Osteosarcoma, seasonality, and environmental factors in Wisconsin, 1979-1989

Osteosarcoma, seasonality, and environmental factors in Wisconsin, 1979-1989

Mark E. Moss

Proxy exposure measures and readily available data from the Wisconsin Cancer Reporting System were used to contrast 167 osteosarcoma cases with 989 frequency-matched cancer referents reported during 1979-1989. Differences in potential exposure to water-borne radiation and fluoridated drinking water, population size for the listed place of residence, and seasonality were assessed. An association was found between osteosarcoma and residence in a population of less than 9 000 (odds ratio = 1.6, 95% confidence interval = 1.1 – 2.4). In addition, an association between month of birth (May through July versus other months of birth) and osteosarcoma among individuals who were less than 25 y of age (odds ratio = 1.9, 95% confidence interval = 1.1 – 3.4). Overall, no association was found between potential exposure to fluoridated drinking water and osteosarcoma (odds ratio = 1.0, 95% confidence interval = 0.6 – 1.5). The association between osteosarcoma and waterborne radiation was weak and was not significant statistically (odds ratio = 1.5, 95% confidence interval = 0.8 – 2.8).

OSTEOSARCOMA is an uncommon tumor of largely unknown etiology. Approximately 900 new cases occur each year in the United States.(1) The only known risk factor in humans is exposure to ionizing radiation.(2) Some studies, however, have noted geographic variations in incidence patterns for osteosarcoma; these variations suggest that environmental factors may be important.(3)(4) An animal study performed by the National Toxicology Program indicated that fluoride may play a role in osteosarcoma.(5) Conversely, other studies of fluoride in laboratory animals and humans have not supported a fluoride-osteosarcoma hypothesis.(6)(7)

We used proxy exposure measures and readily available data from the Wisconsin Cancer Reporting System to assess the role of several environmental exposures among osteosarcoma cases that occurred during the years 1979-1989. Specifically, the present study was undertaken to assess the environmental association between drinking-water fluoride, gross alpha radiation in water, and community population size with the occurrence of osteosarcoma in Wisconsin. Also explored were relationships between seasonality, by month of brith and by month of diagnosis, and occurrence of disease.

Method

Osteosarcoma data were obtained from the Wisconsin Cancer Reporting System. The Wisconsin Cancer Reporting System initiated statewide data collection in 1978. Subsequent to 1979, reporting and processing of data have been estimated to be at least 92.7% complete.(8) Incidence data collected for the period 1979 to 1989 were used. The Cancer Reporting System maintains both histologic (morphology) and site (topography) codes assigned according to the International Classification of Diseases for Oncology (ICD-O).(9) Population data were provided by the Center for Health Statistics at the Wisconsin Department of Health and Social Services.

The design was that of a case-referent records analysis. There were no interviews with study participants to ascertain residence or drinking water history. All information was gleaned from existing records. The odds ratio (OR), accompanied by the corresponding 95% confidence interval (95% CI), was used as the measure of association between exposure variables and osteosarcoma.

Exposure was determined indirectly with an environmental exposure matrix, based on listed place of residence. The environmental exposure matrix consisted of two parts: (1) environmental radiation exposure and (2) fluoridated drinking water exposure.

Fluoridation status for each listed place of residence was determined, using water supply data maintained by the Wisconsin Department of Natural Resources. The range of fluoride values for public water supplies in Wisconsin is less than 0.1 to 2.4 ppm. Communities that are fluoridated artificially adjust the level of fluoride to either 1.0 or 1.1 ppm (i.e., the optimal levels for protection against tooth decay). With respect to each place of residence, fluoridation status was coded as a binary variable that was either less than 0.7 ppm or 0.7 ppm and higher. In counties with low natural fluoride levels, rural addresses (addresses that included a rural route designation or a highway number) were coded as less than 0.7 ppm. In counties with natural fluoride levels at or above 0.7 ppm, rural addresses were coded as 0.7 ppm and higher.

Natural fluoride drinking water concentrations are known to be correlated with natural radium concentrations,(10) and high doses of radium-226 and radium-224 are known risk factors for osteosarcoma.(2) Data on gross alpha radiation in drinking water were used to control for the potential effect of radium on incidence of osteosarcoma. Radiation values for each place of residence were assigned in accordance with estimates obtained from a U.S. Environmental Protection Agency survey of water supplies that was done in 1980 through 1981.(11) Residences in counties that had a water system that contained a gross alpha radiation of 9 picoCuries per liter (pCi/I) or higher were coded as high-radiation areas, whereas all other residences were coded as low-radiation areas.

In an effort to assess seasonal variation, month of birth for cases was grouped into three or four seasons, as was done by Stark and Mantel(12) in their study of the relationship between seasonality and Down’s syndrome in Michigan. A variation of the Ederer-Myers-Mantel procedure(13)(14) was applied to seven different birth month classification schemes, as follows: four seasons with either January, February, or March as the first month of a 3-mo grouping and three seasons for which either January, February, March, or April was used as the first month of a 4-mo grouping. This approach assessed the probability of observing a cluster of a sufficient magnitude in a season, given the total number of cases observed in all months.

A second approach taken to assess seasonality was a comparison of cases with referents, using the same seven classification schemes; the season in which most cases occurred served as the exposure. In addition to month of birth, these two approaches were also applied to month of diagnosis.

Cases were primary osteosarcoma tumors that were occurring in Wisconsin and that were reported to the Wisconsin Cancer Reporting System during the period 1979-1989 (ICD-O morphology codes 91803-91903). Initially, two separate referent groups were identified and analyzed. Brain and nervous system tumors (ICD-O topography codes 191.0-192.9, referred to hereinafter as brain cancer referents), served as the primary comparison group. A smaller second referent group that included digestive system cancers (ICD-O topography codes 155.0-158.9, referred to hereinafter as digestive cancer referents), was selected to assess whether substantial bias resulting from choice of referents resulted.(15) All referents were reported to the Wisconsin Cancer Reporting System during the same period that cases were reported. The digestive cancer group was limited to cancers of the liver, gallbladder, pancreas, retroperitoneum, and peritoneum. Table 1 contains the distribution of the two referent groups, by type of tumor.

Table 1.–Distribution of Selected Referents, by Tumor Type (Wisconsin

Cancer Reporting System, 1979-1989)

No. Percentage Percentage

Tumor Site referents of referent of all

(ICD-O code) (n = 989) group referents

Brain cancer

referents (n = 647)

Brain (191.0-191.9) 594 91.8 60.0

Other nervous sytem

(192.0-192.9) 53 8.2 5.4

Digestive cancer

referents (n = 342)

Liver (155.0-155.9) 79 23.1 8.0

Gallbladder 70 20.5 7.1

(156.0-156.9)

Pancreas (157.0-157.9) 177 51.8 17.9

Retroperitoneum/

peritoneum 16 4.7 1.6

(158.0-158.9)

Statistical power was improved by frequency matching four referents from each of the two referent groups to each case, according to age (0-14 y, 15-24 y, 25-34 y, 35-44 y, 45-54 y, 55-64 y, 65-74 y, 75-84 y, and [greater than or equal to] 85 y); sex; and race (white, black, and other). We matched for age because, unlike the referent cancers, osteosarcoma has a bi-modal age distribution, and large numbers of cases occur around adolescence and also at approximately 50 y of age.(2) A computer program for generating a file of matched case-referent sets was used.(16) A preliminary matched analysis, using Epi Info,(17) was performed to take advantage of the matched design. Conditional logistic regression was achieved by using a modification of PROC PHREG in SAS.(18) The results of the statistical models in which the two referent groups were used separately were nearly identical; therefore, both referent groups were combined. The results, using the combined cancer referent groups, are presented (see Results).

Given that communities with larger populations tend to be fluoridated and, in many instances, have a longer history of fluoridation, odds ratios and confidence intervals were adjusted for population size. Preliminary modeling revealed that the variable for population size was related nonlinearly to case status when it was coded as a continuous variable in logistic regression. We used the approach outlined by Wartenberg and Northridge(19) to explore various cut points for the population size in which the individual resided. This resulted in our choosing a cut point of 9 000 for population size categorization. Whereas the data demonstrated nonuniform differences in the proportion of cases and referents among incremental increases in population size, we thought the best summarization of the data was as follows: in general, populations greater than 9 000 were different from populations of less than 9 000 with respect to risk of osteosarcoma. It should be noted that almost the same parameter estimates were derived from models that used a population size of 5 000 as the cut point; however, cut points above 9 000 did not show a strong difference between cases and referents for this variable.

Results

During the 11-y study period (i.e., 1979-1989), 167 cases of osteosarcoma were reported in Wisconsin. Of this number, 84 were male and 83 were female. Nine-ty-two percent (n = 154) of osteosarcoma cases were white, 4.7% were black, and 3.6% were coded as “other.” The race category termed “other” included 4 cases coded in the cancer registry as being of unknown ethnic background; 1 case of hispanic background; and 1 case as other than white, black, American Indian, or Hispanic.

Among the referent groups, fewer than 4 matched brain cancer referents per case occurred in some of the age groups, resulting in 317 matched brain cancer referents for females, rather than 332, and 330 matched brain cancer referents for males, rather than 336 (Table 2). Incomplete matching resulted because of a limited number of nonwhites in the referent sampling frame. In the digestive cancer referent series, 342 referents were matched with 141 cases. As stated previously, the results of the statistical models that used the two referent groups separately were nearly identical, so both referent groups were combined.

Table 2.–Age and Sex Distribution of Osteosarcoma Cases and Age-,

Gender-, and Race-Matched Referents (Wisconsin Cancer Reporting System, 1979-1989)

Osteosarcoma Brain Cancer Digestive cancer

cases referents referents

Age group (y) Female Male Female Male Female Male

0-14 19 11 66 43 9 8

15-24 16 22 62 85 8 11

25-34 6 9 24 36 14 24

35-44 5 10 20 40 12 32

45-54 3 0 12 0 12 0

55-64 6 9 24 36 24 32

65-74 13 12 52 48 40 28

75-84 14 8 53 30 48 24

85+ 1 3 4 12 4 12

Total 83 84 317 330 171 171

The findings from the matched case-referent analysis of osteosarcoma incidence and the study variables of interest are presented in Tables 3-5. Heterogeneity of effects was examined by dividing our study population into age-specific subgroups (i.e., < 45 y and [greater than or equal to] 45 y) and gender-specific subgroups.

Table 4.–Birth-Month among Cases and Referents Who Were Less Than 25 y

of Age (Osteosarcoma Cases and Combined Referents, Wisconsin, 1979-1989)

Birth No. cases No. referents Odds ratio 95% CI

May-July 27 75 1.9 1.1-3.4

August-April 41 217

[TABULAR DATA OMITTED]

The strongest association was observed between osteosarcoma and residence in a population of 9 000 or less (OR = 1.6; 95% CI, 1.1-2.4). Residence in an area with high gross alpha radiation demonstrated an effect in the anticipated direction, i.e., a point estimate for the adjusted OR of 1.5 (95% CI, 0.8-2.8). The adjusted OR estimate for the association between fluoridated drinking water and osteosarcoma was 1.0 (95% CI, 0.6-1.5).

In both age-specific subgroups, the proportion of cases that resided in a community with a population of less than 9 000 was higher than the proportion of controls. This difference was greater in the older age subgroup, especially after adjustment for the radiation and fluoride exposure variables. Residence in a community that was served with fluoridated drinking water showed wide variability around the point estimate for the odds ratios for both age groups. The ORs for gross alpha radiation in drinking water suggest an association with incidence of osteosarcoma among individuals who are younger than age 45 y. The data did not show an association with drinking water radiation in older individuals.

In gender-specific models, there was an effect of radiation for females but not for males. However, as in all the observed associations, the point estimates were accompanied by wide confidence intervals. The impact of population size was significant statistically and was more pronounced in males.

The data were examined to determine any patterns of seasonality in osteosarcoma incidence, by month of diagnosis and by month of birth, for cases and referents. No suggestion of seasonality, by month of diagnosis, was found. The Ederer-Myers-Mantel procedure suggested clustering for seasonality, by month of birth, for individuals who were less than 25 y of age. This occurred when the calendar year was divided into four seasons, with February as the first month (chi-square statistic = 5.86, p < .05). More precisely, it is shown in Table 4 that birth in May through July was more common among individuals aged 25 or less in the case group, compared with similarly aged referents. Other comparisons, by month of birth, did not achieve statistical significance.

Potential years of fluoride exposure, defined as years of fluoridation for residence or age of subject, whichever was less, are shown in Table 5. This resulted in classification of potential fluoridation exposure status as 0 y, 1-24 y, and [greater than or equal to] 25 y. The adjusted point estimates were all very close to the null value (both above and below OR = 1.0), and, once again, the wide confidence intervals highlight the variability in the data.

Discussion

One potential methodological weakness of this case-referent analysis was the use of cancer referents who themselves may have had an environmental etiology. An association was found between osteosarcoma and living in a town with a population of less than 9 000. This association was strongest among males and individuals aged 45 y and older. This indicates that the referent cancers were more common among municipalities with larger populations, a finding consistent with the study by Howe et al.(20) that investigated differences in population density gradients among different cancers in Illinois. Among the referent cancers used in the present study (Table 1), we found that, among whites, brain and nervous system, liver, and pancreas cancers occurred with greater frequency in areas with higher population densities. Howe et al.(20) did not separate osteosarcomas from all bone cancers; however, bone cancers displayed a convincing trend inasmuch as rates increased as population density increased for white males (p < .001), but this result was less notable for white females (p < .10). Our descriptive approach identified a difference in osteosarcoma occurrence in small towns and, as stated previously, this difference was not apparent when the selected cut point exceeded 9 000 for the population variable (data not shown).

An association was found between month of birth (May through July versus other) and osteosarcoma among individuals who were less than 25 y of age. This finding is supportive of either an infectious agent or an agricultural toxin hypothesis in osteosarcoma development. Whereas laboratory animal studies support the role of a viral agent, sufficient epidemiologic evidence has been lacking in humans.(2) And, in contrast to our findings, a previous study that specifically investigated seasonality found no evidence that month of birth was associated with osteosarcoma mortality in 819 children (i.e., < 20 y of age) throughout the United States, by state or geographic region.(21)

Given the fact that there were multiple comparisons in our analysis of seasonality, our finding of an association with month of birth should be interpreted as a hypothesis that deserves further attention. Prior support for an agricultural toxin in osteosarcoma was presented by Schwartzbaum et al.(22) These authors identified a statistically significant OR of 2.6 among 78 childhood osteosarcoma patients for parents who reported that they gardened with fertilizers, herbicides, and pesticides during the postnatal, prediagnostic period, compared with parents of other childhood cancer patients. Wisconsin has a large amount of land devoted to farming, and it is likely that exposure to agricultural toxins would be more prevalent in smaller, more rural communities. The climate and growing season in Wisconsin may result in a pattern of seasonal exposure to such toxins. In this way, month of birth may be a proxy for seasonal exposure to toxins common to certain agricultural regions, but not necessarily in others.

Indirect determination of exposure status for fluoride and gross alpha radiation in drinking water resulted in an inability to determine any associations across age and gender subgroups, as well as overall. This is not too surprising because this approach carries with it a great potential for nondifferential misclassification of exposure status(23) and the consequent underestimation of effect.(24) The potential presence of nondifferential misclassification and unmeasured confounders in the data makes it reasonable to assume that further research is needed to clarify the impact of fluoride in osteosarcoma, especially among individuals aged 45 y and older.

In 1990, results of a laboratory study of sodium fluoride(5) received wide publicity. Several investigations of the fluoride-osteosarcoma hypothesis have occurred since the presentation of these results, and no such investigation has demonstrated that fluoride exposure confers risk for osteosarcoma.(25)(26)(27) However, as was the case in the present study, two of these studies(25)(26) relied on indirect assessment of exposure to fluoride to estimate the effect on osteosarcoma, and the study by McGuire et al.(27) was limited to individuals who were less than 40 y of age at diagnosis.

In summary, we found residence in a community with a population of less than 9 000 to be associated with osteosarcoma. The data also suggested month of birth to be an important factor in childhood osteosarcoma incidence in Wisconsin. Future studies of osteosarcoma should assess the impact of socioeconomic variables and agricultural contaminants. Further investigation is also needed to determine the role of fluoride in older individuals who develop osteosarcoma. Efforts should be made to ascertain better measures of cumulative fluoride exposure.

References

(1.)Goorin AM, Abelson HT, Frei E, III. Osteosarcoma: fifteen years later. N Engl J Med 1985; 313:1637-43.

(2.)Fraumeni JF, Jr., Boice JD, Jr. Bone. In: Schottenfeld D, Fraumeni JF, Jr., Eds. Cancer Epidemiology and Prevention. Philadelphia, PA: WB Saunders Co, 1982; pp. 814-26.

(3.)Bovill EG Jr., Kungu A, Bencivenga A, Jeshrani MK, Mbindyo BS, Heda PM. An epidemiological study of osteogenic sarcoma in Kenya: the variations in ethnic groups and geographic regions, 1968-1978. Int Orthop 1985; 9:59-63.

(4.)Larsson S-E, Lorentzon R. The geographic variation of the incidence of malignant primary bone tumors in Sweden. J Bone Joint Surg [Am] 1974; 56-A:592-600.

(5.)National Toxicology Program. Toxicology and carcinogenesis of sodium fluoride in F344/N rats and B6C3FI mice. Technical Report series 393, NIH Publication No. 90-2848, 1990.

(6.)U.S. Department of Health and Human Services. Review of fluoride benefits and risks. Washington, DC: Public Health Service, Department of Health and Human Services; February, 1991.

(7.)World Health Organization. Fluorine and fluorides. Environmental Health Criteria 36. Geneva, Switzerland: World Health Organization, 1984.

(8.)Fischer M. An evaluation of the Wisconsin cancer reporting system’s completeness of case reporting. Madison, Wisconsin: Wisconsin Department of Health and Social Services, Division of Health, Center for Health Statistics; 1982.

(9.)World Health Organization. International Classification of Diseases for Oncology. Geneva, Switzerland: World Health Organization, 1976.

(10.)Lynch CF. Relationship of fluoride in drinking water to other drinking water parameters. Arch Environ Health 1987; 42:5-13.

(11.)Horton TR. Nationwide occurrence of radon and other natural radioactivity in public water supplies. (EPA 520/5-85-008). Montgomery, AL: U.S. Environmental Protection Agency, October 1985.

(12.)Stark CR, Mantel N. Lack of seasonal- or temporal-spatial clustering of Down’s syndrome births in Michigan. Am J Epidemiol 1967; 86:199-213.

(13.)Ederer F, Myers MH, Mantel N. A statistical problem in space and time: do leukemia cases come in clusters? Biometrics 1964; 20:626-38.

(14.)Mantel N, Kryscio RJ, Myers MH. Tables and formulas for extended use of the Ederer-Myers-Mantel disease-clustering procedure. Am J Epidemiol 1976; 104:576-84.

(15.)Smith AH, Pearce NE, Callas PW. Cancer case-control studies with other cancers as controls. Int J Epidemiol 1988; 17:298-306.

(16.)Hanrahan LP, Olson J, Anderson HA, Bruns J. STRATRAN: Stratified random selection of controls for matched case-control data. SAS Users Group Notebook 1987; pp 1006-1010.

(17.)Dean AG, Dean JA, Burton AH, Dicker RC. Epi Info, version 5: a word processing, database, and statistics program for epidemiology on microcomputers. Stone Mountain, GA: USD, Incorporated, 1990.

(18.)SAS Institute, Inc. SAS Technical Report P-217 SAS/STAT Software: The PHREG Procedure. Cary, NC: SAS Institute Inc, 1991; pp 37-41.

(19.)Wartenberg D, Northridge M. Defining exposure in case-control studies: a new approach. Am J Epidemiol 1991; 133:1058-71.

(20.)Howe HL, Keller JE, Lehnherr M. Relation between population density and cancer incidence, Illinois, 1986-1990. Am J Epidemiol 1993; 138:29-36.

(21.)Glass AG, Fraumeni JF, Jr. Epidemiology of bone cancer in children. J Nat Cancer Inst 1970; 44:187-99.

(22.)Schwartzbaum JA, George SL, Pratt CB, Davis B. An exploratory study of environmental and medical factors potentially related to childhood cancer. Med Pediatr Oncol 1991; 19:115-21.

(23.)Grembowski D. Measuring length of exposure to fluoridated water. Community Dent Oral Epidemiol 1988; 16:131-34.

(24.)Copeland KT, Checkoway H, McMichael AJ, Holbrook RH. Bias due to misclassification in the estimation of relative risk. Am J Epidemiol 1977; 105:488-95.

(25.)Mahoney MC, Nasca PC, Burnett WS, Melius JM. Bone cancer incidence rates in New York state: time trends and fluoridated drinking water. Am J Public Health 1991; 81:475-79.

(26.)Hoover RN, Devesa SS, Cantor KP, Lubin JH, Fraumeni JF, Jr. Fluoridation of drinking water and subsequent cancer incidence and mortality. In: Review of Fluoride Benefits and Risks. Appendix E. Washington, DC: Public Health Service, Department of Health and Human Services; February, 1991.

(27.)McGuire SM, Vanable ED, McGuire MH, Buckwalter JA, Douglass CW. Is there a link between fluoridated water and osteosarcoma? J Am Dent Assoc 1991; 122:38-45.

COPYRIGHT 1995 Heldref Publications

COPYRIGHT 2004 Gale Group