[CANCER RESEARCH 48. 7298-7303. December 15. 1988]

Cancer Morbidity and Mortality in Phosphate Workers1

Gladys Block,2 GenevièveM. Matanoski, Raymond Seltser,3 and Thomas Mitchell

Departments of Epidemiology [G. B., G. M. A/., R. S.] and Environmental Health Sciences [T. M.J, The Johns Hopkins School of Hygiene and Public Health,Baltimore, Maryland 21205

ABSTRACT

Phosphate ore mining and processing operations are associated withdusty conditions and potential exposure to ->radiation and radon daughter

products. Although most current measurements are within OccupationalSafety and Health Administration (OSHA) guidelines, no measurementsexist for the era preceding modern standards of industrial hygiene andworkplace ventilation. All workers employed by the participating phosphate company between 1950 and 1979 were ascertained, and 3451 malesemployed for approximately 6 months or more comprise the studypopulation. Statistically significant elevations in lung cancer (standardized mortality ratio = 1.62) and emphysema were observed in white butnot in black workers, in relation to I . S. rates. Among workers for whom20 years had elapsed since first employment, there was a dose-responsetrend of increasing lung cancer risk with increasing duration of employment (standardized mortality ratio = 2.48 with 20 years of employment).There was no evidence of excess lung cancer risk among employees hiredafter 1960. Multivariate analyses and internal comparisons of risk by jobtype are consistent with a hypothesis of occupational!) related lungcancer, but small numbers prevent firm conclusions.

INTRODUCTION

In recent years, the possibility of adverse health effects arisingfrom exposure to phosphate mining and phosphate fertilizerproduction has become of increasing interest (1), with concernfocussing on the possibility of a lung cancer risk. In addition,concern about nonoccupational radon exposure has directedattention to occupational populations for information aboutpotential health effects. Potential exposures in the phosphateindustry include chemical fumes (e.g., sulfuric acid, fluorides,sulfur dioxide) from the beneficiation and fertilizer manufacturing processes; high silica dust levels in some job areas; andradiation resulting from uranium and radon daughter productsassociated with phosphate ore (2). The study reported here wasundertaken at the request of a Florida phosphate company todetermine whether there is an excess cancer risk among itsworkers.

METHODS

Workers employed between 1950 and 1979 were ascertained fromcompany and union records, supplemented by a census of all of thecompany's 1RS 941 Quarterly Reports from 1950 to 1978, a process

which assured complete ascertainment. Workers employed for 6months or more were included. Mortality follow-up was through 1981.Females accounted for only about 5% of the study population, andwere not included in these analyses.

The study group was characterized with respect to age, sex, race,date of employment, and duration of employment using companyrecords, if available, supplemented by reference to Florida MotorVehicle records, death certificate information, and by direct questionnaire. If date and duration of employment were unavailable from

Received 3/17/88; revised 7/25/88; accepted 9/21/88.The costs of publication of this article were defrayed in part by the payment

of page charges. This article must therefore be hereby marked advertisement inaccordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1This research was supported by funds from the American Cyanamid Corpo

ration. The authors gratefully acknowledge that support.! To whom requests for reprints should be addressed, at Division of Cancer

Prevention and Control, National Cancer Institute. Executive Plaza North, Room313, 9000 Rockville Pike, Bethesda. MD 20892.

3Current address: Centers for Disease Control. Rockville, MD 20815.

company records, they were calculated from the dates in which anindividual appeared on the Quarterly Reports referred to above. Table1 presents the distribution of the study population by race and durationof employment (defined by the number of Quarterly Reports on whichthe individual appeared). Workers with five or more quarters of employment are referred to below as the long-term cohort.

For some analyses, missing values were imputed using the studypopulation distribution, and the individual's Social Security number.

Thus, for external SMR analyses (but not for the internal comparisonsshown in Tables 6 and 7) approximately 30% of the years of birth wereestimated, using a method described elsewhere (3). To investigatepotential bias, an entirely different imputation method was also evaluated, which was based on the age distribution by race, year hired, andduration of employment. The impact on risk estimates of excluding allworkers with missing values is addressed in the "Discussion" below.

Vital status was ascertained using Social Security Administrationdata, supplemented by motor vehicle and vital statistics records andpersonal contact, and confirmed by death certificate. Results are summarized in Table 2. Causes of death were coded by trained nosologistsaccording to the Ninth Revision of the International Classification ofDiseases. Data on nonfatal cancer morbidity were obtained by a searchof the records from all of the major central Florida tumor registries,and from questionnaire information confirmed by physician or hospitalrecords. All current workers and a large majority of retired workerslived in Central Florida.

An industrial hygiene analysis was conducted, in which detailed jobdescriptions covering the entire study period were examined, currentplant conditions were inspected at the job site, and recent radiationmeasurements in this and other phosphate companies were ascertained.A job coding system was developed based on this industrial hygieneanalysis, which characterized each job with respect to major plant area,dust exposure, chemical fumes, and several other exposures. Becauseprocesses and practices had changed over time and no historical dataon radiation levels existed, it was not possible to assign quantitativeradiation levels to jobs or work areas. Jobs were characterized withrespect to their exposure to dust as a proxy measure, since phosphatedust had been shown to be highly correlated with radon daughter level.'1

Individual job histories were recorded directly from company records,and individuals were categorized according to the job area in whichthey had worked for the longest time.

As a result of this industrial hygiene analysis, only Drying/Shipping,Chemical/Fertilizer, and Maintenance job categories were found tohave the potential for exposure to high levels of dust, chemical fumes,or radiation. Other job types were found to have minimal risk of suchexposures. The Drying/Shipping area had jobs with potential for intermittent exposure to severe or enclosed dust conditions. In addition,time-weighted average exposures to radon progeny greater than twicethe OSHA standards were observed inside poorly ventilated loadingtunnels in the Drying/Shipping area, in other plants in the industry(4); and in both this and other plants, time-weighted average exposuresto long-lived a radiation exceeding the OSHA standard were observedin a few measurements in certain very dusty jobs in this Drying/Shipping category.5' ' The Chemical and Fertilizer area had chemical

fumes and a few elevated y radiation measurements. Workers classifiedas Maintenance tended not to be exposed to a single plant area butworked in a variety of locations, and thus had a variety of intermittentexposures, including those in dusty or chemical areas. Thus, the industrial hygiene assessment suggested a priori that highest risk might exist

4 DeLong, C. T. Evaluation of radiation exposure at Brewster Phosphates:long-lived alpha, silica exposures. September, 1980. Unpublished.

s DeLong. C. T. Evaluation of radiation exposure at Brewster Phosphates:long-lived alpha, exposure. October 1979. Unpublished.

* DeLong, C. T. Evaluation of radiation exposure at Brewster Phosphates:

radon progeny. January. 1980. Unpublished.

7298

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MORTALITY IN PHOSPHATE WORKERS

Table 1 Total population, male phosphate workers 1950-1979 who appeared onthree or more 1RS Quarterly Reports

One Quarterly Report is submitted for each calendar quarter. An individualwho appeared on four Quarterly Reports could have worked as little as 6 monthsand 2 days, or as much as 12 full months.

Numberof quarters3-45-89-2021-4041-8081-112TotalWhiteNo.6936295902812721452610%26.624.122.610.810.45.6BlackNo.230170198937278841%27.320.223.511.18.69.3TotalNo.9237997883743442233451%26.723.222.810.810.06.5

Table 2 Vital status as of December 31, 1981, phosphate workers employed 6months or more, 1950-1979

No.%Total

Known statusAliveDead

Death certificateNo death certificate3451

32252739486468

18100.0

93.5(100.0)

(96.3)(3.7)

Status unknown 226 6.5

in Drying/Shipping, with moderate risks in Chemical/Fertilizer andMaintenance jobs, and little risk elsewhere.

In data presented below, duration of employment by this companyis used as a proxy for exposure to potentially carcinogenic agents, andlength of time since tirsi employment is used as a proxy for latency,the time between cancer initiation and manifestation.

Questionnaires were sought from all workers employed approximately 1 year or longer with this company, the "long-term cohort," or

their survivors. The length of time since last employment by thecompany (as much as 30 years in some cases) and limited informationon many individuals (even first name was missing in some cases) madetracing difficult. After extensive tracing procedures using last knownaddress, State Department of Motor Vehicles records, State vital statistics records, R. L. Polk Company tapes, telephone directories, andthe remarkable memories of some long-term employees, 1078 of the2155 members of the long-term cohort or their survivors were contactedand 992 responded. This represented 46% of the long-term cohort, anda 92% response rate among those who were contacted. Informationwas elicited regarding duration and nature of phosphate work; exposureto selected industries or agents known to be associated with canceroutcomes of interest; residence on reclaimed phosphate land; medicalX-ray exposures; smoking, alcohol, and diet; age, sex, race, and education; and medical history including cancer and other major medicalconditions.

Cancer incidence was traced by means of questionnaires confirmedby medical records, and by tumor registry searches, and SIRs7 were

calculated using Connecticut incidence rates. This registry is the oldestand one of the largest in the nation, and thus has relatively stable cancerrates covering the entire study period, 1950-1979. Since the focus hereis on oral and nasal cancers, it should be noted that Connecticut ratesof these tumors are not higher than Florida rates.

To estimate the study population's risk in relation to general U. S.

rates, SMRs adjusted for age and calendar time were calculated usinga computer program developed by Monson (5), and tested for significance at the 0.05 level using the Poisson distribution. A 1-year lag wasused in the calculation of person-years. Standardized incidence ratioswere calculated in an identical manner, using incidence rates from theConnecticut Tumor Registry. Adjustment for multiple confoundingvariables was carried out by multiple logistic regression using a cohortand person-years approach, with a computer program developed by Dr.James Tonascia. Internal comparisons between job categories werecarried out using the log-linear Poisson model described by Breslowand Day (6), with a computer program developed by Dr. James Ton-

1The abbreviations used are: SIR, standardized incidence ratio; SMR, stand

ardized mortality ratio.

ascia. SMRs of this latter type represent the ratio of the death rate inthe particular job category to the death rate in the study population asa whole, adjusted for differences in age and time of entry into the study,and permit a comparison of risks by job type. Differences in thedistribution of variables among groups were examined using bivariateanalysis and \: tests.

In the choice of an external comparison population, U. S. rates,Florida rates, and Polk County rates (where the plant is located) wereconsidered. However, Florida's role as a retirement state makes the

appropriateness of using Florida rates questionable, since Florida ratesmust include at least some who have moved to Florida with incipientillness or with socioeconomic characteristics associated with excessrisk. [Haenszel, for example, found significantly higher lung cancerrates among individuals who had changed residences, even after controlfor smoking and socioeconomic status (7).] Polk County's rate of in-migration is as great as that of Florida as a whole.8 It could be argued

that observed risks in the study population might be attributable toradiation from underlying phosphate deposits rather than occupationalexposures. However, the available data provide no strong indication ofan elevated background environmental risk in the local population,since while white male rates for lung cancer in Polk County aresignificantly higher than U. S. rates, female rates are not significantlydifferent from U. S. rates (8, 9). Indeed, for most cancer sites PolkCounty rates are either no different or significantly lower than U. S.rates. Therefore, while SMRs based on Florida rates are briefly presented and discussed, results reported below in Tables 3-5 which usean external comparison population are based on U. S. rates.

RESULTS

Standardized mortality ratios for selected causes of death areshown in Table 3. Among white males employed for 6 monthsor more, there is a statistically significant excess of the category"All malignant neoplasms," due largely to the significant excess

of cancer of the respiratory system (SMR = 1.62). Emphysemais also significantly elevated (SMR = 2.19), as is the category"Ill-defined conditions." Diseases of the circulatory system, on

the other hand, are significantly lower than expectation amongboth white and black workers. The all-cause mortality ratioamong white male workers is exactly 1.00, somewhat higherthan is sometimes observed in a healthy worker population.Among black workers, no cause of death is in excess of expectedrates based on U. S. black male rates. Indeed, both all-causemortality and diseases of the circulatory system are significantlybelow expectation, in blacks.

When SMRs for lung cancer among white males are recalculated using Florida rates, a lower but still statistically significant excess (SMR = 1.50) is seen.

It was hypothesized that radionuclide-bearing dust particlesmight produce an elevated risk of oral and nasal cancers, sitesat which malignancies are often not rapidly fatal. Consequently,standardized incidence ratios were calculated using Connecticutincidence rates (data not shown). The SIR for cancers otherthan respiratory is 0.80. Only for cancer of the respiratory tractis there a significant elevation, with an SIR of 1.55. The dataprovide no evidence of an excess incidence of head and neckcancer in the study population.

To explore further the observed excess of respiratory cancerdeaths among white males, the mortality risk analysis wasrepeated on a subset of white males with at least 5 years sincetheir first employment by this company. In this way the riskestimate is not deflated by the inappropriate inclusion of workers who, because of insufficient latency, may not be realisticallyat risk of developing employment-related respiratory cancer.Short-term and longer-term workers were then examined separately in this at-risk cohort.

* U. S. Bureau of the Census. Florida, 1977.

7299

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MORTALITY IN PHOSPHATE WORKERS

Table 3 Observed deaths through 1981, standardized mortality ratios (U. S. population rates as standard). Male phosphate workers who worked 6 months or more1950-1979, by cause of death

Cause ofdeath(ICD8th revisioncode)All

causes ofdeathInfectiveand parasitic diseases(001-139)All

malignant neoplasms(140-209)Cancer,buccal cavity(140-149)Cancer,all digestive organs(150-159)Cancer,esophagus(150)Cancer,stomach(151)Cancer,large intestine(153)Cancer,rectum ( 154)Cancer,liver(155)Cancer,pancreas(157)Cancer

of respiratory system(160-163)Cancer,larynx(161)Cancer,lung(162)Cancer,bone(170)Cancer,skin(172)Cancer,prostate(185)Cancer,

testis(186)Cancer,bladder(188)Cancer,kidney(189)Cancer,brain and other CNS (19 1- 192)Cancer,thyroid(193)Lymphosarcoma

and reticulosarcoma(200)Hodgkin'sDisease(201)Leukemia

(204-207)Cancer,other lymphatic tissue(202)All

lymphopoietic cancer(200-209)Benignneoplasms(210-229)Allergic,

endocrine, metabolic, nutr.(240-279)Diabetesmellitus(250)All

dis. blood and blood-forming organs(280-289)Mental,psychoneurotic disorders(290-319)All

diseases of nervous system(320-389)Alldiseases of circulatory system(390-459)Chronic

rheumatic heart disease(393-398)Arterioscleroticheart disease(410-414)Vascular

lesions of CNS(430-438)Allrespiratory diseases(460-519)Allpneumonia(480-486)Emphysema

(492)Alldiseases of digestive system(520-579)Gastric

and duodenal ulcer(531-532)Cirrhosisof liver(571)All

diseases of genitourinary system(580-629)Chronicnephritis(582)Symptoms,

senility, ill-defined cond.(780-799)Allexternal causes of death(800-999)Allaccidents(800-929)Motor

vehicle accidents(810-829)Suicides(950-959)White

(A' =2607)"Observed

deaths346386219225514392370262012000224044212139210914267111133201151311412SMR1.000.821.26*0.881.051.240.610.862.450.83.08.62*.91.62»0.00.41.803.340.000.560.820.000.000.000.701.210.550.000.700.842.780.440.640.84*0.540.890.721.391.082.19*0.601.330.290.500.002.49*1.070.970.851.13Black

(N =840)°Observed

deaths127126282240009091020200000202022103351196530723116291771SMR0.74*0.210.931.730.911.100.892.350.000.000.000.970.001.047.050.000.850.003.690.000.000.000.000.002.530.000.920.000.630.741.870.001.680.51*0.890.53*0.41*0.530.570.000.732.190.550.240.521.060.911.020.960.53

" One black and three white workers were excluded because their dates of birth were unknown and could not be imputed.*/><0.05.

Table 4 Observed deaths, person-years and SMRs, respiratory cancerWhite male phosphate workers employed 6 months or more, 1950-1974°by

duration of employment.*

Duration ofemployment Deaths SMR

6-11 months1+ years

1127

9,09030,981

2.33"

1.42" That is, at least 5 years had elapsed since first employment.»2607 white males (three workers were omitted because their dates of birth

were unknown and could not be imputed); 38 respiratory cancer deaths. Onedeath in the 6-11 month group is excluded by the 5-year latency restriction.

' Person-years of observation.°><0.05.

Short-Term Workers. Although caution is appropriate incomparing SMRs, Table 4 suggests that workers with 6 to 11months of employment by this company have a risk elevationof 2.33, considerably greater than that of members of the long-term cohort. This apparent excess mortality among very short-term (6-11 months) workers is counterintuitive in an industryin which virtually all workers are exposed to most jobs, including the dusty ones, at some time during their employment. Thatis, longer-term workers are at least as likely to have had 6-11

months of exposure to dusty jobs as are short-term workers, inthis industry. These short-term workers were therefore investigated in an attempt to determine what factors might be associated with the observed excess risk of lung cancer.

10 of the 11 lung cancer deaths among short-term workersoccurred among those hired between 1956 and 1959, a timeperiod corresponding to a major plant construction in thiscompany. Thus, the usual occupations of these workers mayhave included more heavy industrial construction trades (suchas welding) than was the case among workers in the longer-term group. Furthermore, nine of the 11 lung cancer deathsamong short-term workers occurred less than 15 years aftertheir hire date, a fact which suggests the possibility of cancerinitiation at some prior time. These data suggest that most ofthe excess respiratory cancer mortality seen in very short-term(6-11 months) workers may be attributable to characteristicsor exposures other than the phosphate work represented inthese data.

Long-Term Workers. Among workers employed 1 year ormore by this company, the relationship between duration ofemployment, latency and respiratory cancer mortality is seen

7300

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MORTALITY IN PHOSPHATE WORKERS

Table 5 Observed deaths and SMRs, respiratory cancerWhite male phosphate workers employed 1 year or more, 1950-1974, by

duration of employment and time since first hired

Table 6 Deaths and SMRs: Internal comparisons, by job areaLong-term cohort, male phosphate workers employed 1 year or more.

Time since first hired (years)

Duration of1-19 years 20+ years Total

employment1+

years1

-9years10-19years20+yearsDeaths

P-Y105525,41520,7094,705SMR0.980.691.66DeathsP-Y1743105,5662,2929272,336SMRDeathsP-Y1.93°1.221.612.48°27981030,98123,0025,6322,337SMR1.420.861.642.48°

in Table 5. There appears to be a dose-response relationshipwith duration of employment, particularly among those forwhom 20 or more years of latency have elapsed. Among thoseworkers the SMR increased from 1.22 in those employed 1-9years, to 1.61 in those employed 10-19 years, to an SMR of2.48 in those employed 20 years or more. Among workers hiredsince 1960 (data not shown) there is no evidence of excess risk.

Since the apparent findings with respect to respiratory cancermortality could be due to confounding factors, such factorswere examined among the approximately 1000 workers forwhom potential confounder data existed from questionnaires.Smoking, education, and longest phosphate job were significantly associated with both lung cancer and duration of phosphate employment. Among those who responded to the questionnaire, 75.6% of whites had ever smoked, while only 63.9%of blacks had smoked. While smoking rates for whites aresimilar to U. S. population data, among blacks the proportionwho had ever smoked is considerably lower than the 82.8%seen among blacks in the U. S. population as a whole (10). Ifthe questionnaire respondents are representative of the entirestudy population with respect to smoking, one may crudelyestimate the effect of smoking by adjusting the expected deathsproportional to the ratio of smokers in the U. S. and studypopulations. The result is an estimated smoking-adjusted respiratory cancer SMR of 1.22 for blacks and 1.52 for whites.Though the figure of 1.52 for whites is somewhat lower thanthe 1.62 seen in Table 3, it is still a statistically significantexcess; the figure for blacks is considerably higher than wasobtained without this adjustment. These figures must be regarded simply as estimates, since smoking rates in the entirestudy population are not well known, nor have detailed smokingrates by age or calendar time been considered. They do suggest,however, that while smoking may contribute to the observedlung cancer excess in whites, it may not explain it completely.

The role of phosphate job type was examined in internalcomparisons to determine whether or not the risks appeared tobe consistent with the a priori knowledge about exposures,using the log-linear approach of Breslow and Day (6). SMRsfor each job area are based on expected rates derived from thestudy population as a whole. Conclusions are thus independentof any external population and one can examine whether thenumber of cancers among workers in a particular job typefollows the distribution which would be expected, based oncancer rates in the study population itself. The SMRs areadjusted for age, calendar time, and years of observation. Because of the small numbers of workers in individual job types,all workers in the long-term cohort for whom age and job typewere known are included in this exploratory analysis. Again,SMRs should be compared cautiously. The result is shown inTable 6.

Drying and Shipping, the area established a priori as havingthe greatest exposure to dust and radiation, appears to have anelevated risk in this analysis. The all-cancer SMR for this job

All cancer Lung cancer(/>= 0.09») (/>= 0.20»)

Oral/nasal

Number of

Drying/shippingMaintenanceChemical

and fertilizerFlotationRailroad

yardMiningOfficeworkers"38214115415412649Deaths71811193SMR2.461.230.280.310.400.970.76Deaths3500133SMR2.830.910.000.001.080.851.98Deaths1200000SMR4.541.820.000.000.000.000.00

' 637 male phosphate workers for whom the relevant data were known. No

imputed data were used.Probability of observing this distribution of deaths by job area as a result of

chance alone.

area is 2.46, with P = 0.09 for this distribution. Though the

numbers are extremely small in this analysis and none of thedistributions achieves statistical significance, it is notable thatthe direction of these results is roughly consistent with the apriori order of risk; in each case the highest SMR is found inthe Drying and Shipping area, and for All Cancers and Oral/Nasal Cancer there is an intermediate risk for Maintenance andlower risk for other job areas. This pattern does lend somesupport to a hypothesis that jobs in enclosed or dusty areas inthis industry, such as those found in Drying and Shipping, maybe associated with cancer, particularly of the respiratory tract.When this analysis is repeated separately by race (data notshown), the same apparent excess risk in Drying/Shipping isseen among blacks.

A multiple logistic regression analysis was carried out on astudy subset with 1 year or more of employment and at least15 years of follow-up, so that the role of job type and otherfactors could be compared in a population with adequate latencyfor the occupational risk factors. In Table 7, an analysis of all-

cancer mortality reveals a statistically significant elevation forthe Drying/Shipping job type as well as the temporal variables,year hired and age hired, when cigarette smoking, race, education, duration of employment and other industrial exposureswere simultaneously controlled. A multiple logistic analysis oflung cancer mortality produced similar results (data not shown),although the small number of workers with complete dataprevents firm conclusions. Because of small numbers, thesedata should be interpreted cautiously. They are, however, consistent with the a priori knowledge about exposure levels inthese jobs, and independent of any external comparison population.

To examine the possibility that the higher risk estimates incertain jobs might have been caused by a higher rate of smokingamong those workers, smoking rates were examined by jobcategory. This analysis revealed that the proportion who hadever smoked was no higher among Drying/Shipping workersthan among other workers; Maintenance workers did have ahigher, though not statistically significant, rate of having eversmoked. When amount smoked was considered, 70% of Maintenance workers were found to have smoked 10 pack-years or

more, compared with only approximately 55% of workers inboth the Drying/Shipping and other job categories. Thus, whilethese data would support a hypothesis that excess cancer ratesamong Maintenance workers might have been attributable tosmoking, they would not support a hypothesis that excess lungcancer mortality among Drying/Shipping workers was dueentirely to smoking patterns.

7301

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MORTALITY IN PHOSPHATE WORKERS

Table 7 All-cancer relative odds and adjusted rates, multiple logistic regressionMale phosphate workers employed for 1 year or more, hired before 1965."

Simultaneously adjusted for age and calendar time, and each of the variablesshown.

VariableJob

typeDrying/ShippingMaintenance

OtherSmoking

Ever smokedNeversmokedRace

WhiteBlackDuration

of employment10+ yearsLess than 10yearsEducation

8th grade or less9th grade ormoreYear

hiredBefore 19501950-1959

1960-64Age

hired40 or older39 oryoungerAsbestos

Ever exposedNeverexposedLumber

Ever exposedNever exposedRei.

odds5.3

1.9(1.0)2.1

(1.0)0.7

(1.0)1.9(1.0)1.0(1.0)8.42.2

(1.0)4.8

(1.0)2.9(1.0)1.0(1.0)M'alio2.16*

1.351.19-0.830.570.072.35*

0.862.88*1.320.02Crude

rate166.7

62.543.862.3

30.348.2

76.965.2

12.0100.8

37.998.8

40.314.3127.3

44.6117.652.590.9

52.1Adjusted

rate175.1

71.438.662.3

30.349.2

73.760.4

32.256.3

54.6106.1

30.714.0163.4

39.4136.4

51.755.8

55.0" Only individuals for whom the relevant data were known. No imputed data

were used. Includes 24 cancer deaths, 436 workers.*/><0.05.

DISCUSSION

This study has found an elevated lung cancer risk amongwhite phosphate workers employed by a Florida phosphatecompany for 6 months or more, relative to U. S. rates. Thisexcess is also seen when Florida rates are used as the standard,and when cancer incidence is examined relative to Connecticutincidence rates. No comparable elevation was found amongblack workers with respect to U. S. rates, although internalcomparisons and multivariate analyses suggest that blacks aswell as whites may be at elevated risk in certain jobs. Furthermore, a possible association with certain job areas consistentwith the a priori exposure estimates is seen both in internalcomparisons by job within the company, and in multivariateanalyses. However, small numbers in these latter analyses prevent firm conclusions.

The uranium miner experience supports the plausibility of alink between inhaled radionuclides and respiratory cancer (11,12) and for a dose-response effect with level and duration ofexposure (13-15), and temporal patterns are consistent withother observations of a 15- to 20-year latent period for themanifestation of most lung cancer.

Other investigations of cancer mortality in this industry haverecently been reported (16-18). Checkoway et al., in a largestudy of other companies in the industry, also found a healthyworker effect in blacks but not in whites, significant elevationin emphysema in whites but not in blacks, and deficits inarteriosclerotic and CNS disease, especially in blacks. A significant elevation in lung cancer mortality in whites was also seenby that study, which became statistically nonsignificant when

Florida comparison rates were used. Analyses by duration andtime since first employment are not directly comparable, sinceresults were not shown by Checkoway et al. for cross-classification of those two factors, nor were recently hired workersexcluded from their analyses. Nevertheless, their data suggestan increased lung cancer risk when sufficient time (30+ years)has elapsed since first employment, which does not disappearwhen Florida rates are used. Such an interpretation is tentative,of course, since although the SMR reaches 2.08 at the longestlatency in Checkoway's data, statistical significance is not

achieved.Stayner et al. (18) conducted a similar study in one phosphate

company. Only 10 lung cancer deaths were observed, which didnot reflect a statistically significant excess. In blacks, but not inwhites, a significant lung cancer excess was observed in thesmall subset with 20 years or more of both employment andlength of follow-up.

These observations are consistent with the observation reported here, suggestive of a dose-response of respiratory cancerwith duration of employment, among those workers with sufficient latency (Table 5). This pattern is more pronouncedamong workers for whom 20 years or more have elapsed sincefirst employment. The distribution of risk for different job typesalso could be judged to constitute a dose response, since it isconsistent with the a priori information on degree and type ofexposures.

Methodological Issues. The complete ascertainment of thepopulation at risk which is afforded by the use of all Form 941Quarterly Reports minimizes the potential biases which incomplete ascertainment can introduce. On the other hand, it increases the proportion of individuals for whom characterizinginformation is unknown. Thus, complete ascertainment coupledwith incomplete tracing information meant that questionnairedata were obtained from only about half of the long-termcohort. Much but not all of the job information could beobtained from company records, but smoking and other exposure information could not. Thus it is important to note thatinferences from Tables 6 and 7 depend on the assumption thatthose who could be contacted were representative of the wholecohort, and that this assumption may not be completely correct.However, note that much of the incomplete information wasnot due to nonresponse (92% of those contacted provided thequestionnaire data) but to the absence of tracing informationas a result of the company fire. The possible effect of suchincomplete information is discussed below.

Missing data could have introduced a last-job bias amonglong-term workers. However, the direction of the detected risk,toward the Drying/Shipping area, is not only consistent withthe a priori information, but is inconsistent with a last-job biaswhich might have been introduced if company records of senioremployees were preferentially maintained; such jobs are notseniority positions, and any bias should have been away fromDrying/Shipping and toward more "senior" jobs such as those

in the Office or Mining categories.Bias could also have been introduced in the external SMR

analyses (but not in the internal analyses represented in Tables6 and 7), in which year of birth was imputed if unknown. Themagnitude of the possible error was investigated by bracketingthe estimate and by using an unrelated method. A systematicfive-year error in the year of birth in one direction wouldproduce a lung cancer SMR of 1.40, (compared with the 1.62seen in Table 3), still statistically significant; a 5-year error inthe other direction produces an SMR of 1.81. Systematic errorsof this degree are unlikely, since 55% of the year-of-birthestimates were shown to be within 2 years of the correct year,

7302

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

MORTALITY IN PHOSPHATE WORKERS

and 78% within 5 years, in a subset of this population (3).SMRs were also calculated using an entirely unrelated imputation method, based on the age distribution by race, year hiredand duration of employment. This method yielded an estimatedrespiratory cancer SMR of 1.56 for white males, very similarto the 1.62 reported in Table 3.

Thus, the imputation of year of birth for those with missingvalues appears unlikely to have introduced a large or systematicerror in the SMR estimate. While the precise level of riskcannot be known with certainty, a statistically significant elevation in respiratory cancer SMR in the range of 1.4 to 1.8,centering on an estimate of 1.6, would appear to be a reasonablyaccurate and unbiased estimate.

It should be noted that if no imputation is used and thosewith missing values are omitted from the SMR calculation, thisresults in an estimated lung cancer SMR of 2.15, and an all-cause mortality SMR of 1.30. These values are clearly erroneously high, since they result from omitting person-years ofindividuals known to have been at risk but whose year of birthwas missing, while not omitting any deaths since their date ofbirth is available from the death certificate.

The Role of Smoking. Because smoking information wasincomplete, the relative contribution of smoking and occupation to the observed respiratory cancer excess cannot be finallydetermined. Among whites the SMR is reduced only from 1.62to 1.52 after partial adjustment for smoking, suggesting thatthe effect may not be totally explanatory. Furthermore, thedose-response seen in Table 5 suggests that duration of em

ployment exerts an effect independent of smoking. All of theworkers in the "20+ years since first hired" group in Table 5

had been first employed by this company at least 20 years ago;thus they have all had at least 20 years to be smokers, butvarying durations of employment by this phosphate company.Unless we are to suppose that they smoked only while workingfor this company, and gave it up when they moved on to anotherjob, the observation of a dose-response with length of employment, within the 20-year latency group, suggests an employment effect unlikely to be due solely to smoking.

It is also notable that the elevated risk observed in very dustyjobs in Drying/Shipping cannot be explained by smoking patterns; the available data indicate that while Maintenance category workers smoked significantly more than workers in otherareas, Drying/Shipping workers did not.

However, since no lung cancers developed among non-smokers in the group for whom smoking data were available, it wouldappear that occupational exposures in this industry are notpotent lung carcinogens in the absence of smoking. A parallelmay be seen in the uranium miner data, in which only a veryweak dose-response is seen among non-smokers, while a clearsynergistic relationship is seen between smoking and mineexposures (15).

Radiation. No attempt has been made in this study to calculate cumulative radiation doses to workers, for several reasons:no measurements were made in the industry prior to the mid-1970s; most data since then involve site or area monitoringrather than personal dosimetry; mobility is extensive, bothwithin the industry, within the plant, and from area to areawithin a particular job; and conditions have changed as technology has changed. It is notable, however, that although thevast majority of radiation measurements in the industry inrecent years have been well within the range of currently ac

ceptable levels, some elevated radon and long-lived a radiationmeasurements have been observed (4).5'6 It is possible that therisk is limited to a few jobs with very high radionuclide-bearingdust levels. On the other hand, if the risk is more general, theeffect of exposure to low levels, in conjunction with smoking,may be greater than had previously been thought. Until furtherstudy can clarify those questions, it may be prudent industrialhygiene to discourage smoking among individuals exposed toeven low levels of radiation. Cessation of smoking by membersof the public who have elevated radon exposure in the homewould also be prudent, and might reduce lung cancer risk.

The data presented here are not unequivocal. Choice of adifferent comparison population, or the selection of subpopulations with different degrees of latency, could reduce externalcomparisons below the level of significance; the degree ofimportance of smoking cannot be fully settled; and internalcomparisons in these data have been severely hampered bysmall numbers and paucity of information. Nevertheless, theconsistency of externally based SMRs regardless of the comparison group, the apparent dose-response with duration ofemployment, especially after 20 years have elapsed, and thesuggestion of an excess risk in the a priori high-risk group,make it reasonable to infer a modest but real excess risk ofrespiratory cancer among phosphate workers employed by thiscompany in the past.

REFERENCES

1. Carter, L. J. Phosphate: Debate over an essential resource. Science (Wash.DC), 209: 372-374, 1980.

2. Clayton, G. D., and Clayton, F. E. (eds.) Patty's Industrial Hygiene and

Toxicology, Ed. 3, rev., Vol 2A, Toxicology. New York: John Wiley & Sons,1981.

3. Block, <•¿�..Matanoski, G. M.. Seltser, R., et al. A method for estimating yearof birth using Social Security number. Am. J. Epidemici., II8: 377-395,1983.

4. Roessler, C. E., and Prince, R. J. Occupational radiation exposure in theFlorida phosphate industry. Jacksonville, FL: University of Florida Press,1978.

5. Monson, R. R. Analysis of relative survival and proportional mortality.Comput. Biomed. Res., 7: 325-332, 1974.

6. Breslow, N. E., and Day, N. E. Indirect standardization and multiplicativemodels for rates, with reference to the age adjustment of cancer incidenceand relative frequency data. J. Chronic Dis., 28: 289-303, 1975.

7. Haenszel, W., Loveland, D. I!., and Sirken, M. G. Lung-cancer mortality asrelated to residence and smoking histories. I. White males. J. Nati. CancerInst., 28: 947-1001, 1962.

8. Mason, T. J., and McKay, F. W. U. S. Cancer Mortality by County: 1950-1969. DHEW Pubi. 74-615. Washington, DC: National Cancer Institute,1974.

9. Stockwell, H. G., Lyman, G. H., Waltz, J., et al. Risks associated withresidence in the central Florida phosphate mining region. Am. J. Epidemiol.,128: 78-84, 1988.

10. U. S. Department of Health, Education and Welfare. Smoking and Health:A report of the Surgeon General. Washington, DC: DHEW, 1979.

11. Archer, V. E. Health concerns in uranium mining and milling. J. Occup.Med., 23: 502-505, 1981.

12. Cross, F. T. Exposure standards for uranium mining. Health Phys., 37: 765-772, 1979.

13. Lundin, F. E., Lloyd, J. W., Smith, E. M., et al. Mortality of uranium minersin relation to radiation exposure, hard-rock mining and cigarette smoking—1950 through September 1967. Health Phys., 16: 571-578, 1969.

14. Archer, V. E., Wagoner, J. K., and Lundin, F. E. Lung cancer among uraniumminers in the United States. Health Phys., 25: 351-358, 1973.

15. Archer, V. E., ( !illuni. J. D., and Wagoner, J. K. Respiratory disease mortalityamong uranium miners. Ann. NY Acad. Sci., 271: 280-293, 1976.

16. Checkoway, H., Mathew, R. M., Hickey, J. L. S., et al. Mortality amongworkers in the Florida phosphate industry. I. Industry-wide cause-specificmortality patterns. J. Occup. Med., 27: 885-892, 1985.

17. Checkoway, H., Mathew, R. M., Hickey, J. L. S., et al. Mortality amongworkers in the Florida phosphate industry. II. Cause-specific mortality relationships with work areas and exposures. J. Occup. Med., 27:893-896,1985.

18. Stayner, L. T., Meinhardt, T., Lernen, R., et al. A retrospective cohortmortality study of a phosphate fertilizer production facility. Arch. Environ.Health, 40:133-136, 1985.

7303

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from

1988;48:7298-7303. Cancer Res   Gladys Block, Genevieve M. Matanoski, Raymond Seltser, et al.   Cancer Morbidity and Mortality in Phosphate Workers

  Updated version

  http://cancerres.aacrjournals.org/content/48/24_Part_1/7298

Access the most recent version of this article at:

   

   

   

  E-mail alerts related to this article or journal.Sign up to receive free email-alerts

  Subscriptions

Reprints and

  .pubs@aacr.orgDepartment at

To order reprints of this article or to subscribe to the journal, contact the AACR Publications

  Permissions

  Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://cancerres.aacrjournals.org/content/48/24_Part_1/7298To request permission to re-use all or part of this article, use this link

on July 25, 2020. © 1988 American Association for Cancer Research. cancerres.aacrjournals.org Downloaded from