American Journal of Epidemiology Copyright © 1992 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved
Vol. 136, No. 8 Printed in U.S.A.
Is Acute Lymphoblastic Leukemia in Children Virus-Related?
Brian MacMahon1
The sharp peak in incidence of acute lymphoblastic leukemia at ages 2 and 3 years strongly suggests the effect of an agent, whether viral or not, to which either exposure occurs only in the earliest months of life or to which immunity develops very rapidly. Suspected clusters of childhood leukemia in the neighborhoods of two British nuclear reprocessing facilities led Leo Kinlen to postulate that large-scale immigration into areas that had previously been remote and isolated offers opportunities for spread of viral infections to which most urban populations become immune at a very early age: leukemia may be a rare manifestation of infection by one or more of such viruses. He and his group have presented evidence in support of this hypothesis. Lack of increase in childhood leukemia in the contexts of the massive evacuation of mothers and children from British cities during the Second World War, and of the considerable immigration into previously isolated islands of Greece during the last several decades, indicates that some large movements of children have occurred without providing the circumstances postulated by Kinlen. The marked inverse association of leukemia risk with birth order, noted almost 30 years ago, remains unexplained and deserves to be recalled when considering the possibility of viral involvement in the etiology of acute lymphoblastic leukemia of children. Am J Epidemiol 1992;136:916-24. age; birth order; child; infection; leukemia; leukemia, lymphoblastic; viruses
Acute lymphoblastic leukemia is the most common malignancy of childhood, about 80 percent of leukemia in children being, by today's nomenclature, lymphoblastic. The cumulative incidence of the disease up to the age of 15 years is, in Europe and North America, about 1 in 2,000. As shown in figure 1, the disease shows a remarkable peak of incidence in the second and third years of life (1). It is slightly more frequent in males than in females, but both sexes show this early peak. The age pattern for acute
non-lymphocytic leukemia in childhood is quite different (figure 1). There is probably some association of acute lymphoblastic leukemia risk with high socioeconomic status, but it is weak and not free from the possibility of ascertainment or reporting error. Apart from the dramatic age-pattern, there is little remarkable about the descriptive epidemiology of this disease. In 1963, Heath and Hasterlik (2) reported an apparent cluster of eight cases of childhood leukemia during the 4 years, 1957— 1960, in the town of Niles, Illinois. Clusters of leukemia had been reported periodically in the literature for decades, but the Niles report focused considerable attention on the question of whether cases of leukemia, and cases of leukemia in children in particular, tend to come in clusters—a trait characteristic of many diseases caused by microbiologic agents. Extensive investigation of the Niles cluster did not lead to an explanation of it. The literature of the last three decades,
Received for publication July 27, 1992. 1 Department of Epidemiology, Harvard School of Public Health, Boston, MA. Reprint requests to Dr. Brian MacMahon, Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115. This paper is based on an Invited Lecture presented at the 25th Annual Meeting of the Society for Epidemiologic Research, Minneapolis, MN, June 9-12, 1992. The author is indebted to Prof. Dimitrios Trichopoulos for pointing out the potential relevance of the information on leukemia risk and birth order.
916
Is Acute Lymphoblastic Leukemia in Children Virus-Related?
917
14 FIGURE 1. Annual incidence of leukemia by age, sex, and cell type, Great Britain, 1969-1983. Data are from Draper (1). (ALL, acute lymphoblastic leukemia; ANLL, acute non-lymphocytic leukemia).
exploring the broad question of leukemia clustering, has been mathematically complex and has added more to the philosophic and statistical discourse than it has to our understanding of the etiology of acute lymphoblastic leukemia. My own reading—for example, of the review by Smith (3)—leads me to conclude that clustering is not a feature of leukemia, or even of leukemia in children, but, nevertheless, the possibility that there is some low level clustering of this disease in young children cannot be totally excluded.
LEUKEMIA CLUSTERING AROUND NUCLEAR INSTALLATIONS The more recent history of the possible microbiologic origin of acute lymphoblastic leukemia began, somewhat ironically, with
reports of clusters of childhood leukemia in the vicinity of the two nuclear reprocessing facilities in Britain—one a cluster of seven cases extending over a period of three decades in the village of Seascale, near the nuclear operations at Sellafield in Cumbria, and one involving five cases over a period of 5 years, mostly in the town of Thurso near the facilities at Dounreay on the north coast of Scotland. I will not here comment on the validity of these alleged clusters. It is sufficient for the present purpose to note that, in 1987, Darby and Doll (4), in considering possible explanations for them suggested that: Viral infection has often been suspected as the most likely principal cause (of the clusters); ... If an infective agent is responsible for the disease it is possible that mini epidemics would be particularly likely under
918
MacMahon
the conditions that have brought together a young workforce in a comparatively isolated area (4, p. 607, material in parentheses added).
the first half (1951-1967) of the total period for which data were available (1951-1985), because after the opening of the road bridge across the Firth of Forth in 1964 the isolation of Glenrothes ceased (5). The data for leukemia mortality in Glenrothes are shown in table 1. In Glenrothes in the years 1951-1967, there is indeed the predicted excess of leukemia (observed/expected = 2.8, p < 0.01), which is limited to the age group 0-4 years (observed/expected = 4.7, p < 0.001). In the second time period, there is a notable and statistically significant deficit of leukemia. Kinlen suggested that this deficit might be due to exhaustion of the susceptible population by the high rates in the earlier time period. This explanation seems unlikely, since the children in the high risk ages in 1968-1985 belonged to a different cohort than those who were around in the earlier period and experienced the high rates. There is another peculiarity in the Glenrothes situation. This is that the excess in leukemia deaths in the hypothesized time period was predominantly a feature of the age group 0-4 years. This was not true of the cases forming the clusters around either of the nuclear reprocessing facilities. At Seascale, only three of the seven childhood cases were under age 5 years, and in the Dounreay cluster only two of the five cases were under age 5. It seems therefore that the hypothesis stimulated by the nuclear site clusters related to a different category of
Leo Kinlen (5) also noted in 1988 that the construction of the facilities at Sellafield and Dounreay had led to large-scale migration into areas that had previously been remote and isolated—circumstances that offered the opportunity for the spread of viral infections to which most urban populations would have become immune through exposure when they were very young. He postulated that leukemia may be a rare manifestation of infection by one or more of these viruses, for which there are a number of wellcharacterized animal models, the best known of which is in the cat, in which the disease is not only virus-related but infectious (6, 7), as well as the rare human adult T-cell leukemia associated with human Tcell leukemia virus type I (8). Kinlen further offered a test of this hypothesis.
GLENROTHES After reviewing Scottish local authority districts, Kinlen found only one with a large influx in population and other characteristics that he considered similar to those of Seascale and Thurso. This was the District of Kirkcaldy, containing the New Town of Glenrothes. Glenrothes was hypothesized to be similar to Seascale and Thurso only in
TABLE 1. Deaths from leukemia in persons less than 25 years of age in Glenrothes, Scotland, 1951-1985 (data from Kinlen (5)) Yesrs
Age (years)
Leukemia deaths Observed
Expected
Observed/ expected
1951-1967
0-4 5-14 15-24 Total (0-24)
7* 1 2 10*
1.5 1.4 0.7 3.6
4.7* 0.7 2.9 2.8*
1968-1985
0-4 5-14 15-24 Total (0-24)
0 1 0 1
1.3 2.2 1.7 5.2
0 0.5 0 0.2
•p
Vol. 136, No. 8 Printed in U.S.A.
Is Acute Lymphoblastic Leukemia in Children Virus-Related?
Brian MacMahon1
The sharp peak in incidence of acute lymphoblastic leukemia at ages 2 and 3 years strongly suggests the effect of an agent, whether viral or not, to which either exposure occurs only in the earliest months of life or to which immunity develops very rapidly. Suspected clusters of childhood leukemia in the neighborhoods of two British nuclear reprocessing facilities led Leo Kinlen to postulate that large-scale immigration into areas that had previously been remote and isolated offers opportunities for spread of viral infections to which most urban populations become immune at a very early age: leukemia may be a rare manifestation of infection by one or more of such viruses. He and his group have presented evidence in support of this hypothesis. Lack of increase in childhood leukemia in the contexts of the massive evacuation of mothers and children from British cities during the Second World War, and of the considerable immigration into previously isolated islands of Greece during the last several decades, indicates that some large movements of children have occurred without providing the circumstances postulated by Kinlen. The marked inverse association of leukemia risk with birth order, noted almost 30 years ago, remains unexplained and deserves to be recalled when considering the possibility of viral involvement in the etiology of acute lymphoblastic leukemia of children. Am J Epidemiol 1992;136:916-24. age; birth order; child; infection; leukemia; leukemia, lymphoblastic; viruses
Acute lymphoblastic leukemia is the most common malignancy of childhood, about 80 percent of leukemia in children being, by today's nomenclature, lymphoblastic. The cumulative incidence of the disease up to the age of 15 years is, in Europe and North America, about 1 in 2,000. As shown in figure 1, the disease shows a remarkable peak of incidence in the second and third years of life (1). It is slightly more frequent in males than in females, but both sexes show this early peak. The age pattern for acute
non-lymphocytic leukemia in childhood is quite different (figure 1). There is probably some association of acute lymphoblastic leukemia risk with high socioeconomic status, but it is weak and not free from the possibility of ascertainment or reporting error. Apart from the dramatic age-pattern, there is little remarkable about the descriptive epidemiology of this disease. In 1963, Heath and Hasterlik (2) reported an apparent cluster of eight cases of childhood leukemia during the 4 years, 1957— 1960, in the town of Niles, Illinois. Clusters of leukemia had been reported periodically in the literature for decades, but the Niles report focused considerable attention on the question of whether cases of leukemia, and cases of leukemia in children in particular, tend to come in clusters—a trait characteristic of many diseases caused by microbiologic agents. Extensive investigation of the Niles cluster did not lead to an explanation of it. The literature of the last three decades,
Received for publication July 27, 1992. 1 Department of Epidemiology, Harvard School of Public Health, Boston, MA. Reprint requests to Dr. Brian MacMahon, Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115. This paper is based on an Invited Lecture presented at the 25th Annual Meeting of the Society for Epidemiologic Research, Minneapolis, MN, June 9-12, 1992. The author is indebted to Prof. Dimitrios Trichopoulos for pointing out the potential relevance of the information on leukemia risk and birth order.
916
Is Acute Lymphoblastic Leukemia in Children Virus-Related?
917
14 FIGURE 1. Annual incidence of leukemia by age, sex, and cell type, Great Britain, 1969-1983. Data are from Draper (1). (ALL, acute lymphoblastic leukemia; ANLL, acute non-lymphocytic leukemia).
exploring the broad question of leukemia clustering, has been mathematically complex and has added more to the philosophic and statistical discourse than it has to our understanding of the etiology of acute lymphoblastic leukemia. My own reading—for example, of the review by Smith (3)—leads me to conclude that clustering is not a feature of leukemia, or even of leukemia in children, but, nevertheless, the possibility that there is some low level clustering of this disease in young children cannot be totally excluded.
LEUKEMIA CLUSTERING AROUND NUCLEAR INSTALLATIONS The more recent history of the possible microbiologic origin of acute lymphoblastic leukemia began, somewhat ironically, with
reports of clusters of childhood leukemia in the vicinity of the two nuclear reprocessing facilities in Britain—one a cluster of seven cases extending over a period of three decades in the village of Seascale, near the nuclear operations at Sellafield in Cumbria, and one involving five cases over a period of 5 years, mostly in the town of Thurso near the facilities at Dounreay on the north coast of Scotland. I will not here comment on the validity of these alleged clusters. It is sufficient for the present purpose to note that, in 1987, Darby and Doll (4), in considering possible explanations for them suggested that: Viral infection has often been suspected as the most likely principal cause (of the clusters); ... If an infective agent is responsible for the disease it is possible that mini epidemics would be particularly likely under
918
MacMahon
the conditions that have brought together a young workforce in a comparatively isolated area (4, p. 607, material in parentheses added).
the first half (1951-1967) of the total period for which data were available (1951-1985), because after the opening of the road bridge across the Firth of Forth in 1964 the isolation of Glenrothes ceased (5). The data for leukemia mortality in Glenrothes are shown in table 1. In Glenrothes in the years 1951-1967, there is indeed the predicted excess of leukemia (observed/expected = 2.8, p < 0.01), which is limited to the age group 0-4 years (observed/expected = 4.7, p < 0.001). In the second time period, there is a notable and statistically significant deficit of leukemia. Kinlen suggested that this deficit might be due to exhaustion of the susceptible population by the high rates in the earlier time period. This explanation seems unlikely, since the children in the high risk ages in 1968-1985 belonged to a different cohort than those who were around in the earlier period and experienced the high rates. There is another peculiarity in the Glenrothes situation. This is that the excess in leukemia deaths in the hypothesized time period was predominantly a feature of the age group 0-4 years. This was not true of the cases forming the clusters around either of the nuclear reprocessing facilities. At Seascale, only three of the seven childhood cases were under age 5 years, and in the Dounreay cluster only two of the five cases were under age 5. It seems therefore that the hypothesis stimulated by the nuclear site clusters related to a different category of
Leo Kinlen (5) also noted in 1988 that the construction of the facilities at Sellafield and Dounreay had led to large-scale migration into areas that had previously been remote and isolated—circumstances that offered the opportunity for the spread of viral infections to which most urban populations would have become immune through exposure when they were very young. He postulated that leukemia may be a rare manifestation of infection by one or more of these viruses, for which there are a number of wellcharacterized animal models, the best known of which is in the cat, in which the disease is not only virus-related but infectious (6, 7), as well as the rare human adult T-cell leukemia associated with human Tcell leukemia virus type I (8). Kinlen further offered a test of this hypothesis.
GLENROTHES After reviewing Scottish local authority districts, Kinlen found only one with a large influx in population and other characteristics that he considered similar to those of Seascale and Thurso. This was the District of Kirkcaldy, containing the New Town of Glenrothes. Glenrothes was hypothesized to be similar to Seascale and Thurso only in
TABLE 1. Deaths from leukemia in persons less than 25 years of age in Glenrothes, Scotland, 1951-1985 (data from Kinlen (5)) Yesrs
Age (years)
Leukemia deaths Observed
Expected
Observed/ expected
1951-1967
0-4 5-14 15-24 Total (0-24)
7* 1 2 10*
1.5 1.4 0.7 3.6
4.7* 0.7 2.9 2.8*
1968-1985
0-4 5-14 15-24 Total (0-24)
0 1 0 1
1.3 2.2 1.7 5.2
0 0.5 0 0.2
•p
