SEVEN CASES of childhood acute lymphoblastic leukemia (cALL) occurred between 1983 and 1985 among children who lived in Carbonia, a town populated by 32 000 residents and located in southwestern Sardinia (Italy). Headlines in the local press and television broadcasts raised public concern about environmental pollution being the cause of the observed cases of cancer among children. This concern was precipitated by the presence of a large industrial complex on the island, located 10 km northwest of downtown Carbonia.
The analysis of cALL risk, by communes in the Cagliari province, showed that a significant increase existed only in the town of Carbonia. The excess risk was significant statistically during 1983-1985 (i.e., 7 cases observed versus 0.82 expected); two cases were observed versus 0.63 expected during 1986-1988. No cases of cALL occurred in Carbonia between 1989 and 1994; this conclusion was based on a survey conducted by the childhood oncohematology department at the University of Cagliari, on personal communications from the parents of childhood cancer patients (organized in an association), and on personal communications from employees of the municipality of Carbonia (1991 forward). Expected new cases of cALL in the same period were 1.17, estimated by application of the 1986-1989 5-y age-group- and gender-specific provincial rates to the respective strata of the Carbonia resident population. No birth cohort effect was observed. The incidence rate of cALL was significantly higher among children who were born and who resided in Carbonia, compared with children who had been born in Carbonia but who resided elsewhere. However, residences of cases of cALL were not clustered within the town of Carbonia.
We attempted to explore whether the cALL cases that generated the cluster shared common risk factors, and a case-control investigation was conducted among families who resided in Carbonia.
Material and Method
Childhood cancer cases, diagnosed between 1974 and 1989 among residents in the Cagliari province, were identified from pediatric, surgical, neurologic, and orthopedic units of all local hospitals. Cases were also determined from records for referrals to centers outside Sardinia; identifications were made from files of the Regional Administration of Sardinia, which authorizes referrals to health-care centers outside of the region.
Histology was available for all cases. The seven cALL cases that formed the cluster in Carbonia in 1983-1985 and two additional cases that occurred in the same town during 1980-1989 were matched (i.e., 1:4) by gender and date of birth (i.e., month and year) to controls born in Carbonia and who resided in Carbonia at the time of diagnosis of the index case. Controls were selected from the birth register of the municipality of Carbonia. We were unable to trace the families of three controls; therefore, these controls were substituted with a second selection from the same birth register. Following written communication and/or contact by phone, both parents of all cases and controls were interviewed personally by one of the authors (MR) at their residences. All information for one case and one control was provided by the mother only. Prior to the interview, assurance was provided of its confidential nature, and informed consent was obtained. Attempts to corroborate the information obtained from the questionnaires with data provided by local authorities proved unsuccessful.
The questionnaire was divided into three sections. The first section focused on residential history of the family and the home environment, including proximity to sources of potentially harmful exposures (e.g., industries, mechanic shops, electric power lines, power stations). Drinking-water source and type of heating, house design and building materials, consanguinity between parents, and number and health status of other siblings were included in this section. The second included detailed paternal occupational and health history, personal habits of the father, and incidence of cancer cases among first-degree paternal relatives. In the remaining section, similar questions were asked of the mother; in addition, reproductive and pregnancy histories were also determined. Possible occupational exposures to solvents and radon-daughters were assessed by one of the authors (PC), who used a job/exposure matrix, based on the reported occupational histories. Socioeconomic status was defined as low, medium, or high, based on paternal occupation and education.
Associations of cALL with single risk factors were evaluated initially with crude odds ratios (ORs). Risk factors with an OR that exceeded 1.5 and that occurred in at least three exposed cases were also explored with an unconditional logistic regression model. We adjusted the risk estimate for the effects of socioeconomic status, maternal smoking, family history of cancer, and parental birth outside Carbonia. Other models with more covariates were also applied, but there was no improvement in terms of goodness-of-fit. We used SPSS for the logistic regression analysis. Symmetric 95% confidence bounds of the risk estimates were based on the Wald statistics for a null hypothesis.
Case fathers were, on average, somewhat older at the time of birth of the study subjects than were control fathers (33.4 y versus 30.7 y, respectively), but the difference was not significant statistically (p = .137). No difference in mean age at conception of the study subjects was observed for the mothers (26.1 y among case mothers versus 26.9 among control mothers). Birthorder rank was distributed equally among cases and controls: 4/9 cases and 15/36 controls ranked first in birth order, 4/9 cases and 16/36 controls ranked second and third, and 1/9 cases and 5/36 controls ranked fourth or more. No consanguinity was identified among case or control families. Neither study subjects nor their firstgrade relatives suffered from genetic or congenital diseases. None of the case fathers worked at the industrial plants nearby. No case mothers and only 2 control mothers reported illnesses or diagnostic roentgenographs during their pregnancies. Previous spontaneous abortions were reported by 3 case mothers and 14 control mothers (crude 0R = 0.8). Prior to diagnosis of cALL in their children, parents did not know other case parents. Other risk factors (e.g., favorite place for vacation, number of pets at home, number and type of electric appliances, schools attended) were distributed equally among case and control families.
The ORs derived from logistic regression are reported in Table 1. Only risk factors with a crude OR that exceeded 1.5 in at least 3 exposed cases were considered. Nonsignificant fourfold increases in cALL risk were associated with (a) at least 1 parent being born outside Carbonia, (b) presence of a well in the backyard, and (c) use of therapeutic drugs during pregnancy. A medium to high socioeconomic status was also associated with a decreased risk of cALL.
Table 1.-Risks of Childhood Acute Lymphoblastic Leukemia Associated with Risk Factors, Calculated with a Logistic Regression Model
Misclassification of meaningful risk factors and consequent loss of statistical power in establishing greater associations than those observed may have resulted from use of broad categories (e.g., cancer [all types] among relatives, use of therapeutic drugs during pregnancy, parental birth outside Carbonia). We created these broad categories by assembling specific risk factors, represented by very small numbers, to explore risks in a more manageable way.
Further data are required to support or discard a role of any of the risk factors discussed earlier, including an etiological role of possible outbreaks of common infections prior to the appearance of the cALL cluster in Carbonia.
[1.] Cocco P, Bernardinelli L, Biddau PF, et al. Childhood acute lymphoblastic leukaemia. a cluster in southwestern Sardinia (Italy). IntJ Occup Environ Health 1995; 1: 232-38.
[2.] Norusis MJ. SPSS/PC+ Statistics 4.0 for the IBM PC/XT/AT and PS/2. Chicago, IL: SPSS, Inc., 1990.
[3.] Cerquiglini MS, Viviano G, Ziemacki B, et al. Indagine sullo stato di inquinamento atmosferico nella zone di Ponoscuso (Cagliari). Roma: Istituto Superiore di Sanita, ISTISAN 84/7; 1984.
[4.] Mezzorani G, Piga MC, Pompei A. Stima di incremento della radioattivita naturale conseguente l'esercizio della centrale termoelettrica a carbone del Sulcis. Rendiconti Semestrali della Facolta' di Scienze dell'Universita' di Cagliari 1987; 57:1 7-33.
[5.] Robinson LL, Buckley JD, Daigle AK, et al. Maternal drug use and risk of childhood nonlymphoblastic leukemia among offspring: an epidemiologic investigation implicating marijuana (repon from the Children's Cancer Study Group). Cancer 1989; 63: 1904-11.
[6.] Laval G, Tuyns AJ. Environmental factors in childhood leukaemia. Br J Ind Med 1988; 45:843 44.
[7.] Lagakos SW, Wessen BJ, Zelen M. An analysis of contaminated well water and health effects in Woburn, Massachusetts. J Am Stat Assoc 1986, 81 :583-96.
[8.] Consorzio Ambiente Sardegna. Piano di disinquinamento dell'area a rischio del Sulcis-Iglesiente. Roma: Fase A Ministero dell'Ambiente, 1993.
[9.] Kislen L, Clark K, Hudson C. Evidence from population mixing in British new towns in 1946-85 of an infective basis for childhood leukaemia. Lancet 1990; 336: 577-82.
[10.] Alexander FE. Viruses, clusters and clustering of childhood leukaemia: a new perspective? Europ J Cancer 1993; 29A:1424-43.
[11.] Knox EG. Leukaemia clusters in Great Britain. J Epidemiol Community Health 1992; 46:573-76.
The authors acknowledge the collaboration of Mrs. Elena Cancedda, municipality of Carbonia.
Submitted for publication September 5, 1994; revised; accepted for publication August 15, 1995.
Requests for reprints should be sent to Dr. Pierluigi Cocco, Istituto di Medicina del Lavoro, Universita di Cagliari, vid S. Giorgio 12, 09124 Cagliari, Italy.
COPYRIGHT 1996 Heldref Publications
COPYRIGHT 2004 Gale Group