk of developing a second primary neoplasm within 25 years of diagnosis of childhood cancer is about 4 per cent, four to six times the risk in the general population.15,16 However, the risks vary considerably between different types of childhood cancer, being especially high among survivors of heritable retinoblastoma. It should also be stressed that these risk estimates, at least for longer intervals following diagnosis, are based on survivors who were treated before the era of intensive chemotherapy, although of course many received radiotherapy. Also, little is yet known about the risks to survivors beyond the age of about 40, when the population incidence of several common cancers begins to rise markedly.

Some forms of treatment for cancer cause infertility, but many survivors go on to have children of their own. Most children of survivors are still very young. It is particularly important to follow up the offspring of survivors of cancer in childhood for two reasons. Some childhood cancers are known to have a predominantly genetic aetiology and the risk of transmission to future generations needs to be assessed. The question of whether treatmentrelated germ cell mutagenesis causes cancer, congenital malformations, or other genetic disease in the offspring of survivors also needs to be studied, and the risks, if any, estimated. Estimates of the risks to offspring of survivors are discussed in the section on genetic epidemiology below.

Aetiology of childhood cancer

Little is known about the causes of childhood cancer. International comparisons of the type described above, together with case–control and cohort studies, have suggested a variety of possible aetiologic factors, but these have seldom been convincing or replicated in further studies. On the other hand, studies of the genetics of childhood cancer have been much more rewarding, both in their practical application to problems of childhood cancer and in terms of basic science. The literature on aetiology up to 1997 was reviewed in detail by Little17 and more recent studies are covered by Stiller;18 these sources should be consulted for further references.

Environmental factors

Possible environmental causes of childhood cancer have been investigated in a large number of studies. The factors that have attracted most attention are radiation, both ionizing and nonionizing, and the role of infections in the child, the mother, and the community. The other factors studied include parental characteristics, environmental exposures of both child and parents, particularly in utero exposures, and occupational exposures of both parents. Concern about possible environmental causes of genetic damage has led to studies of preconception exposures of the parents.

Ionizing radiation

Probably the first, and still the largest, case–control study of childhood cancer was that carried out by Dr Alice Stewart and her colleagues. An early report from this study claimed that antenatal radiography, used mainly in the third trimester of pregnancy for obstetric reasons, significantly increased the probability of the subsequent child developing cancer. Although this result was initially greeted with considerable scepticism, it is now widely accepted as correct; the most useful presentation of the results is that by Bithell and Stewart.19 For children born between about 1945 and 1965 there was a 40–50 per cent increase in the risk of childhood cancer if the mother was subjected to abdominal radiography about two or three times during pregnancy. It should be emphasized that antenatal radiography probably caused at most about 5 per cent of childhood cancers, even when it was more widely used and doses were higher than they are now. Both the number of pregnant women subjected to radiography and the doses of radiation have greatly decreased since the time of Stewart's work; indeed, radiography has largely been replaced by ultrasound. There is no evidence that ultrasound causes childhood cancer.

Other childhood exposures to radiation, for example that used in treating tinea capitis in early childhood or, particularly, the large doses given during radiotherapy, are also carcinogenic. The possibility that some childhood cancers are caused by natural radiation (gamma, radon) has been suggested. Case–control studies have shown no significant association with radon exposure. On the usual assumption of a linear dose–response relationship for the carcinogenic effects of radiation, one would expect some cases to be caused by background gamma radiation. However, in the only case–control study based on individual household measurements, there was no indication of increased risk of childhood cancer with increasing dose levels.20

The report that caused the most public concern and scientific interest in this area was probably that by Gardner et al.,21 which suggested that paternal preconception exposure to ionizing radiation could lead to childhood leukaemia. Because of the relatively small doses involved (from employment in the nuclear industry) doubts were raised about the validity of this finding, although the study appeared to be epidemiologically sound. Subsequent reports, in particular two large overlapping studies of UK radiation workers including those employed Sellafield and elsewhere,22,23 were reviewed by the Committee on Medical Aspects of Radiation in the Environment.24 This Committee suggested that there was unlikely to be any simple causal relationship.

Non-ionizing radiation

Ultraviolet radiation from sunlight is known to cause melanoma and other skin cancers. Concern has been expressed for more than 20 years over possible carcinogenic effects of electromagnetic fields arising from electric power transmission and use, from both power lines and domestic exposure. However, despite a very large research effort, the question of whether such exposures do have carcinogenic effects is unresolved. Analysis of pooled data from case–control studies has shown no evidence of raised risk of childhood leukaemia with exposure to power frequency fields at the level experienced by more than 95 per cent of children in Western countries.25 A doubling of risk has been found at the very highest exposure levels but the explanation is unknown, although it may be partly due to selection bias; only a tiny fraction of childhood leukaemia cases could be attributable to electromagnetic fields.

Parental occupation and socio-economic status

Any association observed between childhood disease and parental occupations may reflect either specific exposures or lifestyle factors associated with particular occupations. Furthermore, a parent may well be in the same occupation before the conception of the child, during pregnancy, and subsequently; thus it may be difficult to determine whether any observed association is attributable to exposure occurring before conception, in utero, or postnatally.

Many papers analysing occupation in general or specific occupations, using a variety of study designs, have been published. It is inevitable that in a large series of studies, some of which include analysis of many different occupations, a number of statistically significant results will be reported. In such circumstances it is necessary to determine whether there is consistency between different studies and whether there is any evidence of a dose–response effect, i.e. in this instance, whether the risk is higher for children of parents who have been more heavily exposed to the suspected occupational mutagen or who have worked longer in the occupation. As judged by these criteria, there is as yet no sufficient evidence that any parental occupation is causally associated with childhood cancer. However, in a review of 48 published studies some biologically plausible associations were found in more than one study and these, in particular, merit further investigation.26

Of course, occupational category is closely related to socio-economic status. Thus, if there is an effect of lifestyle or standards of living on childhood cancer, this could appear as an effect of an occupation which happens to be in the high-risk group as defined by socio-economic status. The possibility of such an association with socio-economic status has been investigated in both case–control studies and ‘ecologic’ studies, i.e. by comparing incidence rates for areas classified according to the socio-economic status of their resident populations. There is good evidence that the incidence of childhood leukaemia increases with increasing socio-economic status, but the reasons for this are unknown.

Infections

The possible role of infections, especially viruses, in the aetiology of childhood cancer has been studied in a number of ways. Viruses are known to be implicated in some human cancers. Worldwide, the most important numerically among children are Burkitt lymphoma, Hodgkin disease, and nasopharyngeal carcinoma (Epstein–Barr virus), liver carcinoma (hepatitis B), and Kaposi sarcoma (HIV and HHV8).27Case–control studies have included analyses of exposures to infectious illnesses of both the children themselves and their mothers while pregnant. Positive findings have been reported, but there is no conclusive evidence from these studies. Perhaps the most persuasive evidence for the involvement of an infectious agent in childhood leukaemia comes from studies of clustering and incidence rates in different geographic areas.

Kinlen,28 in a remarkable series of studies, has shown that childhood leukaemia rates have increased in a variety of different situations in which people from several different areas come together, resulting in ‘population mixing’. His explanation is that such situations are conducive to the spread of an infectious agent or agents among a previously unexposed, and therefore susceptible, group, and that childhood leukaemia may be represent an unusual response to such agents.28 It is not suggested that leukaemia itself is an infectious disease. Under the ‘delayed infection’ hypothesis, children who are protected from exposure to infection in infancy have an abnormal immune response to infection subsequently, and this occasionally results in leukaemia.29 This is supported by, among other epidemiologic findings, a deficit of common infections and social contact in infancy among children with ALL and a raised risk of ALL in first-born children.

Other possible aetiologic factors

A wide variety of other possible factors has been studied. The cell types involved in childhood cancers and the shape of the age distribution, with a peak for the typical cancers of childhood at a very early age, have led many investigators to concentrate on events occurring during pregnancy or even before conception.

Preconception factors

The fact that some childhood cancers are attributable to germ cell mutations has led many investigators to study parental exposure to possible mutagens. Such studies have included various types of exposure to ionizing radiation (including the occupational exposures mentioned above) and to chemicals. Again, no consensus has emerged.

In utero exposures

The effects of in utero ionizing radiation have already been referred to. It is well known that diethylstilboestrol (DES) given to pregnant women to avert threatened miscarriage can cause vaginal adenocarcinoma in their daughters, although these occur mainly in young women rather than children. DES ceased to be used over 30 years ago and, in the absence of evidence for any transgenerational effect, no more DES-associated cases of childhood cancer are expected to arise. A number of positive findings relating to other drugs given to women during pregnancy have been published, but no firm causal associations have been established. Similarly, smoking in pregnancy has been widely studied but is not established as a risk factor for any specific childhood cancer.30

Postnatal exposures

Ionizing radiation and infections have already been discussed. Many other environmental factors have been studied but, again, there is no consensus. The suggestion by Golding et al31 that intramuscular vitamin K, given to neonates to prevent vitamin K deficiency bleeding, doubles the risk of childhood cancer led to a series of further studies. Analysis of pooled data from six case–control studies found little evidence for a raised risk of childhood cancer with intramuscular vitamin K,32 although interpretation was hampered by the poor quality of much of the vitamin K data. Other medications have occasionally been reported as risk factors and, of course, cytotoxic drugs given in the course of treatment for cancer are known causes of further cancers, but there is no evidence to suggest that any substantial numbers of childhood cancers are attributable to these or other drugs.

Genetic epidemiology

A detailed discussion of the genetic aspects of childhood cancer is given in Chapter 2. In the present chapter we concentrate on the estimation of risks of childhood cancer for family members of affected children and for individuals who have genetic conditions known to predispose to childhood cancer.

The most obvious example of a genetically determined cancer is retinoblastoma. About 40 per cent of cases have the heritable form of this disease. The pattern of inheritance is that of a dominant autosomal gene with about 90 per cent penetrance, but in fact the gene Rb1 is the first example of a tumour suppressor gene; about 90 per cent of individuals who inherit the mutated form of this gene from a parent subsequently suffer a mutation of the wild-type (normal) allele, leading to loss of heterozygosity and the development of retinoblastoma. These individuals are now also known to be at increased risk of a variety of other cancers.

There is an obvious genetic element in some other childhood cancers, notably Wilms tumour, but the pattern of inheritance in Wilms tumour is a great deal more complicated and the proportion of clearly hereditary cases is much smaller.

A variety of childhood cancers and some adult cancers, notable premenopausal breast cancer, are observed in the Li-Fraumeni familial cancer syndrome.33 The risk of childhood cancer is roughly 20 times that in the general population. Germ-line mutations of the TP53 tumour suppressor gene are responsible for the high risk and distinctive pattern of cancer in many Li-Fraumeni families, but other genes have been implicated in some families apparently without TP53 mutations.

Some familial aggregations arise through the association of childhood and adult tumours with known genetic disease such as neurofibromatosis, tuberous sclerosis, Fanconi's anaemia, ataxia telangiectasia, xeroderma pigmentosum, and Bloom's syndrome, although the actual number of cases of childhood cancer in which these conditions occur is rather small.34 There are also well-documented associations with congenital abnormalities; the strongest association with such a condition is Down syndrome, which occurs in a small percentage of cases of childhood leukaemia.

Relatives of affected children

Various authors have studied the siblings, twins, offspring, and, to a lesser extent, the parents of children with cancer.

Siblings and twins

If one child in a family has malignant disease, then, in the absence of any further information about the existence of genetic disease in that family, and excluding twins and retinoblastoma, the siblings of that child have approximately double the risk of the general population, i.e. a risk of approximately 1 in 250 compared with the average risk of about 1 in 500.35 (NB. The estimate for the population risk is larger than that given previously.) As more has become known about the genetic element in childhood cancer, the risk estimates have increased for families where familial syndromes have been identified, while those for the remaining families have decreased. In a more recent study, the excess of cancer among siblings of children and adolescents with cancer could be entirely accounted for by familial syndromes.36 However, in families where there is one child with cancer, a cancer syndrome may not be recognized as such until after a further sibling has been diagnosed with cancer. In providing genetic counselling in the absence of such additional information, i.e. where there is at that time no indication of a cancer syndrome in that particular family, it should be assumed that the increased risk for siblings referred to above applies. A doubling of the risks for siblings compared with the general population implies that half of the families with two affected children are in fact due to chance. (Incidentally, this implies that laboratory studies of these families will be expected to find nothing of interest at least 50 per cent of the time.) It should be emphasized that the risk is less than 1 in 250 for siblings who are a few years old when the affected child is diagnosed. The estimated risk is also lower if there are other children in the family who are not affected.

The risks are higher if there are two affected children in the family and where there is known genetic disease of a type associated with cancer. In the special case of retinoblastoma, the risk for a subsequent sibling following an apparently sporadic case of retinoblastoma, i.e. where there is no previous family history, appears to be about 2 per cent if the disease in the affected sibling is bilateral, and 1 per cent if it is unilateral.37 This is lower than has previously been suggested, and is lower still if there are other siblings who are not affected.

The risk that the co-twin of a twin with cancer will also be affected is of particular concern, and more so if the twins are monozygous. In general, both twins and childhood cancer are too rare for any quantitative estimates of risk to be made. It seems likely that the risk for a dizygotic co-twin of an affected case will be at least as high as that for ordinary siblings, but no data are available. Nearly all the published cases of childhood cancer in twins are like-sexed pairs and are known or assumed to be monozygous. The fact that in case reports of affected twin pairs the two children almost invariably have the same diagnosis and tend to be diagnosed at the same age is an indication that such cases may be genetic in origin. One would expect there to be an increased risk for the monozygous co-twins of affected cases, but in general there are insufficient data to estimate such risks; the fact that the co-twins of the great majority of cases do not develop cancer38 implies that the risk to co-twins is, in general, not very high. The exception to this is childhood leukaemia, where perhaps 25 per cent of co-twins of monozygotic cases also develop the disease(see also Chapter 2). However, many of these cases are due to in utero transfer of leukaemia cells rather than being genetically determined.

Risks to offspring and parents

The risks to offspring of childhood cancer survivors correspond well to those from the studies of siblings. In general they are low, the main exception being for retinoblastoma, although even here, as for siblings, the risks in cases of sporadic retinoblastoma appear to be lower than previously suggested: for a child of a unilateral case, and where it is not known whether the disease is of hereditary type, the estimated risk is 1 per cent.37 After the exclusion of hereditary cancer syndromes there is no evidence of a significantly raised risk of cancer among the offspring of survivors.39 Similarly, there is no evidence that the parents of children with cancer have an increased risk of cancer in the absence of known hereditary syndromes.40

References

1. Kramárová E, Stiller CA, Ferlay J, et al. (1996). The International Classification of Childhood Cancer. IARC Technical Report No. 29. Lyon: IARC.

2. Jensen OM, Parkin DM, MacLennan R, Muir CS, Skeet RG (eds) (1991). Cancer Registration: Principles and Methods. IARC Scientific Publications No.95. Lyon: IARC.

3. Parkin DM, Kramárová E, Draper GJ, et al. (eds) (1998). International Incidence of Childhood Cancer, Volume 2. IARC Scientific Publications No 144. Lyon: IARC.

4. Tyczynski JE, Démaret E, Parkin DM (eds) (2003). Tyczynski JE, Démaret E, Parkin DM, editors. Standards and Guidelines for Cancer Registration in Europe. The ENCR recommendations. Vol 1. IARC Technical Report No.40. Lyon: IARC.

5. Stiller CA, Parkin DM (1996). Geographic and ethnic variations in the incidence of childhood cancer. Br Med Bull 52, 682–703.

6. Hrusák O, Trka J, Zuna J, Poloucková A, Kalina T, Stary J (2002). Acute lymphoblastic leukemia incidence during socioeconomic transition: selective increase in children from 1 to 4 years. Leukemia16, 720–5.

7. Arrundale J, Bain M, Botting B, et al. (1997). Handbook and Guide to the Investigation of Clusters of Diseases. Leeds: Leukaemia Research Fund Centre for Clinical Epidemiology, University of Leeds

8. Gatta G, Corazziari I, Magnani C, et al. (2003). Childhood cancer survival in Europe. Ann Oncol14 (Suppl 5), 119–27.

9. Gatta G, Capocaccia R, Coleman MP, Gloeckler Ries LA, Berrino F (2002). Childhood cancer survival in Europe and the United States. Cancer 95, 1767–72.

10. Stiller CA (1994). Centralised treatment, entry to trials and survival. Br J Cancer 70, 352–62.

11. Stiller CA, Eatock EM (1999). Patterns of care and survival for children with acute lymphoblastic leukaemia diagnosed between 1980–94. Arch Dis Child 81, 202–8.

12. Robertson CM, Hawkins MM, Kingston JE (1994). Late deaths and survival after childhood cancer: implications for cure. BMJ 309, 162–6.

13. Möller TR, Garwicz S, Barlow L, et al. (2001). Decreasing late mortality among five-year survivors of cancer in childhood and adolescence: a population-based study in the Nordic countries. J Clin Oncol 19, 3173–81.

14. Mertens AC, Yasui Y, Neglia JP, et al. (2001). Late mortality experience in five-year survivors of childhood and adolescent cancer: the Childhood Cancer Survivor Study. J Clin Oncol 19, 3163–72.

15. Hawkins MM, Stevens MCG (1996). The long term survivors. Br Med Bull 52, 898–923.

16. Neglia JP, Friedman DL, Yasui Y, et al. (2001). Second malignant neoplasms in five-year survivors of childhood cancer: childhood cancer survivor study. J Natl Cancer Inst 93, 618–29.

17. Little J (1999). Epidemiology of Childhood Cancer. IARC Scientific Publications No 149. Lyon: IARC.

18. Stiller CA. (2004). Aetiology and epidemiology. In: Pinkerton CR, Plowman PN, Pieters R (eds) Paediatric Oncology (3rd edn). pp 3–24 London: Arnold.

19. Bithell JF, Stewart AM (1975). Pre-natal irradiation and childhood malignancy: A review of British data from the Oxford survey. Br J Cancer 31, 271–87.

20. UK Childhood Cancer Study Investigators (2002). The United Kingdom Childhood Cancer Study of exposure to domestic sources of ionising radiation. 2: Gamma radiation. Br J Cancer 86, 1727–31.

21. Gardner MJ, Snee MP, Hall AJ, Powell CA, Downes S, Terrell JD (1990). Results of case–control study of leukaemia and lymphoma among young people near Sellafield nuclear plant in West Cumbria. BMJ 300, 423–9.

22. Draper GJ, Little MP, Sorahan T, et al. (1997). Cancer in the offspring of radiation workers: a record linkage study. BMJ 315, 1181–8.

23. Roman E, Doyle P, Maconochie N, Davies G, Smith PG, Beral V (1999). Cancer in children of nuclear industry employees: report on children aged under 25 years from nuclear industry family study. BMJ 318, 1443–50.

24. Committee on Medical Aspects of Radiation in the Environment (COMARE) (2002). Seventh Report: Parents Occupationally Exposed to Radiation Prior to the Conception of their Children. A Review of the Evidence Concerning the Incidence of Cancer in their Children. Chilton, Didcot, UK: NRPB

25. Ahlbom A, Day N, Feychting M, et al. (2000). A pooled analysis of magnetic fields and childhood leukaemia. Br J Cancer 83, 692–8.

26. Colt JS, Blair A (1998). Parental occupational exposures and risk of childhood cancer. Environ Health Perspect 106, 909–25.

27. Parkin DM, Pisani P, Mu[not tilde]oz N, Ferlay J (1999). The global health burden of infection associated cancers. Cancer Surv 33, 5–33.

28. Kinlen LJ (1995). Epidemiological evidence for an infective basis in childhood leukaemia. Br J Cancer 71, 1–5.

29. Greaves M (2002). Childhood leukaemia. BMJ 324, 283–7.

30. Boffetta P, Trédaniel J, Greco A (2000). Risk of childhood cancer and adult lung cancer after childhood exposure to passive smoking: a meta-analysis. Environ Health Perspect 108, 73–82.

31. Golding J, Greenwood R, Birmingham K, Mott M (1992). Childhood cancer, intramuscular vitamin K, and pethidine given during labour. BMJ 305, 341–6.

32. Roman E, Fear NT, Ansell P, et al. (2002). Vitamin K and childhood cancer: analysis of individual patient data from six case–control studies. Br J Cancer 86, 63–9.

33. Varley JM (2003). Germline TP53 mutations and Li-Fraumeni syndrome. Hum Mutat 21, 313–20.

34. Lindor NM, Greene MH, The Mayo Familial Cancer Program (1998). The concise handbook of family cancer syndromes. J Natl Cancer Inst 90, 1039–71.

35. Draper GJ, Heaf MM, Kinnier Wilson LM (1977). Occurrence of childhood cancers among sibs and estimation of familial risks. J Med Genet 14, 81–90.

36. Winther JF, Sankila R, Boice JD, et al. (2001). Cancer in siblings of children with cancer in the Nordic countries: a population-based cohort study. Lancet 358, 711–17.

37. Draper GJ, Sanders BM, Brownbill PA, Hawkins MM (1992). Patterns of risk of hereditary retinoblastoma and applications to genetic counselling. Br J Cancer 66, 211–19.

38. Draper GJ, Sanders BM, Lennox EL, Brownbill PA (1996). Patterns of childhood cancer among siblings. Br J Cancer 74, 152–8.

39. Sankila R, Olsen JH, Anderson H, et al. (1998). Risk of cancer among offspring of childhood-cancer survivors. N Engl J Med 338, 1339–44.

40. Olsen JH, Boice JD, Seersholm N, Bautz A, Fraumeni JF (1995). Cancer in the parents of children with cancer. N Engl J Med 333, 1594–9.


Previous
Page
Next
Page

Contents


If you find an error or have any questions, please email us at admin@doctorlib.info. Thank you!