Physical work load, fetal development and course of pregnancy

K. Physical work load, fetal development and course of pregnancy. Scand J Work Environ Health 1989;15:404-414. The effects of physical work load during pregnancy were anal yzed in connection with a nationwide case-referent study that screened for associa tions between selected structural malformations and occupational exposures. The strain of the occupa tional activities of 1475mothers of malformed infants and an equal number of mothers of noncase babies was assessed from a description of the work tasks by an expert using a standardized method reflecting energy expenditure. The noncase mothers' experience revealed a relation between physical load and growth retardation that has also been suggested by other epidemiologic studies. No relation was found between an increase in mean physical load and the occurrence of threatened abortion; yet work involving much standing had an increased risk. Mothers whose work included occasional high physical loads had more pregnancy-induced hypertension. The data showed unexpected associations between physical load and structural malformations.

Rather little is kno wn about the effects of a mother's physical work load on the fetus and the course of pregnancy. Exercise during pregn anc y has been described to result in a reduction of birthweight in some anim al studies, but not in all , and some epidemiologic studies ha ve shown an association between low birthweight a nd ph ysically demanding work (I). A recently published study repo rted that the occupational lifting of hea vy weight s retarded fetal gro wth (2). The evidence from studies on spo ntaneous abortion and ergonomic factors suggests that heavy lifting and phy sical effort ma y be associated with an increased risk for this occur rence (3,4). Th ere are result s (5) suggesting that voluntary, moderate physical exercise during pregnancy may actually protect against preterm delivery, whereas heavy physical effort dur ing work and strenuou s work conditions may carry a risk of preterm birth (4).
The occurrence of human birth defects in relation to physical work demands has not been explored to any great extent, except in studies which have categorized women according to occupational titles. An investigation in Montreal (6) analyzed subgroups of structural birth defect s in relation to work , but the report did not includ e result s related to physical work load and structural congenital malformations. Instead , these data showed an associat ion between positional musculoskeletal defe cts and a long work week but no relation to ergonomic factor s.
As a part of a case-referent study screening for associations between selected structural birth defects and 404 occupational expo sure s on a nat ionwide basis (7,8), the present report describ es the results concerning physical work load during pregnancy. To study threatened abortion, pregnancy-induced hypertension, length of gestat io n, and birthweight, we analyzed the information obt ained on the non case mothers. Special empha sis was placed on expo sure evaluation , and physical wor k load was assessed from a description of the mother' s wo rk ta sks by an expert who used a standardized method refle cting energ y expenditure.

Birth defects
The source data of the study, obtained from the Finnish Register of Congenital Malformat ions, included 365 defe cts of the central nervous system , 581 orofacial cleft s, 360 structural defects of the skeleton , 169 card iova scular malformations, and 1475 noncase babies whose deliveries had preceded th e case deliveries in the same maternity health care distr ict (8,9). The noti fication s and autopsy reports of the case babies were checked by a trained pathologist o f the Register , and infant s with kno wn chromosome anomalies were not included. The mothers of the case and non case babies were inter viewed by mean s of standa rd Register pro cedures about particulars of the latest and previous pregnancies, diseases, intake of dru gs and alcohol , smoking, etc. About 99 % of the interviews were successful (10).
Inform ation on the mothers' work conditions was acqu ired through special inter views performed by two trained interviewers from the Finni sh Institute of Occupatio nal Health (8, 9, I I). The interviews took place about two to four months after the delivery. During the interviewer's visit to the matern ity health care center both moth ers o f the case-noncase pair were usually interviewed . Interviews were co mpleted for 96 fIJo of the pair s of mothers with eligible infants (9).
In all , 2073 mothers had worked throughout most of their pregnancy, 274 moth ers had worked only temporarily, and 603 mothers had not worked in th e first trimester. When birth defects were considered, the analyses were restricted to mothers who had work ed in the first trimester of pregnancy.
The physical load of the mot her 's occup at ional activities was assessed fro m a description of her ordinary workday, which had been record ed in detail in the special interview. Th e evaluation was based on a sta nda rdized meth od reflecting energy expenditure (12). Table I shows the a pplied stan dardized activity categor ies (13). An expert in physical education, trai ned at the Department of Phy siology o f the Finni sh Institute of Occupation al Health, assessed the moth er's major activity level at work. If the work included clear cha nges in ph ysical load, the minor activity level was also assessed. Occasional load s tha t exceeded the major and minor levels were evaluated separ ately. (See tab le 2.) Th e assessment was done by th e trimesters of pregnancy without the evaluator knowing the events related to the cour se of gestation or the pregnancy outcome.
Gro ups representi ng sho rt-term ph ysical load were defined accordin g to the physically most strenuous level that had occurred in the work as a main or minor level, or as an occasional load . In the first trimester the work of 1078 mothers had included only light activities (categor ies 3-5 in tables I and 2), 707 mothers had wor ked in jobs with a mod erate ph ysical load (category 6 in table 2), and 516 mothers had had some tasks that entailed a high ph ysical load (categories 7 and 8 in tabl e 2). The mother's mean level of physical load was calculated as an average o f the major and minor levels weighted by the proportion of work time spent at the level concern ed. For example, an office clerk who had been sitting and writing 80 fIJo of her worktime and had been standing and tak ing photocopies 20 % of the time was given a score of 3.2 (0.8 x 3 +0.2 x 4). Occasional load s were not included when the mean level of the physical load was calc ulated.
The mothers were categor ized accor ding to the weighted activity scor e into gro ups representing (i) Lif ting and c arrying bu rdens weighing bet w een 5 and 10 kg , makin g beds , sweeping and vac uuming Lift ing and c arryi ng bu rd ens wei ghi ng bet w een 10 and 25 kg, movi ng a pat ien t toget her wi th anot her person Lifti ng and car ryi ng burd ens weig hin g "" 25 kg , mov ing a patient alo ne • Category 1, lying asleep, and cat ego ry 2, lying awake , we re not appli cabl e. b The energy ex pendi t ure is around 1.5 times the basal metabolic rate, et c . a The to ta l du ration o f the lo ads was less th an 10 % o f th e total wo rkti me. b There were 44 moth ers fo r w hom th e phys ical wo rk load co uld not be asses sed. c A t otal of 973 mot her s wo rked at the same acti vity lev el all the time, and for two moth ers th e di st ribut ion of wo rkt ime between th e di ff erent act ivity levels co uld no t be defined. sede ntary work (3:5 weighted score < 4), (ii) sta nding work (4:5 weighted sco re < 5), (iii) work invo lving walking (5:5 weighte d score < 6), (iv) work with a mo derate mean ph ysical load (6 :5 weighted score < 7), and (v) work with a high mean physical load (weighted score~7). Th ere were only two mothers whose mean physical load in the first trimester had been high . These two women were included in the moderate gro up . In all, 1241 moth ers belonged to the first category, 574 to the secon d, 187 to the third , and 299 to the fo urth. Office cler k, keyboard operator, sewer, and production-line worke r are examp les of occupations ca tegorized as sedentary work . Job titles such as shop assista nt, coo k, a nd baker typically represen ted sta nding wo rk. Nurse's aid s, dom estic aids, and waitresses usually did much walking. Agricultural wor kers, cleaners, an d packers, but also so me waitresses and produ ction-line workers, had a mo derate mean physical load . O f the two mothers with a high mean physical work load one was a plywood-fiber board worker and the other a fore stry worker.
In the analyses the levels of th e gro up mea ns and short-term physical loads were used as nominal activity categories . Tab le 3 show s the mo thers according to these ca tego ries.
Mos t o f the mot hers with a moderate mea n ph ysical load were manu al workers when classed on a socioeco nomic basis (14) (ta ble 4). Mo reover, the groups represent ing distinct mean physical loads differed with respect to ot her work cha rac teristics. For example, almo st all of t he mothers in agricu ltural work belonge d to the gro up with a mod erate mea n load (ta ble 5). In add ition to studying all th e mot hers who wo rked in the first tri meste r, we restr icted some analyses to no nagricultu ra l wo rk ers in th e two socioeconomic classes of lower-level emp loyees with ad ministrative and clerical occupations and manual workers while adju sting for shift work, expos ure to noise an d solvents, a long work week, temporariness of employmen t; and mother's belonging to either of the two socioeconomic classes.
The group with a moderate mean physical load included more olde r mo thers, fewer mothers witho ut a previous pr egnancy, and fewer mothers who had consumed alco ho l dur ing pre gnan cy (table 6). When we studied the relation between phy sical work load and malfor mati on s, we adjusted for the aforementioned work characteristics and also for maternal age of~35 years, birth order higher than th ree, two or more induced abortions, prev ious miscar riage, prev io us ma lformed child , previous stillbirth, regu lar smok ing, alcohol consumption, intake of drug s in the first trimester , and co mmo n co ld or fever in the first tri mester.
The matc hing procedure had correlated the case and non case series with respect to agricultural work, and, as a con seque nce, with respect to physical work load . When all the moth ers who worked duri ng early pregnancy were studied , the data were analyzed as matc hed Table 3. Groups repr esent in g mea n and sh or t-te rm phy sical load in the first t rimest er of pregnancy. a Ac c ord in g to refe rence 14.
b Ow n-ac cou nt =self-employed.
C Women in occu pational training . a Farm , co m merc ial garde n, park, and fo restry work . b Level of noi se an s-n equivalent continuous A·w eight ed so und level o f aro und 80 dB or high er. C Continuou s concent rat ions of at lea st one-t hird of the th reshold limit valu es for chem ical substances in workroom air (15) or peaks highe r than t he thresho ld limit value. d Weekly worktim e > 40 h. Tabl e 6. Mat ern al characte ristics and mean phys ic al load in t he fi rst trim est er of preg nancy . series, and mutually consistent maximum likelihood estimates of odd s ratio s for several exposure levels were computed (16).
In the restricted study group , from which th e mothers in ag ricultural work had been excluded , th e case and noncase series were not correlated with respect to phy sical work load, and therefore , to gain efficiency (17,18), the restricted analyses were performed for unmatched series. The confidence inter vals for the crude odd s ratios were calculated with the modified Corn field method (19,20). The adjusted odd s ratio estim ates and the ir confidence intervals were calculated from the results of unconditional logistic regression analyses which were execut ed with the SAS software system (21). The ind ependent variables were entered in the mod els as bin ary code s or with the use of cat egory indi cators .

Birth weight and course of pregnan cy
The effects on birth weight and co urse of pre gnancy were determined from the noncase mothers' experience. Thi s approach ha s been describ ed in mo re detail elsewhere (22,23). Th e Registe r questionnaires and the records of the maternity health care center were used to obtain information on the noncase mothers' pregnancies. The mother was asked in the Register inter view if she had had a threatened abortion during her pregnancy; for 88 "70 of the mothers who had experienced vaginal bleeding the symptoms had occurred in the first or second trimester. An increase of at least 20 mm Hg (3 kPa) in the mean arterial blood pres sur e between the moth er' s first visit to th e maternity health care center and her last visit before delivery was considered a sign of pregnancy-induced hypertension (24). The length of gestation was calcu lated from the first da y of the last normal menstrual period. We anal yzed th e occur rence of pregnancies that were shorter than 280 d becau se the noncase series was too sma ll to permit a study of preterm birth. Th e referen ce value for the defin ition of a sma ll baby for his or her gestational age was the tenth perce ntile birthweight of the babies of the same sex in the same gestational age gro up born to mothers in work that included only light acti vities when sho rt-term ph ysical load was considered. When mean physical load was examined, the reference birthweight was obtained from the group in sedentary work .
Th e applied gestat iona l age gro ups were < 37 weeks, 37-39 weeks, 40-41 weeks, and~42 weeks. (Fo r mo re details on th ese out com es, see refe rences 22, 23, 25.) In all , 1042 of the non case mothers with a singleton birth had worked throughout most o f their pregnancy and had descript ions of their work that enabled the assessment of the ph ysical load by trim ester s of pregnancy. Of these moth ers, 952 (91 070) had had a ph ysical load that was ap prox imat ely similar during their entire period of wor k durin g pregnan cy. Respectively, 140 mothers had been employed on a tempora ry basis dur ing their pregnancy. A tot al of 270 women had not worked du ring pregnan cy, 15 of the working wom en did not have singleton bir ths, a nd for eight mothers th e physical wor k load co uld no t be assessed. A tot al of 1152 mothers (97 !t/ o) had wor ked in the first trime ster , 1141 (97 0J0) in the second, and 1035 (88 0J0) in th e third. When th e occurrence of threatened abortion-was studied , all the mother s who had had a singleto n birth and who had worked were included in the anal yses, but , when oth er aspects of pregnan cy were examined , onl y mo thers who had worke d thro ugho ut mo st of their pregnan cy were considered. (For mor e information on th e moth ers who had worked temporaril y, see reference 25.) The mothers' socioeconomic class and the features of the ir work were contro lled in the same manner as in the analyses on malformation s.
The work of 564 mo thers "in the first trimester had includ ed only physically light activiti es, 338 mothers had work that included activities with a modera te physicalload , and 250 mothers had been in jo bs in which some tasks had had a high ph ysical load . According to the weighted activi ty score 647 moth ers had don e sedentary work , 279 had don e sta nding work, 98 had done work which included walking, and 128 had done work with a mod erat e mea n physical load in the first trim este r.
For threatened ab ortion we adjusted for maternal age, parit y, outcome of previous pre gnan cies, history of menstrual irregularity, inta ke of dru gs, alcohol consumption, and smoking. For pregnancy-induced hypertension ma tern al age , par ity, outcome of previous pregnan cies, alcohol inta ke, and smo king were adju sted for. The mod els for length of gestat ion included the covari ates maternal age, parity, outcome of previous pregnan cies, histor y of menstrual irr egulari ty, alco hol intake , and smo king. When studying the occurrence of bab ies small for their gestational age, we controlled for maternal age , parity, outcome of previous pregn ancie s, maternal prepregnancy weight, alcohol inta ke, and smoking.
Th e estima tes and confidence limits for the crude risk rat ios were calcul ated according to the chi-sq uare function pro cedur e of Miettinen & Nurminen (20). To test for trend in th e proportions accord ing to the ordinal level o f ph ysical load , we used the chi-sq ua re 408 statistic o f Armitage (26). Th e estimates for the adju sted risk rati os were calculated fro m the results of binomial regression anal yses executed with the generalized linear interactive mod eling (GU M) program and the macros writte n by Wa cholde r (27). Independent variables were defined, and analyses were perform ed that corresponded to the pro cedures that were used to study malformation s.

Malfo rmations
T he crude odd s rati o of all birth defects pooled for work with a mean ph ysical load at the mod era te level in the first trimester was 1.5 with a 95 0J0 confidence interval (95 !t /o C I) of 1.1-2.0, for work which involved walkin g the odds ratio was 0.9 (95 0J0 CI 0.6-1.3), and for standing wor k it was 1.1 (95 0J0 CI 0.9-1.4), as compa red with sedentary work . The cru de overall od ds rati o for work which occas iona lly included activities with a high physical load was 1. 1 (95 0J0 CI 0.8-1.4), and for work that involved activities with a mod erate ph ysical load th e od ds ra tio was 1.2 (95 % CI 0.9-1.5), as compa red with work which included only physically light tas ks.
When the analyses were restricted to nonagricultural workers in the two socioeconomic categories of lowerlevel employees with administrative and clerical occupa tions and manual work ers, the mothers with a moderate mean physical load sho wed increa sed cru de and adj usted odds ra tios for defects of the centra l nervous system (table 7). Similarly, for high short-term ph ysical loa d , the cru de and adj usted odds ra tios for defects of th e central nervou s system were raised (adjusted odd s ratio 1.6 , 95 0J0 C I 1.1-2.4) when compared with wor k that had included only physically light activities. For mod erat e mean ph ysical load the crude an d adjusted od ds rat ios for oro facial cleft s were also slightly elevated (table 7), but this association was not seen for short-term physical load . The odds rati o point estimates of cardiovascular defects were somewhat raised for bot h the mean and shor t-term physical load . The occu rrence of skeleta l defects was not associated with ph ysical work load .

Vaginal bleeding (threatened abortion)
Th ere was no relat ion between a n increase in mea n physical load a nd th e occur rence o f vaginal bleedin g, but the crud e risk ratio was 1.6 (95 0J0 C I 1.1-2.3) for all the moth ers who, in the first trimester, had been in work that involved standing as compared with mothers in sedenta ry work. Th e restriction of the subject s to non agricultu ral workers in the two socioecono mic categori es of lower-level emp loyees with administrati ve and clerical occupations and manu al workers yielded similar results (tab le 8). The moth ers whose work had occasionall y included activities of high short-term ph ysical load in the first trime ster o f pregnan cy had had slightly mor e vaginal bleed ing than tho se whose work had involved only phy sically light tasks (tabl e 8).

Pregnancy-induced hyp ertension
An increase in the mean physical load of the third trimester was not associated with pregnancy-induced hypertension . Th e crude risk ratio was 1.2 (95 1110 CI 0.6-2.3) for wor k with a moderate mean load in the third trimester as compared with sedentary work . Similarly , the mean physical load in the second trime ster was not related to pregnancy-induced hypertension. When the data were restri cted to th e nonagricultural workers in the two socioeconomic classes of lower-level employees with admini strative and clerical occupations and manual workers, the adjusted risk ratio for moder-ate mean physical load in the third trimester was 1.1 (0.4-3.2) as compared with sedentar y work in the same tr imester. (See table 9.) Among all the mothers who worked in the third trimester, those who had done work that occasionally included activities with a high short-term ph ysical load had had more pregnancy-induced hypertension than the mothers whose occupational tas ks had all been physically light (crude risk ratio 1.6, 95 % CI 1.0-2.7). Similarly, work in task s with a high shortterm ph ysical load in the second trimester involved a slightly elevated risk of pregnancy-induced hypertension (risk ratio 1.5,95 % CI 0.9-2.4). When the data were restricted to the nonagricultural workers in the socioeconom ic categories of lower-level employees with administrative and clerical occupations and manual workers, the ad justed risk ratio for work th at occasionally included activities with a high ph ysical load Table 7. Birth def ects accordi ng to the mean physi cal load in the firs t trimeste r of pregnancy among the nonagricultural workers in the soc ioeconomic cl asses of lo wer -level employees with adm in istrative and clerical occupations and manual work ers.   3

Crud e risk ratio>
Adju sted risk rati o> Table 9. Pregnancy-induced hyp ertension and physical wo rk load in th e thi rd t rimes ter of pregnan cy amon g the non agricu ltural workers in the soc ioec onomic classes of lower-level employees wi th adm in istrative and cle rical occupat ions and manual wor kers. a Ther e were three mothers wi th miss ing data on blood pressure . b 95 % confidence interval in parentheses.

Length of gestation and birth weight
Alto gether, 59 % of the mothers with a moderate mean ph ysical load in the third trime ster had a gestation that was shorter than 280 d , whereas in sedenta ry work the correspo nding per cent age was 40 (cru de risk ratio 1.5, 95 % CI 1.2-1.8; adj usted ris k ratio 1.4, 95 % C I 1.1-1.7). In all, for 48 % of the mothers whose shor tterm ph ysical load had been high , the length of gestation was less than 280 d , and th is percentage was 39 for mothers whose work tasks had all been light (crude risk ratio 1.2, 95 % C I 1.0-1 .4; adjusted risk rat io 1.2,95 % CI 1.0-1.4). Table 10 shows the distribution s of length of gestati on for mean and short-ter m ph ysical load . When the study mothers were restricted to nonagricultural workers in the two soc ioeco nomic classes of lower-level emplo yees with administrative and cler ical occupations and manual wor kers , the adj usted risk rat io of shortened leng th of gesta tion was 410 1.1 (95 % C I 0.7-1.6) for mean ph ysical load at the mode rate level as compared with sedenta ry work. For high short-term physical load th e correspon ding adju sted risk ra tio was 1.1 (95 % C I 0.9-1.4) as contrasted to work which included only ph ysically light tasks. Among all the mothers who worked in th e third trimester , the tenth percentile birthweight was 2860 g for th e bab ies bo rn to mothers whose mean physical load had been at the mo derate level, whereas the resp ective birt hweig ht fo r the gro up in sede nta ry wo rk was 3000 g (ta ble II ). Th e crude risk rati o for mothers in wo rk with a mod er ate mean ph ysical load in the third trimester giving birth to babies tha t were sma ll for their gesta tional age was 1.3 (95 % C I 0.7-2.3) when the y were compared wit h the mothers in sede nta ry wor k, and the co rrespond ing adjusted risk ratio was 1.5 (95 % C I 0.8-2.8). When th e st udy gro up was restr icted to nonagricultural workers in the two socioeconom ic classes o f lowe r-level emp loyees wit h administr at ive an d clerica l occupati on s and ma nual workers, the mot her s with th e heaviest ph ysical load acco rd ing to the mea n and short-term level had an excess o f small a There was one mothe r with miss ing data on the birthweight of the baby. tested in occupati onal grou ps with th e use of measurements of heart rate and oxygen co nsum ptio n (13), bu t the gro u ps have not included pr egn ant women . In our study no physiological mea surements were taken, nor d id we have any data on the mother ' s physical work capacity . Thus, th e as sessment of physical load a pplies to lo ading fac tors at work a nd not to the mother' s actu al strain . Th e correspondence of work load with wo rk stra in in earlier stud ies ha s indicated that the Edholm method can overestimate the wo rk load at lower levels, but th e corresp ondence is good at moderate and high level s (13). Un expectedl y, mean ph ysical lo ad and the oc curren ce o f central nervous system defect s were related in the study da ta (table 7), and indications of a relation were seen also for sho rt-term ph ysical load. Simila rly , the so mewha t elevated odds ra tio s for orofac ia l clefts and ca rdiovascular defects were un expected.
(C o mpare the results pr esented in reference 6.) Our study involved the simulta ne o us assessment of associations be tween m an y occupat ional fa ctors and Birthweight(g) .
Tenth Median percentile Number of babies-babies as co m pared with th e mothers at the ligh test level o f physica l load (ta ble 12). The ad justed risk ratio wa s 2.4 (95 % C I 1. 3-4.6) for th e mothers with a moder at e mean physical load as co mpa red with mothers in sed entar y work.

Placental weight
In a ll, 5.2 070 o f the mothers whos e mean ph ysical load had been a t the moder a te level in th e th ird trim ester had placentas weighi ng 400 g or less as ag a inst 3.6 % of th e mothers in sedentary work (ris k ra tio 1.4 , 95 % CI 0. 6-3.6). Th e corresponding per centages for mother s with high a nd ligh t sho rt-term ph ysical loads were 4.4 and 3.7, respectively (risk ratio 1.2, 95 % C I 0. 5-2.6).

Discussion
Many of the mothers with th e same occupational title had had a very different physical load when assessed from the de scription o f the mother' s ordinary wo rkday. In o ur dat a , there were onl y seven mothers whose main ac tivity level had been high or very high. On th e o ther hand, so me 20 % o f the mothers had wo rked part o f th e tim e a t a high or very high level o f ph ysical lo ad. (See tabl e 2.) Thus, as assessed with th e use of an av erage of th e major and minor levels weighted by the proportion of wo rkti me sp ent a t th e level co ncerned , the gro up with th e ph ysicall y most load ing wo rk in o ur study had a mod erate mean level. Mo st of th e mothers whose sho rt -te rm physical load had been a t a high o r very high level had had a hea vy load only spo ra d ica lly. The two he aviest levels of th e sta ndardized categories of physical load have sugg estive metabolic reference values indicating tha t the energy expend it ur e in the bod y co uld be around eight to ten tim es greate r th an if the person wer e in a sta te of complete rest but awake . (See table I and refere nce 13.) Th e applicability of the basic acti vit y ca teg o ries a nd the co ncerned met a boli c reference va lues ha ve been Table 12. Babies who were small for their gesta tional age and phys ical work load in the third trimester of pregnancy among the nonagricultural workers in the socioeconomic classes of lower-level employees with administrativeoccupations and manual workers. four groups of structural birth defects from the same data. The study was not designed to test certain specific hypotheses, and thus caution is needed in the interpretation of the observed relations. Spec ific mechanisms of teratogenesis are poorly understood, and the effects of any single factor are obscured by other variables . Most of the mothers with a moderate mean physical load during work were manual workers (table 4), and, moreover, half of them had worked in agriculture as against only a small proportion in the physically lighter work load category (table 5). Therefore further analyses were restricted to nonagricultural workers in the two socioeconomic classes of lower-level employees with administrative and clerical occupations and manual workers. In addition , we controlled some aspects of work and also maternal background characteristics in the analyses. However it is possible that the observed relations were attributable to some unrecognized confounders, and then physical load would be a surrogate for something else. Besides, biases in maternal recall (during the interviews or in the assessment of exposure) are possible when information is acquired retrospectively. The mothers of the case-noncase pairs were usually interviewed at the same time after delivery, but especially the mothers of the babies with central nervous system defects gave birth earlier tha n the noncase mothers. Their length of gestation averaged about three weeks less; thus, at the interview , the time since the beginning of pregnancy was usually shorter for these case mothers than for the non case mothers. Whether this phenomenon might cause differences in recalling the course of an ordinary workday in the first trimester of pregnancy remains obscure. The trained interviewers, who recorded the mothers' free-form workday descriptions, were unaware of any specific hypotheses that might be tested, and they gathered information on several factors connected with work. The assessment of ph ysical work load was done by an expert on the basis of the WOI kday description, and she did not know the course or outcome of the pregnancy. Therefore at least the presence of the two last biases seems unlikely. However, the prior probability that physical load could induce birth defects was low, and thus we would like to emphasize the possibility that our findings are chance or proxy association s. Moreover, concerning central ner vous system defects, the possibility of differences in maternal recall could not clearly be ruled out.
In our study data, there was no relation between an increase in mean physical load in the first trimester and the occurrence of threatened abortion (vaginal bleeding); yet stand ing work was associated with a 1.5-fold risk (table 8). The mothers whose short-term physical load had been high also showed indications of an elevated risk . It should be noted that some 40 % of the mothers in standing work had occasionally had task s in the ir work that involved a high short-term physical load (table 3).

412
The mothers whose short-term ph ysical load in the second or third trimester had been high had experienced more pregnancy-induced hypertension than the mothers in work that included only light activities (table 9). Nevertheless, an increasing mean physical load was not associated with this outcome. There was variability in the time of the first blood pressure measurements, and women with pregnancy-induced hypertension who had registered late at the maternity health care center might have been lost. However, a connection between physical work load and the time of the first visit that could have biased the observed results do es not seem plausible becau se of the comprehensive maternity care system in Finland. (For more information on th is subject , see reference 25.) Among all the mothers who had worked throughout most of their pregnancy, physical load appeared to shorten the gestation, but the restricted analyses showed no such associations. Moreover, our study was too small to permit an evaluation of preterm delivery.
Indications of a relation between physical activity in the third trimester and reduced prenatal growth were seen for both mean and short-term physical load. The excess of small babies among the mothers with a moderate mean physical load or a high short-term load co uld be seen especially when the analyses were restricted to nonagricultural workers in the two socioeconomic classes of lower-level employees with administrative and clerical occupations and manual workers (table 12), but in the more heterogeneous data the excess was smaller. Furthermore, for mothers in standing work, a moderate risk of giving birth to small babies could not be ruled out in light of the obtained results .
The original objective of our study was to explore possible factors of teratogenecity in the work environment , and, later , the available noncase mothers' experience was used to study outcomes of pregnancy other than malformations. Related to the obtained results, the issue of possible chance or proxy findings is similarly a cause for concern as in evaluations of associations linked to birth defects. A reduction in birthweight has been considered one of the most likely possible effects of physical load on the fetus (I). Some epidemiologic studies on maternal work conditions and pregnancy have reported suggestive findings concerning phy sical load and reduced birthweight (2,(28)(29)(30)(31). A higher risk of spontaneous abortion and shortened gestation has also been reported for ph ysically demanding jobs (2,(30)(31)(32)(33)(34), but in other studies no relation was observed between preterm birth and physical strain (5,35).
In the aforementioned studies , the definitions of physical load have included physical effort, heavy lifting, and standing, and a persistent difficulty has been , and is, the categorization of women according to varying physical work demands (36). Consequently, the interpretation of the concerned exposure is not evident , and the comparison of results from different studies is cumbersome . To increase the complexity, physical work loads typically occur in environments which contain chemical and other types of occupational exposures that are also potentially harmful during pregnancy. It is also possible that different kinds of physical load (static versus dynamic, long-term versus shortterm), if indeed harmful to the fetus or pregnancy, might not have similar effects upon the different outcome variables included in different studies.
Still, the prior probability of a possible relation between physical load and reduced birthweight or an adverse course of pregnancy was higher than for malformations. Our results gave some indications of possible effects, and our findings also suggested that prolonged and short-term physical load could have different consequences. However, the effects of the increased metabolic demands and other loading aspects of physical activity on the course and outcome of pregnancy need to be studied further.