Lung cancer in males: an epidemiological study in northern Sweden with special regard to smoking and occupation

UMEÅ UNIVERSITY MEDICAL DISSERTATIONS

N e w Series N o 167 - IS SN 0346-6612

From the D epartm ent o f O ncology, U niversity o f Umeå, Umeå, Sweden

Lung cancer in males

A n epidemiological study in northern Sweden w ith special regard to smoking and occupation

by

LENA DAMBER

Umeå 1986

Lung cancer in males

A n epidemiological study in northern Sweden with special regard to smoking and occupadon

AKADEMISK AVHANDLING

som med vederbörligt tillstånd av re k to rsä m b e te t vid Umeå Universitet för avläggande av doktorsexamen i medicinsk vetenskap

kommer a t t offentligen försvaras i hörsal F, Humanisthuset, Umeå Universitet torsdagen den 29 maj 1986, kl. 09.00.

av LENA DAMBER

Umeå 1986

i

UMEÂ UNIVERSITY MEDICAL DISSERTATIONS New Series No 167 ISSN 0346-6612

ABSTRACT

Lung cancer in males. An epidemiological study in northern Sweden with special regard to smoking and occupation.

Lena Damber, D epartm ent of Oncology, University of Umeå, S-901 87 Umeå, Sweden.

In a case-control study comprising 589 cases of male lung cancer in northern Sweden longitudinal data concerning occupations, employments and smoking habits were collected by questionnaires.

Pipe smoking was as common as c ig arette smoking and gave very similar relative risk. The pipe smoking cases, however, had significantly higher mean age and mean smoking years at the time of diagnosis than the cigarette smoking cases. In ex- smokers, the relative risk gradually decreased from 5 years a fte r smoking cessation but this decrease was much less pronounced in ex-pipe smokers than in ex-cig arette smokers. High relative risks were obtained for small cell and squamous cell

carcinomas. For adenocarcinoma the relative risk was considerably lower but still significantly increased. The population etiologic fraction attributable to smoking was about 80% in this series.

Some occupational groups (underground miners, copper smelter workers,

electricians and plumbers) exposed to previously known lung carcinogenic agents had considerably increased odds ratios, which persisted a fte r adjustment for smoking. A slightly elevated odds ratio was observed in a group of blue collar workers potentially exposed to lung carcinogenic agents but this elevation generally disappeared a f te r adjustm ent for smoking. For two specific subgroups, asphalt and concrete workers and pulp workers, the overrisk persisted after adjustment for smoking. Farm ers and foresters had strikingly low odds ratios, which could only partly be explained by their more moderate smoking habits. The population etiologic fraction attributable to occupation was in the reported m aterial assessed to 9 per cent.

Professional drivers had higher average tobacco consumption than non-drivers, which explained the slightly increased crude odds ratio found for the occupational group as a whole. Smoking drivers in an upper age group (70 and over), however, had a high relative risk of lung cancer, while in a lower age group (under 70) no significant increase was found. The results in the older age group suggested a multiplicative effe c t between smoking and the occupational exposure.

The study clearly verified the increased lung cancer risk in underground miners. An obvious dose-response relation was found with high risk a fte r long time exposure.

All analyses concerning underground miners suggested an interaction of a multiplicative type between underground mining and smoking in the causation of lung cancer. The cases of small cell carcinoma among the underground miners had shorter average latency time and in co n trast to the other part of the m aterial, shorter average age than the cases with epidermoid cancer.

Key-words: case-control study, epidemiology, lung cancer, occupational risk factors, smoking, radon, cell types, validity.

ISBN 91-7174-240-9

UMEÅ UNIVERSITY MEDICAL DISSERTATIONS

N ew Series N o 167 - ISSN 0346-6612

From the Departm ent o f O ncology, U niversity o f Umeå, Um eå, Sweden

Lung cancer in males

A n epidemiological study in northern Sweden w ith special regard to smoking and occupation

by

LENA DAMBER

Umeå 1986

ABSTRACT

Lung cancer in males. An epidemiological study in northern Sweden with special regard to smoking and occupation.

Lena D am ber, D e p a rtm e n t of Oncology, U niversity of U m e å , S-901 87 U m e å , Sweden.

In a c ase-co n tro l study com prising 589 cases of male lung c a n c e r in n o rth e rn Sweden longitudinal d a ta concerning o ccupations, e m p lo y m en ts and smoking habits were co llected by q uestionnaires.

Pipe smoking was as common as c i g a r e t t e smoking and gave very sim ilar re la tiv e risk. The pipe smoking cases, how ever, had significantly higher m ean age and mean smoking years at the tim e of diagnosis than the c i g a r e t t e smoking cases. In ex- sm okers, th e re la tiv e risk gradually d e c re a s e d from 5 years a f t e r sm oking cessation but this decrease was much less pronounced in ex-pipe sm okers than in e x - c i g a r e t t e sm okers. High r e la tiv e risks w ere obtained for small cell and squamous cell

carcinom as. For a d e n o carcin o m a the r e la tiv e risk was considerably lower but still significantly increased. The population etiologic fra c tio n a t t r i b u t a b l e to smoking was about 80% in this series.

Some occupational groups (underground m iners, copper s m e l t e r workers,

e le c tric ian s and plumbers) exposed to previously known lung carc in o g e n ic agents had considerably increased odds ra tio s , which persisted a f t e r a d ju stm e n t for smoking. A slightly e le v a te d odds ratio was observed in a group of blue collar workers potentially exposed to lung carcinogenic agents but this elevation generally disappeared a f t e r a d ju s tm e n t for smoking. F or tw o specific subgroups, asphalt and c o n c re te workers and pulp w orkers, the overrisk persisted a f t e r ad ju stm e n t for smoking. F a r m e r s and f o r e s te r s had strikingly low odds ra tio s , which could only partly be explained by th e i r more m o d e r a te smoking h ab its. The population etiologic fr a c tio n a ttr i b u t a b l e to occupation was in the re p o rte d m a te ria l assessed to 9 per c e n t.

Professional drivers had higher average to b a c c o consum ption than non-drivers, which explained the slightly in creased crude odds ratio found for th e o ccupational group as a whole. Smoking drivers in an upper age group (70 and over), how ever, had a high re la tiv e risk of lung c a n c e r, while in a lower age group (under 70) no significant increase was found. The resu lts in the older age group sug g ested a m ultiplicative e f f e c t b e tw e e n smoking and th e occu p atio n al exposure.

The study clearly verified the increased lung c a n cer risk in underground m iners. An obvious dose-response r e la tio n was found with high risk a f t e r long tim e exposure.

All analyses concerning underground m iners suggested an in te r a c tio n of a m ultiplicative type b e tw e e n underground mining and smoking in th e c a u sa tio n of lung c a n c e r. The cases of sm all cell c a rc in o m a among the underground m iners had sh o rte r av erag e laten cy tim e and in c o n t r a s t to the o th e r p a r t of th e m a te r ia l, s h o rte r av erag e age than the cases with epiderm oid c a n c e r .

Key-words: case -c o n tro l stu dy, epidemiology, lung c a n c e r , occupational risk

fa c to r s , smoking, radon, cell ty p es, validity.

To Jan n e

H elena, Sara and Erik

ORIGINAL PAPERS

This thesis is based on the following papers:

I. D am ber L, Larsson L-G: Smoking and lung c a n c e r w ith special re g a r d to type of smoking and ty p e of c a n c e r . A c a s e -c o n tro l study in north Sweden. Br J C a n c e r 1986; 53: In press.

II. Dam ber L, Larsson L-G: O ccupation and male lung c a n c e r . A c a s e - c o n tr o l study in no rth ern Sweden. Subm itted for publication.

III. Dam ber L, Larsson L-G: Professional driving, smoking and lung c a n c e r : a c ase r e f e r e n t study. Br J Ind Med 1985; 42: 246-252.

IV. Dam ber L, Larsson L-G: Combined e f f e c t s of mining and smoking in th e causation of lung c a rc in o m a . A c a s e -c o n tro l study in n o rth e rn Sweden. A c ta Radiol Oncol 1982; 21: 305-313.

V. Dam ber L, Larsson L-G: Lung c a n c e r in m ales and ty p e of dwelling. An epi­

demiological pilot s tu d y . A ccep ted for publication in A cta Radiol Oncol.

VI. Dam ber L, Larsson L-G: Underground mining, smoking and lung c a n c e r : A c ase -c o n tro l study in t h e iron ore m unicipalities in n o rth e rn Sweden. JNCI 1985; 74: 1207-1213.

In the t e x t th ese papers will be r e fe rre d to by th e ir Roman num erals.

C O N T E N T S

INTRODUCTION

Page

MATERIAL AND METHODS 8

Selection of cases and controls. Data collection. 8

Statistical methods 9

METHODOLOGICAL CONSIDERATIONS 12

Representativity of the cases 12

Microscopic types of cancer 12

Completeness and validity of the data 13

Smoking habits 13

Occupation 14

Municipality 15

Building material 16

Comparison between cases and controls 17

RESULTS 18

Smoking and lung cancer (Paper I) 18

Occupation and lung cancer (Paper II) 19

Professional driving, smoking and lung cancer (Paper III) 21 Radon, smoking and lung cancer (Papers IV, V, VI) 23 Can the findings explain the geographical incidence differences? 25

COMMENTS AND CONCLUSIONS 29

Selection of controls 29

Reliability of data 30

Results 31

ACKNOWLEDGEMENTS 34

REFERENCES 35

INTRODUCTION

In many co u n tries the lung c a n c e r incidence r a t e has in creased rapidly during r e c e n t d ecades. Large d iffe re n c e s in incidence or m o r t a l i t y r a te s exist b e tw e e n d if fe re n t p a rts of the world and have also been observed within several c o u n trie s (1, 2, 3, 4); generally higher r a te s have been found in larg e m etro p o lita n regions than in urban a reas. R em ark ab ly low r a t e s have been observed in special religious groups as Mormons and Seventh-D ay A d v e n tists in th e USA (3, 6). As for m ost types of m alignant tu m ours such d iffe re n c e s a re a t p r e s e n t reg ard ed mainly as expressions of d iffe re n c es in life -sty le a n d /o r o th e r en v iro n m e n ta l fa c to r s .

Tobacco smoking is recognized today as the dom inating cause of lung c a n c e r . This is supported by many epid em iological in v estig atio n s (cf 7). In several W estern c ountries, such as the US, the UK and Sweden, it has been e s tim a te d t h a t smoking is responsible for about 80-90 p e rc e n t (population etio lo g ic fa c to r) of all lung c a n c e r cases in males (8, 9, 10). A dose-response relatio n sh ip betw een c i g a r e t t e smoking and th e risk of lung c a n c e r has been well e sta b lish e d in several s tu d ie s (cf 7).

Several ch em ical and physical a g en ts are id en tified as lung carcinogenic and can be of considerable im p o rta n c e for specific o ccu p a tio n a l groups. These ag e n ts include radon, a sb esto s, chrom ium , arsenic, nickel, m u s ta rd gas, bischlorom ethyl e t h e r and polycyclic hydrocarbons in soot, t a r and oil (8). Some of th e s e agents a p p e a r to in t e r a c t synergistically with to b acco smoking (11, 12, 13, 14, 13).

G eneral air pollution (in th e form of polycyclic hydrocarbons) has been much discussed concerning lung c a n c e r risk. F o r se v e ra l reaso n s, the size of the risk has been d iffic u lt to assess by epidem iological m eth o d s. H o w ever, for the individually exposed persons, it seem s, to be low c o m p a re d to smoking and some occu p atio n al exposures (8). Another risk f a c t o r much discussed in r e c e n t years is passive smoking (environm ental to b a c c o smoke). F r o m a t h e o r e t i c a l point of view,

exposure to passive smoking ought to in c re a s e th e lung c a n c e r risk, and this is also suggested by some epidem iological stu d ies (16, 17).

Ionising rad iatio n is an epidem iologically fairly well v e rifie d cause of lung c a n c e r

(18) both as reg ard s photon radiation (x-rays, y -rays) and a - r a d ia tio n (radon

daughters). E xperience from underground m in ers exposed to high c o n c e n tr a tio n of

radon d aughters gives convincing evidence about the lung carcinogenic e f f e c t of

this ty p e of exposure (19). During r e c e n t y e a rs in creased c o n c e n tra tio n s of indoor

radon (due to building m a te r ia l or leakage from the ground) have caused concern about th e lung c a n c e r risk. Some epidem iological stu d ies suggest a risk of increased indoor radon co n c e n tra tio n s but, so fa r t h e r e is no firm epidem iological evidence concerning the size of th is risk.

O ther suspected risk f a c to r s for lung c a n c e r t h a t can be m entioned are h ere d ita ry d iffe re n c es in th e en z y m a tic m odification of some c a rc in o g e n ic substances (20) and a dietary lack of p r o te c tiv e substances as vitam in A (21).

The relationship b etw een risk f a c to rs and a specific disease such as lung c a n c e r, can be elu cid ated by several types of epidem iological in v estig atio n . H ypotheses can be g e n e ra te d by descriptive epidemiologic stu d ie s showing geographical incidence variations, d iffe re n c es b etw een ethnic groups or tim e - r e l a t e d trends. C o rrelatio n studies (for instance b e tw een lung c a n c e r incidence and smoking habits) can give support to a causal in te rre la tio n . H ow ever, the most convincing evidence for an association b etw een specific risk f a c to r s and lung c a n c e r has been obtained using cohort studies and c a se -c o n tro l studies. Both types of studies have been extensively used in o rd er to verify or e lu cid ate d if f e r e n t risk f a c to r s for lung c a n c e r.

The p resen t re p o rt is exclusively based on c a s e -c o n tro l stu d ies in which d a ta were obtained by questionnaires. The incentive for th e stu d ies was th e finding t h a t the male lung c a n c e r incidence r a t e within the no rth ern region of Sweden showed large geographic variations, with very low r a t e s in most m unicipalities in th e inner p art of the region and considerably higher r a t e s in m ost c o a sta l m unicipalities and in the 2 m unicipalities Kiruna and Gällivare with ex ten siv e iron ore mining (22, 23). It was f e lt t h a t a large ca se -c o n tro l study m ight yield in fo rm atio n of in te r e s t

concerning etiologic f a c to rs , p artic u la rly smoking and occu p atio n al exposures.

The principle aims of the study were:

- to analyse the e f f e c t s of smoking

- to analyse occupational risk f a c to r s and the in te ra c tio n betw een such f a c to rs and smoking

- to study the possible e f f e c t s of building m a te r ia l in dwellings (with re g a rd to

risks from radon)

MATERIALS AND METHODS

Selection of cases and controls. Data collection.

P a p e r I - V.

The study com prised m ale p a t i e n t s with lung c a n c e r diagnosed in the t h re e m ost n o rthern counties in Sweden. A survey of th e Swedish C a n c e r R eg istry re v ealed 660 cases re p o rte d froin th e s e c o u n tie s from 1972 to 1977. F o r e th ic a l reasons,

however, only th e 620 c a s e s who had died a t le a st one y e a r befo re the s t a r t of the study (May 1979) w ere included in the investig atio n . Sixteen cases for whom no close r e la tiv e s could be found w ere also excluded before th e s t a r t of th e study. To each of the rem aining 604 c a se s, one d eceased control was drawn from the

N ational R eg istry for C a u se s of D eath , m a tc h e d acco rd in g to sex, year of death , age and m unicipality. Lung c a n c e r cases and suicides w ere not a c c e p te d as c ontrols. The following c a u s e s of d eath w ere re g is te re d am ong the controls:

m alignant tu m o u rs 20%, c a rd io v a s c u lar diseases 45%, c e re b ro v a s c u la r diseases 9%

and o th e r causes 26%. This d e a th cause p a t t e r n did not e ssen tially d e v ia te from the general p a t t e r n . A living co n tro l was also included for e ach case, and this control did not e x ceed an age of 80 y ears. The living controls (467 controls) w ere drawn from the N ational Population R egistry and m a tc h e d a g a in st the cases according to sex, year of birth and m u n icip ality .

Individual in fo rm atio n c o n cern in g several f a c t o r s was c o lle c te d through postal questionnaires for all c a s e s and controls. The q u estio n n aires w ere answ ered by close r e la tiv e s of the c a se s and of the d eceased controls and by the living controls th em selv es. Both among c a s e s and d ec e a se d controls th e s u rro g a te respondent was for 50% the w ife, 35% an a d u lt child, and 15% o th e r r e la tiv e s . The respondents were by an in tro d u c to ry l e t t e r asked for p a rtic ip a tio n and inform ed t h a t th e study concerned re la tio n b e tw e e n e nvironm ent and health.

The questio n n aire was c o n s t r u c t e d to yield longitudinal d a t a (tim e periods) c o n c e r­

ning:

1. m unicipality and type of dwelling (building m a te r ia l and co n stru ctio n year).

2. occupation and e m p lo y m e n t held for a t le a s t 1 year w ith n am e of company.

3. smoking habits including year of s t a r t , daily number of c i g a r e t t e s , o th e r types of smoking and y ear of possible cessation.

The questio n n aires answ ers w ere sc ru tin iz e d by a in te r v ie w e r not physically

a sso c ia te d with th e d e p a r t m e n t and not knowing w h e th e r th e answers concerned

cases or co n tro ls. A ccording to defined c ontrol c r i t e r i a s th e answers were

supplem ented by telep h o n e interview s. All questionnaires with incom plete smoking d ata were s u p p lem en ted by telephone interview s, which w ere required in th e sam e proportion (about 30%) among both cases and controls. F u r th e r m o r e , all

questionnaires with lack of d a ta on m unicipality, type of dwelling, occupation or em ploym ent for any period betw een age 20 and the tim e of diagnosis (or

corresponding tim e for th e controls) w ere su p p lem en ted . The occupational history required s u p p le m e n ta tio n in about 40% of the cases and the deceased controls, and in about 30% of th e living controls.

Answers w ere obtained in 589 of 604 cases (98%), in 582 of 604 deceased controls (96%) and in 453 of 467 living controls (97%).

The inform ation obtained in the questionnaires was coded. For the occupations a 5- digit code was used (24). Special codes w ere used for th e s p ecific com panies.

P a p e r VI

Lung c a n c e r cases re p o rte d from the m unicipalities of Kiruna and Gällivare during the period 1972-1982 w ere included in this study. Five c ases w ithout r e la tiv e s were excluded before the s t a r t of the study. The study m a te r ia l then consisted of 69 cases, e ach with one d e c e a se d control. Sixty of these c ases also had one living control. The selectio n and m atching p rocedures for the con tro ls were the sam e as described for paper I-V.

The sam e type of questionnaires and p rocedures of d a ta collection were used in this study as in th e previously described stu d ies. Answers w ere obtained for all persons with the exception of two deceased controls.

Statistical methods

The e s tim a tio n of odds ra tio s was perfo rm ed with dissolved m atching in all studies.

Several e s sen tial resu lts w ere, however, c ontrolled by parallel analyses with

individual m atching (25). The e f f e c t of the m atching was also evaluated by

examining th e c o rre la tio n of exposure b e tw een cases and th e ir m a tc h e d controls

(26). A ccording to M iettin en (27) the m atch in g can be ignored in the analyses if this

co rre la tio n c o e ffic ie n t is low. On th e whole, low c o rre la tio n c o e ffic ien ts were

obtained in the p re se n t study, and the size of the two risk e s tim a te s were very

sim ilar (Table 1). Accordning to Seigel and Greenhouse (28) th e ex p e c te d odds ratio

resulting from an analysis incorporating th e m atching is always more e x tr e m e than

the e x p e c te d odds ra tio obtained ignoring the m atch in g . A ctually, a tendency

to w ards so m e w h a t higher e s t i m a t e s was o b tain ed with individual matching in the p re se n t study.

Table 1. C om parison b e tw e e n re la tiv e risks c a lc u la te d with and without dissolved m atch in g

Absolute value P a p e r F a c to r Odds ra tio s e s tim a te d with

of the c o rre la tio n individual dissolved

c o e f f ic ie n t m atch in g matching

P a p e r I smoking in

0.05 general 8.1 7.3

P a p e r II plumbers and e l e c t r i -

0.10 cians 1.7 1.9

P a p e r II pulp

0.08 workers 1.8 1.7

P a p e r III professional

0.04 driving 1.3 1.4

P a p e r IV underground

0.17 mining 3.1 2.5

P a p e r V non-wooden

0.05 houses 1.2 1.2

P a p e r VI underground

0.02 mining 5.2 4.6

T est-b a se d confidence in te rv a ls were e s t i m a t e d in paper III and IV according to M iettin en (29). A m eth o d for c a lc u la tio n of e x a c t confidence limits (30) was c o n sisten tly used in papers I, II, V and VI and for small subgroups in paper III.

M ie ttin e n 's m eth o d produces a valid c o n fid en ce in terv al for odds ratios only r a th e r close to unity (31, 32). The e x a c t confidence lim its are th e r e f o r e used in most analyses.

The hom ogeneity of the odds ra tio across age s t r a t a was te s te d with an asy m p to tic

likelihood t e s t (27). If a h e t e r g e n e ity of th e odds ra tio was found, the analyses were

p e rfo rm e d with s e p a r a t e age groups. O th e rw is e , the age adjusted re la tiv e risks

w ere c o m p u te d by th e M antel and H aen zel m ethod (33).

The e s tim a te of the population etiologic fra c tio n (EFpop) in all studies was calculated according to th e form ula: EFpop= C Ffrx(R R -l)/R R , w here C F ^ is the case fractio n (proportion of exposed cases and RR re la tiv e risk) (34).

To study occupational exposure with a d ju s tm e n t for smoking, a logistic model was used in paper II, with th r e e d is c re te levels of occupational exposure and four levels of life-tim e to b acco consum ption used in th e analyses. In paper VI the logistic model was used to e lu c id a te the joint e f f e c t s of underground mining and smoking.

Maximum likelihood e s t i m a t e s w ere p e rfo rm e d for the logistic regression

co efficien ts (35). An index of synergy was e s tim a te d on the basis of principles

described by R oth m an (36, 37) and was used in papers III and VI to assess deviation

from additivity.

METHODOLOGICAL CONSIDERATIONS

Representativity of the cases

At the s t a r t of this study in 1979, the Swedish C an cer R egistry s u ffered from several years of delay in d a t a e n try . The lung c a n c e r cases for the years 1975-1977 were th e r e f o r e c o lle c te d manually among th e notifications from the counties studied. Due to in co m p lete sorting a t th e re g istry , 90 cases w ere lost in this collectio n . Including cases w ith o u t close re la tiv e s and cases collected from the re g is te r but alive in May 1978 (and t h e r e f o r e excluded from the study), th e to ta l d e fic it was 146 cases (19%) of the 750 c a se s re g is te re d in all. There were no obvious d iffe re n c e s in d is trib u tio n of age and microscopic types b e tw e e n th e t o ta l re g istry m a te r ia l and the study m a te r ia l (Table 2).

Table 2. D istribution of m icroscopic ty p es for the to ta l reg istry m a te ria l and the study m a te r ia l.

Microscopic type The 604 cases The 750 cases

lung c a n c e r Code used in included in to tally

Swedish C a n c e r the study re g is te re d

R e g is try Per cent Per c en t

Bronchiolar and alveolar ca

076 1 2

Adenoca 096 13 13

Squamous cell ca 146 46 49

U nspecified ca 196 40 36

incl small cell ca

Mean age 68 year 67 year

Microscopic types of cancer

Copies of all cytology and histopathology r e p o rts were c o lle c te d . The r e p o rts were

read by one oncologist (L.-G. L.). In c ases with questionable lung c a n c e r diagnosis

copies of the hospital re c o rd s w ere also c o lle c te d . From this m a te ria l, the cases

w ere classified in the following groups:

1. Small cell c a rcin o m a

2. A deno carcin o m a, alveolar cell

150

c a rc in o m a and bronchiolar carcinom a 3. Squamous cell c a rcin o m a

4. Poorly d if f e r e n t i a t e d carcin o m a

81 285

(not sp ecifically classified) and la rg e cell a n a p la stic carcinom a

5. Squamous cell ca rc in o m a + a d en o carcin o m a 6. M icroscopically verified but not

43 7

classified

7. Only c lin ic a l/ro e n tg en o lo g ic a l

13

diagnosis. Probably prim ary lung c a n c e r 8. M icroscopically verified malignancy but

5

probably not prim ary lung can cer 5

589

Groups 7 and 8 were very small in re latio n to the t o ta l m a te r ia l and were included in papers II-V. H owever, in papers III and V none of the cases with the specifically stu died occu p atio n al exposures (professional drivers and underground m iners, re s p ectiv ely ) belonged to groups 7 and 8. In paper I, th e se groups were excluded from th e analyses.

Completeness and validity of the data S m o k i n g h a b i t s

In order to e lu cid ate the reliability of smoking d ata c o lle c te d by questionnaires to close re la tiv e s , these d a ta , for a sam ple of th e cases, were com pared to d a t a in hospital reco rd s deriving from the p a tie n ts th em selv es. This sample (86 cases) consisted of all cases a d m i t t e d to the d e p a r tm e n t for re sp ira to ry diseases a t the University Hospital in U m eå. It was the routine in this d e p a r tm e n t to r e g is te r fairly d e ta ile d smoking histories. The re su lt of this com parison is illu stra te d in Fig.

1. Among th e 86 p a tie n ts , 3 had no smoking d a ta re g is te re d in the records. For the rem aining 83 cases, th e tw o sources showed good a g r e e m e n t (99%) as regards dis­

crim ination betw een sm okers and non-sm okers. In 53 re c o rd s, in form ation about

th e t o ta l num ber of smoking years was available. The a v e ra g e number of smoking

years given in the questionnaires was 47.9 and in the re c o rd s 47.7. The individual

d iffe re n c e was < 5 years in 39 of the 53 sm okers.

It was possible to distinguish betw een c i g a r e t t e , pipe or cigar sm okers in 61 reco rd s, and th e a g r e e m e n t b e tw een th e tw o sources was good in 48 of th e s e . In 37 of these reco rd s, the daily to b a c c o consumption was re g is te r e d . This could be used for s t r a t i f i c a t i o n into low, m o d e r a te and heavy tobacco consum ers. A ccording to this s t r a t i f i c a t i o n , the in fo rm atio n in 31 of these records agreed with the

questionnaires. In 3 of th e 6 records, with d isag reem en t b e tw e e n the tw o sources, the re la tiv e s had re p o rte d a higher to b acco consumption than the su b jects

th em selv es.

AGREEMENT

□

EVALUABLE SUBJECTS.

NUMBER

83 SMOKING / NON-SMOKING

61

TYPE OF SMOKING ( C IG A R E T T E S , P I P E ,

CIGARS)

NUMBER OF SMOKING YEARS.

ACCEPTED DIFFERENCE

£5 Y.

' IN T E N S IT Y OF SMOKING.

_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ LOW, MODERATE OR HIGH

F ig u re 1. Com parison b e tw een smoking d a ta from questionnaires answered by close r e la tiv e s and from hospital reco rd s.

O c c u p a t i o n

The years betw een age 20 and the tim e of diagnosis or r e t i r e m e n t were well- surveyed concerning the occu p atio n al history both for cases and controls. Including education periods, tim e of s ick -leav e, pension and unem ploym ent, this period could be m apped out in an a v e ra g e of 44.3 years for the cases, 44.9 years for th e

d eceased controls and 41.7 y ears for the living controls. If only the real occupation tim e was included th e corresponding figures were 38.9 y, 38.4 y and 38.1 y

re sp e c tiv e ly .

The answers deriving from two specific occupation groups, the non-iron m iners and th e copper sm e lte r w orkers, w ere c ontrolled against em p lo y m en t re g is te r s in order to validate the inform ation obtained from the questionnaires. The f irs t of these groups consisted of 10 cases and 11 controls which had been employed by one company (Boliden). Only 2 of the 21 questionnaires were misleading concerning underground work (+ 8 years and - 6 years). In none of th e rem aining 19

questionnaires did the d a ta about underground work d e v iate more than one year from th e re g iste r.

The second occupational group consisted of em ployees a t a copper s m e l t e r (Rönnskärsverken). The lung c a n c e r risk among th ese em ployees had previously been analysed in a co h o rt study including all male workers, employed a t l e a s t 3 months any tim e betw een 1928 and 1966 (38). Of th e t o t a l number of 76 lung c a n c e r cases found in the cohort, 26 w ere included in the p re se n t study. In 25 of these cases, the em ploym ent a t the s m e lte r was obvious in the questionnaires.

Three of th ese 25 cases were employed less than 1 year and were t h e r e f o r e not classified as copper s m e lte r workers in the p re se n t study. In 21 cases d a t a in the questionnaires concerning length of em p lo y m en t could be com pared with

corresponding d a ta in th e cohort, which derived from th e em p lo y m en t r e g i s t e r at the f a c to r y . In only one of th e 21 cases did the d eclared length of e m p lo y m e n t tim e d ev iate m ore than 2 years from the re g istry d a ta .

M u n i c i p a l i t y

As reg ard s municipality, the tim e period betw een age 20 and tim e of diagnosis for th e cases (corresponding tim e for the controls) was well covered in th e answ ers.

Only a m ean of 1 year per case and 1/3 y ear per control was missing. To study the

validity of these d a ta , a sam ple of cases and controls w ere se le c te d . This sam ple

consisted of all cases who lived in one sp ecific municipality (Sundsvall) a t the tim e

of diagnosis and th e ir co n tro ls. Inform ation about all m unicipalities w here these

cases and controls had lived during th e ir lif e - tim e was obtained from th e parish

au th o ritie s. This in fo rm atio n was co m pared with d a ta in the questionnaires from

the age of 20 years and fo rw ards. By a c c e p tin g a d iffe re n c e of 4 years b etw een the

two sources, a g re e m e n t was obtained in 79% of the cases, 86% of th e d e c e a se d

controls and 85% of the living controls. (Table 3). In only 2% of the cases, 5% of

th e dec e a se d controls and 3% of the living controls th e questionnaires gave d a ta

th a t dev iated with more than 20 years from the tr u th . The d a ta betw een the birth

and the age of 20 years was, however, m o re o fte n o m itte d or inad eq u ate in the

questionnaires. Inform ation about m unicipalities covering the whole period from

birth to diagnosis (corresponding tim e for controls) was thus obtained in only 65%

of th e cases, 69% of the d e c e a s e d co n tro ls and 74% of th e living controls. A ten d en cy to more c o m p le te and s u ffic ie n t d a ta was obtainded by self-resp o n d en ts (living controls), especially reg ard s th e ea rly life -tim e . However, the d iffe re n c e in d a ta quality betw een s u r r o g a te - and se lf-re sp o n d e n ts was less pronounced than e x p e c te d .

Table 3. D a ta about m u n icip alities o b tain ed by the questionnaires com pared to d ata from the parish a u th o ritie s . C oncerns only d a t a a f t e r age 20.

D iffe re n c e in years C ases

No.

D eceased co n tro ls

No.

Living controls

No.

< 4 72 78 53

5-10 10 4 4

11-20 7 4 3

>20 2 5 2

Total 91 9! 62

B u i l d i n g m a t e r i a l

A special study of the validity of the in fo rm a tio n in the questionnaires concerning

building m a te r ia ls was p e r f o r m e d a t th e N ational In s titu te of Environmental

Medicine (39). For a sam ple of 30 cases and 30 deceased controls, the d a ta given by

the s u rro g a te responders w e re c o m p ared to inform ation obtained from the local

public health boards. The analysis c o n c e rn e d 54 of these 60 persons, for whom data

w ere obtained from both sources and only dwelling periods longer than 2 years were

included. As regards 2 broad c a te g o r ie s of building m a te r ia l, wood/not wood, the

in fo rm atio n was identical for 117 of 121 dwellings in the tw o sources (Table 4).

Table 4. Comparison of d a ta on building m a te ria l of dwellings for deceased subjects obtained from a n ex t-o f-k in and from local public h e a lth boards.

N ext-of-kin

Wood Not wood T otal

Public Wood 90 3 93

health Not wood 1 27 28

board

Total 91 30 121

Concerning specific type of non-wood building m a te ria l (such as lightw eight concrete) the questionnaires did not give useful in fo rm atio n . No question about the b asem ent con stru ctio n was included, as le akage of radon fro m the ground not was recognized as a source of indoor radon when the p resen t stu d y s t a r t e d .

Comparison between cases and controls

The m a te ria l included th e following groups of cases and controls:

(a) all cases

(b) cases born a f t e r 1900 (c) all d eceased controls

(d) deceased controls born a f t e r 1900 (e) all living controls (born a f t e r 1900)

In papers I and II cases (a) were com pared to controls (c) (study model

) and cases (b) com pared to controls (e) (study model

).

In papers III, IV and V cases (a) w ere c om pared to controls (c) (study model i) and cases (b) com pared to con tro ls (d) + (e) (study model II).

In the d i f f e r e n t papers and in th e thesis, e s t im a te s based on study model B or II are presented in p a re n th e se s a f t e r the e s t i m a t e s based on study model A or I.

The philosophy behind th e d iffe re n t com parisons is discussed under C o m m en ts and

conclusions.

RESULTS

Smoking and lung cancer (Paper I)

In this study th e m a te r ia l was used for analyses concerning the lung c a n c e r risk of smoking with special em phasis on dose-response relatio n s, type of smoking, type of c a n c e r and e f f e c t s of smoking cessation.

The in fo rm a tio n from th e questionnaires concerning smoking habits included a p p ro x im a te year of s t a r t of smoking, daily number of c i g a r e t t e s , o th er types of smoking and y ear of possible cessation of smoking:

Smokers □ Yes □ No

If th e answ er is yes:

Age a t which smoking began ...

Smoking break m ore than 1 year ...

C i g a r e t t e s

□ occasionally

□ 1-7 cig/day

□ 8-15 cig/day

□ 16-25 cig/day

□ m ore than 25 cig/day Pipe

□ occasionally

□ less than 1 p a c k e t (100 gr)/w eek

□ m o re th a n 1 p a c k e t (100 gr)/w eek C igarillos

□ occasionally

□ 1-4 cigarillos/day

□ 5-10 cigarillos/day

□ m o re th a n 10 cigarillos/day Cigars

□ occasionally

□ 1-2 c ig a rs/d a y

□ m ore than 2 c ig a rs /d a y

Stopped smoking □ Yes □ No

If yes, a t which a g e ? ...

Individuals who had sm oked a t le a st one c i g a r e t t e daily or equivalent am ount of

to b a c c o for one year or m ore a t any tim e were classified as smokers. With this

definition 93% of the cases, 64% of the deceased controls and 62% of the living

con tro ls w ere sm okers. The odds ratio s obtained for a ü smokers were 7.3 in study

model A and 9.0 in study model B. The population etiologic fraction a t tr ib u ta b le to

smoking was 80% (83%).

The m a te ria l in this stu dy included d if f e r e n t types of sm o k e rs. Pipe smoking was quite as com m on as c i g a r e t t e smoking, and also a co m b in a tio n of pipe and c i g a r e t t e s was fre q u e n t, while cigar and cigarillo sm okers w ere r a r e . The r e la tiv e risks obtained for pure c i g a r e t t e sm okers was 7.0 (9.2), for pure pipe sm okers 6.9 (8.1) and for com bination sm okers 8.9 (11.8). The pipe smoking cases, how ev er, had significantly higher m ean age and m ean smoking y ears a t t h e tim e of diagnosis than the c i g a r e t t e smoking cases. With m easu res such as smoking tim e , smoking

intensity and lif e - tim e to b a c c o consumption obvious d ose-response re la tio n s w ere observed both in c i g a r e t t e and pipe sm okers.

The re la tiv e risk gradually declined with increasing y ears of smoking c e ssa tio n . However, this reduction was dependent upon the previous sm oking tim e and the de c re a se was less pronounced for smokers with long sm oking h istories.

F u r th e r m o r e , the d e c re a s e of the re la tiv e risk in e x -sm o k e rs was d ependent on the type of smoking and was m ore pronounced in e x - c i g a r e t t e sm okers th an for ex-pipe smokers.

The re la tiv e risk for sm oking was re la te d to the m icro sco p ic subtypes. A strong association b e tw een lung c a n c e r and smoking was observed for sm all cell

carcin o m a (

R=13.8 (44.6)), squamous cell ca rc in o m a (

=11.8 (9.8)), and a group of poorly d if f e r e n t i a t e d c a rcin o m a (large cell ca or poorly diff. ca not f u r t h e r classified) (RR=7.3 (7.4)). The association b e tw een ad e n o c a rcin o m a and smoking was much w eaker (RR=2.4 (3.1)) but still s ta tis tic a lly s ig n ific a n t. Some d iffe re n c es betw een d if f e r e n t ty p es of c a n c e r w ere observed con cern in g th e distribution of c i g a r e t t e and pipe sm okers. C ig a r e tte sm okers w ere m ore common among th e smokers with small cell c a rc in o m a than am ong th e sm o k ers with squamous cell carcin o m a (41% versus 34%), while the r e v e r s e was t r u e for pure pipe sm okers (30% versus 41%).

Occupation and lung cancer (Paper II)

In this p ap er, odds ra tio s for d iffe re n t o ccupations are p re s e n te d concerning lung ca n c e r, with and w ithout a d ju stm e n t for smoking.

In the questionnaire, one overwiew question co n cern ed gen eral d a ta on o ccupations,

e m ploym ents, branches of business, company n am es, s p e c ific tasks and tim e

periods. F o u r te e n additional questions c o n cern ed sp e c ific occupations or groups of

occupations such as m iners, professional drivers, s m e l t e r w orkers, m ech an ics,

m achinists, w elders, plum bers, e le c tric ia n s , c a rp e n d e rs , p a in te rs e t c . Most of these

questions contained se v e ra l subquestions aim ed a t a clo ser id e n tific a tio n of the

type of work. One question also concerned occupational exposure to asbestos (regardless of th e sp ecific occupation.)

A person who had been a c tiv e a t least one year in a sp e c ific occupation was assigned to this occupation in the analyses. One person could th e r e f o r e be counted for more than one profession.

The lung c a n c e r risk was calc u la ted for d if fe re n t occupations; some occupations, however, were pooled into groups and not s e p a ra te ly p r e s e n te d . The odds ra tio for a specific occupation was e s tim a te d re la tiv e to all rem ain in g persons and

p erfo rm ed both with and w ithout a d ju stm e n t for smoking. F a r m e r s , fo re s te r s , c a rp e n te rs and navvis had low odds ratio s (OR=0.4-0.8) while underground miners, copper s m e lte r workers and ele c tric ian s and plumbers had high odds ratios

(OR=1.6-2.8). The corresponding e s tim a te s a f t e r a d ju s tm e n t for smoking were very sim ilar O R=0.5-0.9 and OR=1.5-2.8, re sp e c tiv e ly . Professional drivers, pulp

workers, one group including te a c h e rs, clerks, salesm en e t c ("white collar workers") and one group including em ployees in m ech an ical industry, work shops, m e ta l industry, garages and m achineries ("mechanics") had all a tendency tow ards slightly increased odds ratio s. This increase for m echanics and professional drivers did not, however, p ersist a f t e r ad ju stm e n t for smoking. The odds r a tio s obtained for some occupations or groups of occupations are shown in Fig. 2.

With the use of a linear logistic regression model, smoking adjusted re la tiv e risks

for tw o d if fe re n t groups of "risk occupations" were e s t i m a t e d . One group included

w orkers with more than 5 years em ploym ent in occupations known or definitely

suspected from previous stu d ies to increase th e lung c a n c e r risk. The o th e r group

consisted of workers with occupations so m e w h a t su sp ected of increasing the lung

c a n c e r risk. A third group (all the rem aining persons) w ere used as a r e f e r e n c e in

relatio n to which the re la tiv e risks were e s tim a te d . A f te r a d ju stm en t for smoking

only th e first group showed an increased re la tiv e lung c a n c e r risk. With some

assum ptions, the population etiologic f ra c tio n a t tr ib u ta b le to occupation was

e s tim a te d to 9 per ce n t.

Odds ratio

(vs deceased controls)

Odds ratio (vs living controls) 3.0 2.0 1.0

d

Farmers Foresters Carpenters

Navvis

i IB

1.0

H —

2.0 3.0

White collar workers

"Mechanics"

Professional drivers Pulp workers

Plumbers and electricians Underground

miners Copper smelter

workers

Figure 2. Odds ra tio s for d if f e r e n t occupations (or groups of occupations) concerning lung c a n c e r. V ertical solid lines: w ithout a d ju s tm e n t for sm oking.

Vertical dasked lines: with ad ju stm e n t for smoking.

Professional driving, smoking and lung cancer (Paper III)

Through th e exhaust of diesel and gasoline engines, professional drivers a re exposed to benzo(a)pyrene and o th e r carcinogenic polycyclic a r o m a tic hydrocarbons. Some previous studies had suggested an in creased lung c a n c e r risk in this occupational group, e.g. a study based on the Swedish C a n c e r-E n v iro n m en t R eg ister (40). In the p resen t case -c o n tro l study, professional driving was a r a t h e r common and

geographically well d istrib u ted occupation which, in com bination with the

longitudinal smoking d a ta and o ccu p atio n al d a ta , made th e m a te ria l suited for

analyses of this specific group.

In paper III, mainly analyses p e rfo rm e d w ith th e 456 cases (all born in 1900 or la te r), which had one d e c e a s e d and one living co n tro l, a re p re se n te d . Analyses based on all 589 cases w ith one dec e a se d co n tro l were also p e rfo rm e d . They gave very sim ilar results and a re n o t p re s e n te d in th is overview .

The questionnaire included a s p ecific question concerning th e type of professional driving (tru ck , van, bus o r tax i) and ti m e periods for th e d if f e r e n t ty p es of driving.

All persons who had been a c tiv e in the profession for a t l e a s t one year w ere classified as professional drivers. O c c u p a tio n -y e a rs was t h e only m easu re of exposure used.

S ix ty -th re e cases and 95 co n tro ls w ere classified as professional drivers. Bus, taxi and van drivers were few and tru c k drivers c o n s titu te d th e only subgroup la rg e enough fo r s e p a ra te analyses.

The unadjusted odds r a tio for professional driving for m ore than 1 y e a r was 1.4 and fo r m ore th a n 20 years 1.6. A h e te r o g e n e ity of th e odds r a tio across t h e age s t r a t a (age a t diagnosis and corresponding tim e fo r controls) was found and s e p a r a t e analyses w ere th e re fo re p e rfo rm e d fo r two age classes; <70 y and >70 y.

F o r both professional drivers as a whole and t r u c k drivers, a significantly in creased re la tiv e lung ca n c e r risk was found in th e higher age group, 3.2 and 5.7

re s p e c tiv e ly , while th e r e l a t i v e risk in th e younger age group did not e x c e e d 1.0.

Both am ong cases and co n tro ls, th e sm oking h a b its w ere d i f f e r e n t in driv ers and non-drivers. The drivers included r e la tiv e ly m o re sm okers th a n th e no n -d riv ers and th ey w ere m ore often high to b a c c o co n su m ers. Without consideration of the age d istribution, th e smoking drivers showed an odds ra tio of th e sam e size as smoking non-drivers. The s t r a t i f i c a t i o n according to age, h ow ever, gave a much higher r e la tiv e risk for smoking drivers in th e group aged 70 or over (RR=23.0) th a n for smoking non-drivers in th e sam e age group (RR=6.7). F u r t h e r s t r a t i f i c a t i o n according to daily to b a c c o consum ption showed a very high re la tiv e risk fo r professional drivers with heavy consum ption (>20 cig/day) aged 70 or over

co m p ared to a corresponding group of non-drivers. A slightly (but not significantly) increased odds ra tio was obtained for non-smoking d rivers. However no e s t i m a t e could, be obtained for this c a te g o r y in th e older age group due to lack of non­

smoking cases.

Radon, smoking and lung cancer (Papers IV, V, VI)

Underground iron ore m iners in sev eral Swedish s tu d ie s have been shown t o run an increased risk for lung c a n c e r , in all probality due to in creased radon dau g h ter c o n c e n tra tio n in poorly v e n tila te d mines (41, 42, 43). Within t h e region fo r this study la rg e iron ore mines e x iste d in the m unicipalities Kiruna and G ällivare. The c a s e -c o n tro l m a te r ia l was t h e r e f o r e utilized fo r a special s tu d y concerning th e lung c a n c e r risk in th e se m iners and analyses of in te r a c tio n b e tw e e n smoking and underground mining.

Inform ation on the sp e c ific miners, e m p lo y m en ts and e m p lo y m en t tim e s was obtained by an overview question. One sp e c ific question c oncerned ty p e of mining (surface mining, underground mining and o th e r tasks) w ith corresponding tim e periods.

In this p a p e r, only underground mining in iron ore mines was regarded as exposure.

H owever, t h e resu lts w e re in principle sim ilar if o th e r ty p e s of underground mining (sulphide mines) were also included as exposure (44).

A num ber of 23 (22) cases and 10 (16) w ere classified as underground miners (> 1 year) and the a v e ra g e tim e of em p lo y m en t for the c a s e s was much longer than for th e c o n tro ls. The e s t i m a t e d re la tiv e risk of lun§ c a n c e r fo r underground miners was 2.5 (2.8). In th e lung c a n c e r cases exposed to underground mining th e av erag e tim e from t h e s t a r t of this work to diagnosis ("latency tim e " ) was 34.8 y ear. An o v e rre p re s e n ta tio n of sm all c e ll carc in o m a was also found among the cases exposed to underground mining c o m p ared to th e t o t a l m a te r ia l. The la te n c y tim e was considerably s h o r te r for sm all cell ca rc in o m a (28.7 years) than for epiderm oid c a rcin o m a (40.8 years). Among cases exposed to underground mining those with small c e ll c a rc in o m a w ere an average younger (58.4 y ears) than those with epiderm oid cell c a rc in o m a (67.0 years). This observ atio n was in sharp c o n t r a s t to the m a te r ia l as a whole, in which no age d iffe re n c e was found betw een sm all cell ca rc in o m a and epiderm oid ca rc in o m a cases. (P aper II).

The r e l a t i v e risk for sm oking underground m iners co m p a re d with non-smokers

w ithout underground mining was 18.2 (16.1). Very high r e l a t i v e risks w ere

e s tim a te d fo r underground m iners with high to b a c c o consum ption. The e s tim a te s

suggested a syn erg istic i n te r a c tio n b e tw e e n smoking and th e occupational

exposure.

The above r e p o r t (Paper IV) was based on th e to ta l c a se -c o n tro l m a te ria l which included many cases and controls w ithout exposure to underground mining in iron ore m ines. A supplem entary study was designed, lim ited to the m unicipalities of Kiruna and G ällivare, w here th e iron ore mines are lo c a te d , and a larg er proportion of exposed cases and co n tro ls could t h e r e f o r e be e x p e c te d . This study included c ases r e p o r te d to th e C a n c e r reg istry during th e e xtended period 1972-1982 (Paper VI).

The m a te r ia l consisted of 69 cases, 67 d eceased controls and 60 living co n tro ls. The analyses in this re p o rt were perfo rm ed acco rd in g to tw o study models in a sim ilar way as in P a p e r IV. Of th e 69 cases included in this stu dy, 42 were classified as underground m iners. The c a se s in underground miners diagnosed 1978-1982 were mainly squamous cell carc in o m a s and the o v e r -re p re s e n ta tio n of sm all cell c a rc in o m a found in P a p e r IV could not be observed in this m a te ria l. H ow ever, in this m a te r ia l the median age and la te n c y tim e were also sig nificantly lower for sm all cell th an for squamous cell c a rc in o m a .

The ag e -a d ju ste d re la tiv e risk for underground mining was 4.5 (3.3) in this study.

An obvious dose-response relationship b e tw e e n em ploym ent tim e (underground work) and lung c a n c e r risk was found; > 20 years of em ploym ent gave a re la tiv e risk of 7.7 (5.1). E s tim a te s obtained a f t e r s t r a t i f i c a t i o n fo r smoking and

underground mining (Paper VI, Table 5) strongly suggested a m ultiplicative e f f e c t of smoking and underground mining. This ty p e of in te r a c tio n was supported by a linear logistic, regression analysis. The re s u lts did not change in principle if ex- sm okers of m ore than 10 years w ere re c la ssifie d as non-smokers.

The c a lc u la te d population etiologic fra c tio n was 48 (40) % for underground mining and 78 (79) % for smoking in this m a te r ia l.

The q uestionnaires contained one question about the type of building m a te r ia l in th e dwellings and tim e periods for living in th e se dwelling. From the answ ers, only wooden and non-woodes houses could be s e p a r a te d . The d a ta was used for a case- control study m e a n t as a pilot inv estig atio n (Paper V). Living in wooden houses strongly p redom inated in this population. No overrisk for lung c an cer due to living in non-w ooden houses could be found a f t e r ad ju stm e n t for smoking. The s tro n g predom inance of wooden houses suggested t h a t increased indoor radon

c o n c e n tra tio n s due to building m a te ria l played a negligible role in this region. The study could, of course, not exclude t h a t o th e r sources of increased radon

c o n c e n tra tio n s (leakage from a ground) could have played a role.

Can the findings explain the geographical incidence differences?

Large geographical v ariations in th e a g e -s ta n d a rd iz e d incidence r a t e s of m ale lung ca n c e r was found within the studied region (Fig 3) and was actu a lly one in cen tiv e for th e c a s e -c o n tro l stu d y . The principle aim of this study was, however, n o t to find explanations of th e se d iffe re n c es but r a t h e r to e lu c id a te e f f e c t s of smoking and occu p atio n al exposures. Against the background of t h e larg e incidence

v ariations t h e r e was reason to suspect g r e a t variations concerning th e s e exposures, which could be a good basis for a c a s e -c o n tro l study. The co n tro ls w ere m a tc h e d against th e cases according to m unicipality in order to re d u c e confounding from fa c to r s o th e r than smoking and occupation (for in stan ce, g en eral air pollution). The design of th e study was thus not aim ed a t m easuring the e f f e c t s of geographical domicile.

C o u n ty /m u n ic ip a lity Age s ta nd a rd ize d inc idence r a te

Year: 59 -68 6 9 -78 59 -78 1969-78

VÄSTERNORRLAND 26 .7 3 6 .2 3 1 .4 134513

Ange 17.1 3 3 .0 25 .1 7319

Timrå 22 .4 3 3 .7 28.1 9299

Härnösand 2 5 .4 ,59.0 4 2 .2 13227

Sundsval 1 37.1 4 3 .9 4 0 .5 46448

Kramfors 2 8 .8 3 4 .3 3 1 .6 14169

S o lle fte å 1 7 .8 2 8 .2 2 3 .0 13581

Ö rnsköldsvik 19.1 2 4 .3 21 .7 30469

VÄSTERBOTTEN 16 .7 3 1 .2 2 4 .0 119433

Nordmaling 11 .6 3 1 .0 2 1 .3 4006

Vindeln 18 .8 4 3 .3 31 .1 3705

R o b ertsfo rs 9 .9 2 3 .5 16 .7 3917

Norsjö 8 .7 2 3 .2 15 .9 5459

Storuman 9 .5 1 6 .4 12 .9 4479

S ors ele 3 .9 3 .7 3 .8 2192

Vännäs 6 .6 - 2 6 .5 16 .6 6027

V ilh e lm in a 3 .9 7 .2 5 .5 4627

Asele 2 .9 1 4 .6 8 .7 4856

Umeå 2 4 .2 4 0 .0 32 .1 36366

Lycksele 8 .9 1 2 .0 1 0 .5 7500

S k e lle f t e å 25.1 4 1 .4 3 3 .2 36293

NORRBOTTEN 2 3 .9 3 4 .3 29.1 133493

A rv id s ja u r 6 .0 1 9 .4 12 .7 4276

A rje plo g 7 .9 7 .5 7 .7 2272

Jokkmokk 11.4 2 6 .5 1 9 .0 4084

öve rka l i X 6 .6 3 .3 2968

K a lix 2 1 .8 2 6 .3 2 4 .0 9398

öve rtorne å 12 .2 1 8 .6 1 5 .4 3577

P a ja la 12.0 2 0 .3 1 6 .2 5376

G ä lliv a r e 2 6.1 5 4 .7 4 0 .4 13237

Älvsbyn 12 .5 4 1 .4 2 7 .0 4585

Luleå 3 2 .4 4 2 .7 3 7 .6

P ite å 2 8 .2 35.1 31 .7 17428

Boden 2 5 .6 3 8 .0 3 1 .8 13755

Haparanda 2 8 .3 4 6 .6 3 7 .5 .

Kiruna 4 5 .2 35.1 4 0 .2 16097

WHOLE REGION 2 2 .8 3 4 .0 2 8 .4 387439

Figure 3. Age sta n d a rd iz e d incidence r a t e s (per 100.000 and year) of m ale lung ca n c e r in th e m unicipalities 1969-1978 (43). The rig h t m ap illu stra te s th e

deviations from the a v e r a g e for the whole region during th e period 1969-1978. The Swedish population 1970 se rv e d as s ta n d a rd population.

25

N e v e rth e le ss, some d iffe re n c es observed b etw een the m u n icip alities w ere of in t e r e s t in relation to th e observed incidence v ariatio n s. The controls fro m the d if f e r e n t m unicipalities could thus serv e as sam ples of the male populations in th e se domiciles with an age distribution defined by th e sam ples.

In o rd e r to com pare smoking habits and occupational p rofiles in d i f f e r e n t p a rts of th e region the m unicipalities were grouped with guidance of the a g e -s ta n d a rd iz e d in cidence r a te s during the period 1969-1978:

P e r c e n t of a v e ra g e in whole region

Area 1

(Mlow -in cid en ceM)

A rea 2

("m edian-incidence")

A rea 3

("high-incidence")

<90 per c e n t of the a v e r a g e in the whole region

90 - 130 per c e n t

> 130 per c e n t

The persons were r e f e r r e d to the m unicipality where th e y were r e g is te r e d a t the tim e of th e rep o rt to th e C a n c e r R e g istry . Moving fro m one are a to a n o th e r had only m arginal im p o rta n c e in this m a te r ia l and was th e r e f o r e n e g le c te d .

Table 5 shows smoking habits among the controls from th e se 3 a re a s . The highest r a t e of non-smokers and the lowest a g e ra g e l if e - tim e to b a c c o consum ption was observed in area 1 for both deceased and living c o n tro ls.

Table 5. Smoking habits among the co n tro ls from the geographically defined areas.

Geographical Non-smokers A verage lif e -tim e to b a c c o

a r e a per c e n t consum ption. lCp cig

D eceased Living D e c e a se d Living con tro ls controls co n tro ls controls

1 41 36 111 109

2 38 , 37 147 122

3 31 37 177 132