Document mK5ddq0b8bGrRx6Vw2rkbpXg

Journal ot Occupational Modicina (ISSN 009frl73$) October, 1981 V I. 23 No. 10 Editor UOYD B TEPPER, MD Associate Editor ROBERT R i HILKER, MD Executive Editor DORIS L FLOURNOY. M S Editorial Board CHARLES E BECKER, MD H DEAN BELK, MD PATRICIA Y HAGENDORN, MD HENRY R HERBERT. MD BRUCE W KARRH. M D PAUL KOTIN. MD THOMAS A LINCOLN. M D BRIAN MacMAHON, M.D. IAMES W MITCHELL. M D. DAVID B ROBBINS. M D Production Manager LUCILLE ROBERTS Manuscript Coordinator BARBARA VASUMPAUR Circulation Manager VIDA HEPLER The ioumel ot Occupational Medicine is owned, copyrighted, and published monthly by the American Occupational Medical Association, 150 North Wacker Dr, Chicago. IL 60606 Second clast postage paid at Chicago. IL. Title registered. United State* Patent Office. Subscriptions S25 00 per year, U S; $30 per year, Canada and elsewhere. Single copies $3 50 Allow four weeks for first issue Publisher cannot assume respon sibility for replacing undelivered issues, or issues damaged in transit unless notified not later than the end of the second 4 month following the month of issue. Although all advertising material ac cepted is intended to conform to ethical medical standards, acceptance does not imply endorsement by the Journal. Change of Address Notices should in clude old address as well as new Attach label from recent issue if available. Be sure to include tip codes for both old and nm* address Allow four weeks for change to become effective. Manuscripts submitted for publication should be sent to Doris Flournoy, Ex ecutive Editor, Journal of Occupational Medicine. 150 N. Wacker Drive. Chicago, IL 60606. ADVERTISING REPRESENTATION CHARLES B. SLACK. INC. Thorofire. N| 06086 (609) M6-1000 Marketing Director R1CHAR0 N. ROASH Advertising Sales SUSAN CROWLEY ' ts- POSTMASTER: I Form 357* to oDprivteio. nCahilcagMoe, dILidCnOe.MB1Mr4 .*NvaL-4 1 e0* ORIGINAL ARTICLES Cancer Mortality of a Group of Canadian Workers Exposed tq Vinyl Chloride Monomer G Theriault, M D., Dr.P H, and P. Allard, M.D., M.Sc. Mortality Among Rubber Workers III. Cause-Specific Mortality, 1940-1978 Elizabeth Oelzell, SD, and Richard R. Monson, MO. Reproductive Hazards of the Workplace L. E. Sever, Ph D. Diagnostic Sensitivity Bias -- An Epidemiologic Explanation for an Apparent Brain Tumor Excess Peter Creenwald, M.D., et al Tuberculosis Surveillance of Hospital Personnel lohn V. Gaeuman, M.D. Effects of Iron and Ascorbic Add Supplementation on Hemoglobin Level and Wotk Effidency of Anemic Women......................... . Cecilia A. Florencio, Ph D. 671 677 685 690 695 699 CURRENT COMMENTARY Responding to a Health Crisis.................................................................. Richard M. Cooper, J.D. 705 DEPARTMENTS Lettetvto-the-Editor Follow-up of Methylene Chloride Study................................................... 660 A Medical Decision Only?.......................................................................... 660 Correction...................................................................................................... 661 * OM Forum................................................................................................... 664 Book Reviews Handling Chemicals Safely 1980 ......................................................... 667 Occupational Epidemiology................................................................ . . 667 Woods Injurious to Human Health.......................................................... 668 White Collar and Professional Stress........................................... 669 Selected Reviews from the Literature................................... 710 Association Attain Summary ot the Meeting ot the Board of Directors. 714 General People and Events Classified Advertising.. ii ................... j ..... . -^'v.c-'iayv.- - - 71* 72BV (Classified Ads also Advertisers' Index.......... . . , . * . a > t * v .>** -m *_* t**.*"*^4r* ' ' 9 >'mTV MANUSCRIPTS: Contributors are requested to read Iniamttionja^ i Jishotf to the lanugry. Marknd'WtwntjeridMlef ofc ~ ~ ' from the Editorial Office upon lequnt Failure to fdlbw cause delay in the cons(deration of manuscripts for publication. OLI 6621 Diagnostic Sensitivity Bias -- An Epidemiologic Explanation for an Apparent Brain Tumor Excess Peter Greenwald, M.D.; Barry R. Friedlander. M.D.; Charles E. Lawrence, Ph.D.; Terry Hearne, M.S.; and Kenneth Earle, M.D. Preliminary data showing over-representation of the Eastman Kodak Company (Rochester) on death certificates of brain tumor patients, and higher risk for older workers when compared to the general population, led to a casecontrol epidemiologic study Chemical exposure histories of 56 workers with brain tumors were compared with those of other Kodak employees. No differences were found in exposure to a variety of chemicals. In addition, employees with brain tumors were compared to other upstate New York brain tumor patients; there was no difference in histol ogy. However, the Kodak employees'had diagnoses more frepuently confirmed by histologic examination and more thorough diagnostic studies. Thus, the apparent initial ex cess of diagnosed tumors may have resulted from a "diag nostic sensitivity bias"arising from more complete medical evaluation of Kodak employees. Hypothesis-generating studies, done coincidentally by the New York State Department of Health and East man Kodak Company in 1976, raised the question of whether there was a brain tumor excess among Kodak employees. In order to investigate these preliminary find ings. a cooperative study was undertaken by the Depart ment of Health and the company. The Health Department conducted a case-control study of 20-to-64-year-old men dying in 1968 through 1973 with brain tumors (International Classification of Disease, Eighth Revision Code -- ICDA-8:191, 238.1) in New York State, exclusive of New York City. Two controls for each decedent with a brain tumor were randomly selected from men dying from all other causes matched by five-year age group at death and the^same multi county area of the state. The study, exploratory in nature. iii From th* Division of Epdwnioloay. Nw York SUM DwMrtmont of Hwlth Toww Building, Room SO), Tho Covomor Notion A. Rocfcofollor Cmpm Suit Flit*. Albany. NY 1)2)7 (Of Cronwald. Director, and Or Lawrence Director. Operations RetearchL Eastman Kodak Company. Bide )20. Kodak Park, IBM Lake Aye, Rochester. NY 14650 (Dr Friedlander. Epidemiologist. Health. Saiety and Human Factors Laboratory, and Mr Hearne. Statistician. Management Saw vices Dhhsionl and American Regnay at Pathology. Armed Forces liMMitR at Pathology. 14th and Alaska Ave. Room 1111. Wastington. DC 20)06 (Or Earle focused on occupation Death certificates in New York State note "usual lifetime occupation" and "kind of busi ness " The name of the employer is commonly entered under kind of business for 12 major New York State com panies because of their large size and ease of identity A relative risk of four (Fishers p * 0 04) for employment at Eastman Kodak was found both by the initial analysis of 1968 through 1969 data (six cases) and by a confirmatory analysis for 1970 through 1973 (six additional cases). The possibility that this finding may have been due to an excess of brain tumors in the region of the Kodak plant prompted further study. A second case-control study identical to the first but limited to Monroe County (in-' eluding the city of Rochester) and covering the interval 1958 through 1976 was conducted. A highly significant al though less-marked relative risk of 1.62 (p = 0.004) emerged. Concurrently, Eastman Kodak investigators had been reviewing sickness-absence data. Noticeable absences due to brain malignancy prompted the estimation of risk (for both mortality and incidence) for persons working during the period 1964 through 3975 for which population denominators were available. The expected number of deaths from brain tumors (ICDA-8,191) was computed by applying age-sex-specific mortality rates for the U.S. white population, 1950-1969,' to the Kodak Rochester age-sex-specific person-years-at-risk. Incident cases were estimated from age-sex-race-specific data from the U.S. cancer survey data 1969-1971 .* Only person-years con tributed by active (or disabled) employees were included for incidence analysis since completeness of information was uncertain for persons diagnosed after leaving the company. Neither the overall standardized mortality rabo for brain malignancy (SMR 128. lower 9596 confidence limit [CL] of 91.8) nor the standardized incidence rabo (SIR 132, lower 9596 CL of 92.4) was significantly increased However, two groups did show elevated risk. 1. The SIR for women was 217 (based on 10 cases; 9596 a104 -399k i. > -V Z The incidence ratio,for the 55-to-59-year age group - (both sexes) was 281 (based on 14 cases; 959k CL *-153^- t 4A7-2mJ- ......... ~ T ri.i .J : If ^ *W OLI 6622 A relative risk declining with time explains the dif study subjects A structured exposure evaluation form ference m significance levels found in the two studies. containing the complete |ob history was provided for Thus, the two studies clearly indicated the need for fur each study subject Physicians and the investigators ther study and suggested special emphasis on women and worked with plant chemists, industrial hygienists, depart on the 55-to-59-year age group for both sexes ment supervisors, long-time employees, department per In view of the consistency in these preliminary results, sonnel officers, safety coordinators and others who might an intensive joint investigation was carried out, the results be able to assist them to identify exposures Based on a of which are reported. review of the major processes of the company and rele vant carcinogenesis literature, the 18 categories of expo Methods sure shown in Table 1 were used. Heat was considered a The general design is that of a retrospective, case-con- control variable, since no case-control difference was ex rol, epidemiologic study Cases are all Eastman Kodak pected Physicians were queried about any gaps or incon Rochester employees identified as having died with pri sistencies in the data they collected mary intracranial neoplasms* between lanuary 1. 1956, In general, it was not possible to distinguish between and December 31,1975 Two control groups of Eastman potential for exposure and actual exposure There was no Kodak employees were selected -- one from employees method of determining who actually had absorbed or in who had died from causes not including brain tumors; the gested chemicals into his body Thus, throughout this second from a historical file of full-time employees. report, "exposure" means "potential for exposure" Chemical and other occupational exposures were the key Relative risks and significance levels were calculated by study variables. the method of Miettinen.* In order to identify the case group. New York State Histopathology observed in cases was compared to death certificate files were abstracted for ail residents of that noted in other brain tumor decedents in order to see Monroe and adiacent counties whose stated cause of whether the pattern of cell types of brain tumors among death was a brain tumor during the study period. These Eastman Kodak employees differed significantly from were matched against the Eastman Kodak mortality data those of others in upstate New York. Two tumor compari file (January 1,1956, to December 31,1975) and popula son groups were used. These were obtained by matching tion data file (January 1, 1964, to December 31, 1975X cases to New York State's death certificate files for other Rochester area Eastman Kodak employees who died out persons of the same sex, five-year age group, and year of side of Monroe and adjacent counties were identified death who died with brain tumors. One tumor compari from the same file. son group consisted of residents of the nine-county Deceased controls were^elected from the company's Rochester Health Department region, excluding Kodak mortality data file (January 1, 1956. to December 31, employees; the other comprised residents of New York 1975). matched to cases on sex and five-year age group, State, exclusive of New York City and the Rochester area. and picked at random from within these groups in a 2:1 Case and tumor comparison slides were obtained from control to case ratio. Another control group, including hospital pathologists and reviewed by the neuropatholo both living and dead persons who were also matched to gist (Dr. Earlel who was blinded as to case or tumor com cases on sex and five-year age group with a 2:1 ratio, was parison status. The World Health Organization classifica chosen from the company's population data file(January tion was used, supplemented by coding of anatomic site 1,1964, to December 31,1975), This latter control group within the brain. If pathology slides could not be ob was derived from a computerized employee record sys tained. or were found not to show a primary brain tumor, tem begun on January 1, 1964, and was used only for an alternate comparison tumor was drawn using the same comparisons to cases who had worked after that time. matching criteria. The rationale for the second control group was that it Hospital records *and death certificates were ab might be more representative of the employed popula stracted, in order to compare demographic, medical and tion than controls selected from the mortality file. family data on cases with those of the tumor comparison Cases and controls were combined so that investigators group. Information about procedures used in making the searching for exposure data did not know whether an indi diagnoses was recorded in the hospital records abstract vidual was a case or a control, job histories for cases and form. controls were obtained from mortality and population files, supplemented by abstracting divisional personnel Results records. The combined case and control groups were then Case selection procedures provided a case group of 56 separated ioto divisions. Review of job titles and organi employees who had died with primary brain tumors dur zational units within the company demonstrated that ing the study period. Based on matching criteria, 112 de these were not sufficient for identifying industrial expo ceased controls were selected from the company's mor sures of the study subjects. Thus, it was necessary to have tality data file; a subgroup of 31 cases working after the occupational physicians responsible for each division January 1,1964, was matched to the company's popula determine potential exposures, including time and dura tion file, providing a second control group of 62 alive or tion, for each study subject who had worked in their divi dead persons. , . sion These occupational physicians knew neither what, ,. The exposures of the cases arid controls were analyzed disease was under study nor the case-control status of the: Sf" - ' -A.` for each set of controls. Results are shown in Table 1. No significant increases were found. With 18 hypotheses it - r- *<a>* in. H2.i, ms.rae, mi ra* ravraira* youId not be at all surprising to find one difference signif- r:% ''Journal of Occupational P _ i-jUt.:h i A' vw OLI 6623 Tibi* 1. - Exposurc of Caste Dying witb Tumor* tho Central Nervous Syttam (1958-1975) and Matclwd Control*. _________ Tbo Numbtr Exposed Rotors to Those Ever Having tbo Potential lor Eaposur* White Employed. Exposure Citsgory Chlorinated solvents Other solvents Plasticizers Amines Oil Mist Other and mixed cnemicals Silver Other heavy metais Radiation Heat Animal tissue Color forming agents (dyes and couplers) Slack and white developers Color developers Hardeners Wetting and spreading agents Photographic sensitized product Film and photographic paper dusts He. et Cates Exposed 16 32 8 7 7 13 10 9 0 9 13 6 7 6 7 S 5 4 AD Casas (N 68) vs. Dead Cantrali (H - 112) No. el Oeed Controls Exposed Keltttw Risk (MK Cl) 36 0 82 (0 44-1 55) $6 1 40 (0 77-2 55) 18 0 86 (0 39-1 89) 13 1 09 tO 47-2 53) 21 0 65 (0 31-1 36) 25 1 06 (0 54-2 09) 21 0 93 (0 44-1 97) 24 0 69 (0 33-1 43) 6 -- (0 00-0 95) 21 0 79 (0 36-1 77) 22 1 32 (0.63-2.80) 12 1.00 (0 39-2.54) 11 1 31 (0.56-3 07) 6 2.16 (0.76-5 93) 18 0 73 (0.33-1 65) 13 0.75 (0.30-1 86) 13 0 72 (0.27-1 88) 8 1.00 (0 37-2 74) Casa* oltn Iasi Employment 1964 or Lstar (N 31) n. Populous* Contrail IN 62) He. el Coses Exposed No. el PepControls Expssed Rftfttff *b* 1*5* CL) , 7 13 1 10 (0 45-2 71) 18 32 1 40 (0 61-3 26) 5 7 1 55 (0 53-4 54) 3 5 1.23 (0 34-4 49) 4 7 1 16 (0.39-3 43) 8 19 0.78 (0.34-1 79) 5 7 1.48 (0 54-4 08) 5 17 0 48 (0 19-1 25) 0 5 -- (0 00-1 25) 6 10 1.23 (0 49-3 08) 6 11 1.09 (0 47-2 51) 2 4 1.00 (0.24-4 16) 5 6 226 (0 63-8 15) 4 6 1 50 (0 43-5 27) 4 7 1.23 (0,34-4 49) 2 4 1 00 (0.24-4.16) 1 4 050 (0.08-3 15) 2 4 1.00 (0.21-4.76) icant at the 0.05 level. Thus the significant decrease in risk observed for radiation exposure (which included work in rooms with only very low-level radiation source, eg., tritium buttons) is well within the realm of chance. Com parison of time of exposure, duration of exposure and exposure-to-diagnosis intervals among exposed cases ver sus exposed controls also showed no differences. The case group included 18 women. Twelve of the 18, including four of six in the 55- to 59-year age group, had had no exposures. The "other solvents" category, with four cases, was the only category containing more than one woman. The relative risk estimates for this category were less than one. The 55- to 59-year age group of men included eight cases. Again there were no clusters in any exposure category, and the extent of exposure for con trols was about the same as for cases. Few of these work ers had been located in any one setting. A review of age and employment data showed no evidence of a cohort ef fect Findings on the histological type of tumor are shown in Table 2. Here, cases are compared to two groups of other upstate New York residents diagnosed as having brain tumors. It can be seen that the diagnosis was pathologi cally confirmed and reconfirmed by the review patholo- Tible 2. - Histologic Type* of Eastman Kodak Cast* and Tumor Comparison Group* far Whom SMas Wara Avaltabte far RaaxsnKnation. Numbers in Each Category Art Shown. Cai Type (Wortd HoeM Oraatexteten CbssMcattM) Kadtk Gliooiastome 32 - Giant Call Glioblastoma 3 Glioblastoma with sareonwtous 4 Pnmrtlve polar spongtoteMtoma Anaplashc istroc^toma 4 Astrocytoma, protoptaMUe 3 Astrocytoma, gamistocytic Mixed otigo-astrocytona Oligodendroglioma Anaplastic oUgodondrogtani Maduiiootastoma Chromophobe adeno"ma __ , ,* ' Naurttemmoraa *'v Cranlapharynglano m.v^-y-v. .. ik. * k , J*-' Martngothiflomitoua -- fibrous noninglamn:^SOLaitAi^if^sm w%t-, . .Tramltional maningiomg - . ToOt.-aT ..--,n.. .... - . , Csopwlnn 30 2 J. 2 '4 1 2 * * w .* M 'SniwltMty :Expli 26 2 3 '1 : s i 1 2 t OLI 6624 Tbl* 3. - Ffiqmncy o( PstMagy Conflrmitl*B and Diagnostic Procadurii m Cam and Tumor Campanian Groups. Coded as Brain Tumor on Deatn Certificate no hospital records available Not cancer or metastatic cancer Clinical diagnosis (no oatnoiogyl Pathologically con- tirmedt Pathological or Clinical Diagnosis of Brain Tumor Craniotomy Oram surgery or biopsy Bram scan Angiogram EEG Pneumoencephalogram Autopsy Skull x-ray Eastmee K*dt> Can* Tolll NO.* 64 3t 4 7 64 7 10 9 64 3 4 7 64 51 79 7 54 48 88 9 54 33 61 1 54 41 75 9 54 20 37 0 54 19 35.2 54 2B 51.9 54 50 92.6 RecftesW ina Tamar Cempsrisen Bnupt Total No.* % * Statewide Tumor Campanian Greups TiW N#.* % P 84 15 17,9 0 036 84 10 11 9 0 623 84 7 83 0 381 84 52 61 9 0 019 too 13 13 0 0 166 100 17 17 0 0 170 too 10 10 0 0 219 100 60 60 0 0 008 59 49 83.1 0 373 59 25 42 4 0 046 59 40 67 8 0.338 59 20 33 9 0,727 59 9 15.3 0.014 59 28 47 5 0782 59 50 847 0 191 70 53 75 7 0 061 70 21 30 0 0 0005 70 45 64 3 0 163 70 23 32 9 0 627 70 12 17 1 0 021 70 24 34 3 0 049 70 67 95 7 0 455 'Number that hospital records indicated procedure was performed tTwo ot these are included with the 54 pathologically or clinically diagnosed cases lor the study el exposures. giving the 56 cases shown m Table i (see text for details) tot the pathologically confirmee cases, slides could be obtained tdr review on 51 cases. 46 Rochester area matched controls, and 44 statewide matched controls as shown in Table 2 p - Significance levels lor differences in proportions having procedures between Kodak cases and comparison groups gist for 51 of the 56 cases listed in Table 1. Three cases were based on a clinical diagnosis: The diagnosis for one patient was supported by a ventriculogram and a cranio tomy showed necrotic tissue unsatisfactory for histologi complete hospital records were available and obtained. It can be seen that the cases had a significantly higher fre quency of brain scans and pneumoencephalograms than did either tumor comparison group. It is noteworthy that cal examination; for the second by a brain scan; and for the third by clinical signs of a brain mass and an autopsy that excluded a non-brain primary but permission was denied to examine the brain. Two additional cases were included on the basis of death certificate reports, when all eight procedures were more commonly performed in cases than in Rochester-area controls and that seven of eight procedures were more commonly performed in cases than in statewide controls. It should be noted that the study covered a period before the availability of com no hospital records could be located. One certificate puted tomography. ytated "glioblastoma multiforme of left cerebellum' and With one exception, no family history or demographic the other, "right temporal brain tumor." Another death differences were observed between the cases and the certificate report noting "brain tumor" was excluded; this tumor comparison groups. The hospital records indicated report had come from a nursing home and no other infor that three study case? had family histories of epilepsy, mation could be found. while the less complete hospital records of those in the The distribution of central nervous system tumors by tumor comparison groups showed no such family history. cell type showed no differences between cases and either tumor comparison group. Clioblastoma and its variants Ditcuaion accounted for about 70% of diagnoses for the case and Employees in this study had a more thorough medical each tumor comparison group. There were no case-com diagnostic examination than did either of the tumor com parison group differences in anatomic location of the parison groups. That is, workers in this large company brain tumors. who died with brain tumors had pathologically confirmed A significant contrast is seen in the extent of pathologi diagnoses and sophisticated diagnostic procedures more cal confirmation for Eastman Kodak cases when they are frequently than did other brain tumor patients living in compared with those in the Rochester area or statewide the same region or scattered throughout upstate New tumor groups (Table 31 This finding of a higher percent York. Such differences in diagnostic effort may account age of histologically confirmed brain tumors among East for the initial observations that led to the study. This man Kodak employees suggests that diagnosis may be study showed an absence of exposure differences be more thorough for these workers than for either tumor tween employees with and without brain tumors. Comparison group. The top section of Table 3 refers to Age-specific incidence rates for brain tumors appear to those coded on death certificates as having brain tumor* peak in the 5S-to-64-year age group and then declines*, mther than the final study grtxips, Tabie 3 also shows the most population based reports,* but the incidence coo- frequency of diagnostic procedures for the cases and for ' tinues to rise with age in Rochester, Minnesota (locatordecedents jn the tumor comparison groups for wboni v^of the Mayo Clinic).* Schoenbergef al* offer as a possil ` Journal of Occupational Modicine>W. 23, Wl iOfDctobnMraf explanation that more complete case ascertainment and statistics difficult This is called the "healthy worker ef higher autopsy rates in the Mayo Clinic area account for this difference Higher rates are seen for all age groups of adults, although the difference is most marked m the elderly It appears that many brain tumors in the general fect." In order to evaluate the possibility of a "diagnostic sensitivity bias" in occupational studies, we suggest the use of diseased comparison groups and detailed attention to the level of diagnostic medical procedures used population are undiagnosed or misdiagnosed. Retrospec Furthermore, this bias may have an importance outside tive industrial or other studies that determine expected of the occupational setting It could, for example, distort rates on the basis of general population data thus may the perception of the influence of socioeconomic vari show artificially high rates because of underdiagnosis in ables or geographic patterns On the other hand, this bias the general population It remains to be seen whether may be important only in a few specific studies, such as computed tomography will change this in the future this one The extent and importance of the diagnostic sen The improved diagnosis and case ascertainment of sitivity bias will be resolved only through further study workers under certain circumstances of employment such as observed here might be called the "diagnostic sensitivity bias " Sensitivity is the probability that persons with the disease are correctly classified, and the bias comes from a difference in sensitivity between the study population and the non-exposed comparison population This bias would appear to pertain especially to conditions Wf jr# indebted to Or Abuhgm M Ld^nfHd tor wving 4\ 4 consultant >n rhi* ytudv W* arg indebted to Or* Noffnjn Aiht'tbu'l Robert L Raleigh lulian AJ*ader Albert Anderson, Ms Shirlev H*J| god the medical staff at the East' man Kodak Company and Manor* lackson Dr William Surnett and others at the s** York State Department of Health for their assistance, to Dr Albert Paulson tor statistical assistance, and to Dr Lowell W Lapham and many pathologists and hospitals throughout New York State for providing pathology that are difficult to diagnose, such as brain tumors, and to industries where workers have such benefits as health in- slides and hospital records information and for the reporting of cancer to the New York State Cancer Registry surance and high-quality employee medical services with referral and follow-up Of course, the opposite could oc cur in an industry with substandard medical services or in a community where the disease is overdiagnosed The References I Mason Tl. McKay FW. Hoovar R, et l Atlas of Cancer Mortality , for U S Counties 1950-1969 National Cancer Institute. DHEW Publi cation No (NIH) 75-780 bias would also apply to nonindustrial studies having this 2. Cutler SI and Young JL |r Third National Cancer Survey Inci type of difference in diagnosis. A method for detecting this potential bias may be an essential part of many industrial and other studies. Ideal ly. one would detect this bias throQgh the identification of dence Data National Cancer Institute Monograph 41, March 1975 DHEW Publication No (NIH) 75-787 3 Miettinen 0$. Individual matching with multiple controls in the case of all or none response Biometries 21-339-3$$. 1969. 4 Schoenberg BS. Christine BW, and Whisnant IV. The descriptive persons with the disease under study who were misdiag nosed as having other diseases. In reality this ideal ap proach may rarely be feasible. Thus, surrogate measures will often be required. In this study, the diagnostic sen- epidemiology of primary intracranial neoplasms. The Connecticut ex perience Am / Epidemiol 104 499-510, 1976. 5 Percy AK, Elveback LR, Okazaki H. et al. Neoplasms of the ner vous system- Epidemiologic considerations. Neurology 22 40-48.1972 6 Tabershaw IR and Caffey WR Mortality study of workers in the sitivity bias was revealed through the use of tumor com manufacture of vinyl chloride and its polymers. I Occup Med parison groups to examine histopathological and diagnos tic procedures Such diseased comparison groups rarely have been used in epidemiologic research. In this study they only provided partial information about the extent 16509-518, 1974. 7. Thomas TL and Decoufle P Mortality among workers employed in the pharmaceutical industry: A preliminary investigation. I Occup Med 21619-623, 1979 8. Thomas TL. Decoufle P. and Moure-Eraso R: Mortality among of bias, as any brain tumors that may have been misdiag nosed were still undetected. The medical literature contains several reports of ap parent industrial excesses of brain tumors that might be workers employed in petroleum refining and petrochemical plants. I Occup Med 22:97-103.1980. 9 Mancuso TF, Ciocco A. and El-IXtlar AA. An epidemiological ap proach to the rubber industry -- A study based on departmental ex perience I Occup Med 10:713-232.1966. influenced by medical diagnosis and ascertainment dif ferences. but study designs do not allov for assessment of this possibility ` *1 Studies showing no general excess could be influenced in a reverse manner if industrial detection is poor. Rubber industry worker studies*'1 have 10. Fox Al. Linders DC, and Owen R: A survey of occupational cancer in the rubber and cablemaking industries. Results of a fiveyear analysis. 1967-71 Br I Ind Med 31:140-151.1974. II McMichael A), Spirtas R, and Kupper U,: An epidemiologic study of mortality within a cohort of rubber workers, 1964-72.1 Occup Med 1*458-464.1974. shown conflicting results suggesting that tumor compari son groups might be vital if we are to distinguish the possibilities of artifact of diagnosis from true exposure ef fect A study'* in another, large company with an exten 12. Andjelkovic 0, Taulbee), and Symons M Mortality-experience of a cohort of rubber workers, 1964-1973.1 Occup Med 18:387-394, 1976. 13 Memon RR and Nakano KK: Mortality among rubber workers I White male union employees in Akron, Ohio. Am /'Epidemiol sive medical program also reported rates that were some what higher for brain tumors than for other cancers. Higher brain tumor rates have been reported for upper socioeconomic levels, particularly professional and man 103 284-296, 1976. 14. Pell S. O'Berg MT, and Karth BW: Cancer epidemiologic surveil lance in the DuPont Company. I Occup Med 20:725-740.1978. 15 Registrar Ceneral's decennial supplement for England and Wales, 1970-72 Occupational mortality. Saha* OS no. 1. London: Her agerial", this increase may be attributable to better diagcno*s*>- -7 (..The diagnostic sensitivity bias adds to the list of prob lems associated with occupational studies. For example,* Majesty's Stationery Office, 1978. IB McMichael A|, Haynes SC, and Tyrofer HA. Observations on the evaluation of occupational mortality data, t Occup Mad 1712*131.197V-.; -' ; -V./ , : - ' 7 . 17. McMkhaef A- Standardized mortality ratios and die "Healthy ,'McMichaei et a!'*,r described the selection of healthy ; Worker. CttRcCVSfratthin*' baneath; the-surfac* f. Occup Mad Sfear***.by incfcjstiy. fts^ung^cp^er^on^to ii * ^ ~ ' ___ '7 -__;77_ j______' vw r ` ' f-: > tWXagnoosyltic Sensitlvfty Bias tpildm|oliBSc fexpanatioiCiSrewwad et al OLI 6626