This study examined the relationship between drinking water source, amount consumed, and trihalomethane (THM) levels during pregnancy and risk of adverse pregnancy outcomes including miscarriage, preterm delivery, and low birth weight. Data were collected through interviews of women in central North Carolina who experienced these outcomes as well as term, normal birth weight controls. Exposure was assessed based on reported water source, amount consumed, and measured community THM levels during relevant periods of pregnancy. Adjusted odds ratios were calculated to estimate relative risks for various exposure categories while controlling for potential confounding factors. The results provide limited evidence of small increased risks for some outcomes associated with surface water use and higher THM levels. However, exposure assessment limitations and residual

1. muffol mm
Drinking Water and Pregnancy Outcome in slightly in relation to both surface (versus
ground) water use and elevated THM lev-
Central North Carolina: Source, Amount, and els, with adjusted OR, in the range of
Trihalomethane Levels 1.1-1.4. Except for a closer temporal rela-
tion between the pregnancy and the mea-
David A. Savitz, Kurtis W. Andrews, and Lisa M. Pastore surement, the same limitations noted for
Department of Epidemiology, School of Public Health, University of North Carolina, the Iowa study are applicable to the New
Chapel Hill, NC 27599 USA Jersey study.
Other reports of less direct relevance
.W.~: .C. the
Inspi.te ofe:. reco-ton of potentiay
....ii A
re a
".
improved estimation of contaminant con- found that stillbirth risk was associated
: .emc.a.s in cli .nated drinidng
-ch: centration. The seasonal variation in chlo- with chlorinated versus chloraminated
watrsouAce,
with'nna rination by-product levels (higher in the water supplies in Massachusetts (8) and
.:
:
* :h .:. ....h..... :: ..p summer) can be incorporated into repro- that the use of bottled water rather than
.:
~ ..... .. ....
...c..t".e..~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ .....eri. ... .. ..l... b ^ .
ductive health studies as a component of tap water may be associated with a reduced
... .... ... ... .: .. .. ..s .. .. .. .
the exposure variability that is analyzed. risk of spontaneous abortion in Northern
Laboratory research relevant to chlori- California (Se. However, the investigators
c. traion. Water s.o.ce w
. ..... .... ......l ..., ..... . ...
a.te.d to...n
...d
nation by-products and reproduction is suggested that a reporting bias may
limited (3,4) with most evaluations account for the latter association (JO).
*ranyo 'hsampregancingb5_twA
incras pog tof ted outces in
water s
i
... .. ........... .....
:f: focused on single chemicals rather than the
..
complex mixture encountered by humans
Given current knowledge, additional
work is clearly needed to evaluate whether
a.so..ated iti.decre.sed s of. .11 thse. ..
in treated water. At exposure levels orders the previous suggestions of small adverse
outcomes '( i....r......a....anUoe(cnern0
TH'Ocentao r d .5 o.
..l f
..< of magnitude higher than those encoun- effects on pregnancy are likely to be causal.
~tine~ ~ xrnout were no reaedopUgon
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~., ....:...............:.*:: X siEi!
...
d w _ tered naturally, developmental toxicity in Data collected from interviews of women
dafa.lin to OJ'fer 4+:. glsae per day)*:
..*.
a .. ~~~~~~~... ..... .. ....... ..... .. ....... .....
..
the form of reduced fetal weight, heart who experienced miscarriage, preterm
malformations, and reproductive toxicity delivery, and low birth weight births as well
related to adverse effects on sperm has as term, normal birth weight controls allow
y .outcme with thepossibl eeppro of been demonstrated for chloroform, bromo- for an examination of individual water con-
an .c... f m in the high-
..
e. ..a.e. of TH.M concetratin (ad.u form, haloacetic acids, and related com- sumption in relation to water source and
d .....
odds rto28,95% cniec itra
:.. 4499 THUi pounds (3). However, it is not clear
whether they produce toxic effects at the
measured community THM levels.
1.1-.). which was not pa.t.of n.. o.ll
a Methods
dosesos graien :These dat .do not. low exposure levels of concern.
a ~trongassocationbetween:. 0hoi
indicte Wa ...
.......... -no . . . . .. .......
Prior research on the outcomes of A population-based case-control study of
nation y-prouct andA adverse pregnnc
b interest, fetal loss, preterm delivery, and miscarriage, preterm delivery, and low
..outcome.b.hu~t: givena the.lmited qultyof low birth weight, is limited in both quanti- birth weight was conducted in Alamance,
our e.posure asesmn and the inclae ty and quality (4). In the most thorough Durham, and Orange Counties in central
msa.. rage risk in the highst eore effort to evaluate preterm delivery, low North Carolina. The region was originally
groupmor refined evuat is warrante.
* vd choination, ltw birth w h, birth weight, and small for gestational age chosen for the concentration of textile
prterm ..00 ".;n on,: tiM-
del.:.::iverytspont::aneousa
s (SGA), Kramer et al. (5) conducted a study industry employment in Alamance County
-P Mt.. in Iowa. Exposure was classified based on and expanded to include the larger, more
baomtaness Envirean Health Peripec:t
103:-96 (195) the community of residence in conjunction sociodemographically diverse populations
with a survey of chlorination by-products. of Durham and Orange counties.
Chloroform concentrations above 10 ppb All medically treated miscarriage cases
It has been known for nearly 20 years that in drinking water were associated with a among women in Alamance County dur-
chlorination of surface waters produces small increase in risk of low birth weight ing the period September 1988 through
small amounts of chloroform and other [adjusted odds ratio (OR) = 1.3] and a August 1991 were identified through med-
potentially toxic by-products (1). Ingestion somewhat greater risk of SGA (adjusted ical care providers, including hospitals and
of these agents by large numbers of people OR = 1.8). As noted by the investigators, private clinics, as described in detail else-
over extended periods of time has generat- there was no opportunity to consider fluc- where (11). Preterm deliveries (<37 weeks
ed considerable concern with potential tuations in the contaminant levels over completed gestation) and low birth weight
adverse health effects. Most of that concern time or individual variability in water con- infants (<2,500 g) were identified at six
has focused on carcinogenicity, with stud- sumption. area hospitals covering virtually all births to
ies providing mixed support for an associa- The only other study that considered area residents during the period September
tion between chlorination by-product con- trihalomethane (THM) levels in the com- 1988 to August 1989 in Orange and
centrations and the risk of bladder and munity supply was conducted in northern
colon cancer (2). New Jersey (6) using birth and fetal death
Reproductive outcomes, known to be certificates to identify birth weight, low Address correspondence to D.A. Savitz,
sensitive to environmental toxicants, have birth weight (<2500 g), very low birth Department of Epidemiology, CB 7400,
University of North Carolina, Chapel Hill, NC
received much less attention. In addition to weight (<1500 g), term low birth weight, 27599 USA.
the obvious public health impact of con- preterm delivery, SGA births, and fetal We thank Kate M. Brett, Laurie E. Evans, and
genital malformations and fetal and infant deaths. Mean birthweight was reduced Nancy Dole Runkle for work on data collection
death, studies of reproductive consequences slightly in relation to use of surface water and analysis and acknowledge funding from the
have the logistical advantage of a shorter supplies and in relation to use of water with National Institutes of Health (HD23862,
interval between exposure and disease man- THM concentrations above the federal HD23862S), March of Dimes (15-157), Centers
for Disease Control (R48/CCR402177), and U.S.
ifestation. This briefer period of interest standard of 100 ppb (7) as reflected in the EPA (CR 820076-02).
facilitates more accurate recall of consump- nearest quarterly sample. The risk of Received 5 January 1995; accepted 10 March
tion over the relevant time period and adverse outcomes was generally elevated 1995.
592 Environmental Health Perspectives

2. A J a * e.-. * e 9 -
Durham Counties, and September 1988 to your primary source of drinking water at for analyses of miscarriage and the live
April 1991 in Alamance County. There was home? Was it supplied by the community birth outcomes (as indicated in Tables
substantial overlap between the preterm water company, from a private well, or 2-4); 5) THM dose (glasses per day x con-
and low birth weight groups, with 34% of bottled water? This was followed by the centration): analyzed as continuous mea-
eligible live birth cases preterm and not low question, About how many glasses of water sure and divided into tertiles based on dis-
birth weight, 17% low birth weight and did you drink per day around the time of tribution of controls, categorized separately
not preterm, and 50% both preterm and your pregnancy? for analyses of miscarriage and the live
low birth weight. After analyzing water source and birth outcomes as indicated in Tables 2-4.
Controls were selected in a one-to-one amount, we restricted the sample to women The ORs were calculated comparing
ratio to live birth cases from the deliveries who were served by public supplies and who exposures of cases to that of controls, e.g.,
immediately following a preterm or low reported drinking one or more glasses of community versus private source or higher
birth weight case of the same race and hos- water daily (omitting approximately 30% of versus lower THM concentrations. The
pital as the case, but restricted to term, nor- eligible subjects; see Table 1). A woman's ORs provide an estimate of the relative risk,
mal weight births. The controls selected for address was used to assign her to one of the i.e., the magnitude of increased risk associ-
preterm and low birth weight cases in five public water supplies serving residences ated with the exposed versus the referent
Alamance County also served as controls in this region. Although we did not have category. The confidence intervals (CIs)
for the miscarriage cases. We considered information on changes in water consump- provide an indication of the statistical pre-
the controls as a hospital- and race-strati- tion during pregnancy, we were able to con- cision of those estimates.
fied sample from the population. sider the changes in THM concentrations Based on preliminary analyses to iden-
Therefore, we did not analyze the data as a over time. The dates of pregnancy were tify factors associated with adverse preg-
pair-matched sample but controlled as used to assign the reported quarterly average nancy outcomes, potential confounders
needed for race and hospital in the analysis. THM value from the appropriate supplier included maternal age, race, hospital (for
Ten to fifteen percent of cases and con- as her THM score. For miscarriage cases preterm delivery and low birth weight
trols were lost due to subject refusal (high- and their controls, the fourth week of preg- only) education, marital status, poverty
er for miscarriage cases than the other nancy was the time period used for making level, smoking, alcohol consumption,
groups), and an additional 11 to 16% were that assignment, and for preterm delivery employment, and nausea (for miscarriage
lost due to being untraceable (Table 1). An cases, low birth weight cases, and their con- only). Crude and adjusted ORs were com-
abbreviated form of the questionnaire trols, the 28th week of pregnancy was used pared, with adjustment for each of the
which did not include the questions per- to assign the nearest THM value. These above covariates one at a time. When the
taining to drinking water was used for sub- periods reflect the most likely intervals in adjusted OR differed from the crude by
jects who would have otherwise refused which any adverse effects would occur. 10% or more, the variable was considered
(short questionnaire). Final response pro- With this information, we were able to as a confounder and incorporated into the
portions ranged from 62 to 71%, lowest analyze several indices of water exposure: adjusted ORs presented. When multiple
for miscarriage cases and highest for 1) source: community supply, private well confounders were identified, a logistic
preterm delivery cases. (referent), bottled water; 2) amount (glass- regression model was developed to simulta-
Telephone interviews were used to es per day): 0, 1-3 (referent), 4+; 3) source neously adjust for those confounders.
ascertain information on a wide range of x amount: private well, 1-3 glasses per day Therefore, the adjusted ORs presented in
potential risk factors for adverse pregnancy (referent); private well, 4+ glasses per day; the tables can be interpreted as free from
outcome, including sociodemographic community supply, 1-3 glasses per day; confounding by all of the above variables,
attributes (age, race, education, marital sta- community supply, 4+ glasses per day; bot- but only the subset that influenced the
tus, income), pregnancy history, tobacco tled (regardless of amount); 4) THM con- results (if any) were considered directly.
and alcohol use, prenatal care, physical centration: analyzed as a continuous mea- Given the small numbers of subjects in
exertion, psychological stress, and employ- sure and divided into tertiles based on dis- some cells and the lack of a biological basis
ment. Each woman was asked, What was tribution of controls, categorized separately for postulating effect modification, we did
Table 1. Number of eligible and interviewed participants: Alamance, Durham, and Orange counties, North Carolina, 1988-1991
Miscarriage analysis Preterm and LBW analysis
Cases Controls Preterm cases LBW cases Controls
No. % No. % No. % No. % No. %
Eligible 418 100.0 341 100.0 586 100.0 464 100.0 782 100.0
Physician refusal 9 2.2 1 0.3 3 0.5 4 0.9 5 0.6
Patient refusal 63 15.1 39 11.4 52 8.9 53 11.4 85 10.9
Untraceable 48 11.5 34 10.0 70 11.9 74 15.9 84 10.7
Other 18 4.3 4 1.2 6 1.0 7 1.5 4 0.5
Short questionnaire 12 2.9 22 6.5 39 6.7 25 5.4 51 6.5
Missing water data
Source only 1 0.2 0 0.0 1 0.2 1 0.2 0 0.0
Amount only 0 0.0 0 0.0 1 0.2 1 0.2 3 0.4
Source and amount 6 1.4 4 1.2 2 0.3 3 0.6 7 0.9
Complete water data 261 62.4 237 69.5 412 70.3 296 63.8 543 69.4
Restriction for THM analysis
No water consumed 40 9.6 22 6.5 44 7.5 35 7.5 40 5.1
Bottled or well water 77 18.4 70 20.5 104 17.7 67 14.4 141 18.0
Missing date/THM data 18 4.3 23 6.7 20 3.4 16 3.4 29 3.7
Complete THM data 126 30.1 122 35.8 244 41.6 178 38.4 333 42.6
Abbreviations: LBW, low birth weight; THM, trihalomethane.
Volume 103, Number 6, June 1995 593

3. CEMAMMZ.l
not examine interactions between water est risk in the referent group but no trend related to the risk of adverse pregnancy
exposures and other variables. of increasing risk across the middle and outcome, with the possible exception of an
highest categories. Analysis using the con- increased risk of miscarriage among bottled
Results tinuous dose did not indicate a positive water versus private well users. We consid-
Relative to women served by community association. ered only medically treated miscarriages
supplies, women served by private wells and found some evidence of differential
were more likely to be white and some- Discussion under-ascertainment related to social class
what more likely to be married, but were Overall, drinking water source was not (11). The association between miscarriage
otherwise similar with respect to educa-
tion, tobacco and alcohol use, and repro- Table 2. Miscarriage in relation to drinking water characteristics: Alamance County, North Carolina,
ductive history. Bottled-water users were 1988-1991
less likely to use tobacco and were more Cases Controls Crude OR Adjusted ORa 95% Cl
highly educated. The amount of water Water source
consumed was somewhat lower for women Private well 78 68 1.0 1.0
who were parous, white, and less educated. Community 171 159 0.9 1.0 0.7-1.6
THM concentrations were somewhat Bottled 12 10 1.0 1.6 0.6-4.3
higher for women who reported using Water amount (glasses per day)
tobacco or marijuana during pregnancy. In 0 40 22 1.6 1.6 0.9-2.8
general, these potential confounders
1-3b 137 120 1.0 1.0
4+ 84 95 0.8 0.8 0.5-1.1
(which were controlled, as needed, in the Water source x amount
analysis) were not strongly associated with Private well/1-3b 36 29 1.0 1.0
water characteristics. Private well/4+ 29 31 0.8 1.2 0.6-2.4
Risk of miscarriage was slightly Community/1-3 95 85 0.9 0.8 0.4-1.4
increased among women who reported Community/4+ 49 60 0.7 0.6 0.3-1.2
Bottled/l+ 12 10 1.0 1.0 0.3-3.1
using bottled water compared to those THM concentration (ppb)
who used private wells (Table 2), but this 40.8-59.9" 37 35 1.0 1.0
observation was based on few exposed 60.0-81.0 43 44 0.9 1.0 0.5-2.0
cases. Also, bottled water users were not at 81.1-168.8 46 43 1.0 1.2 0.6-2.4
increased risk relative to women served by Per ppb change 126 122 1.5 1.7 1.1-2.7
community supplies (Table 2). Regardless THM dose (gpb x glasses/day)
40.8-139.9 50 47 1.0 1.0
of water source, women who reported not 140.0-275.0 45 40 1.1 1.0 0.6-1.9
drinking any water were at highest risk 275.1-1171.0 31 35 0.8 0.6 0.3-1.2
and those drinking the largest amounts at Per 250 unit change 126 122 1.0 1.0 0.7-1.2
slightly decreased risk. This pattern was Abbreviations: OR, odds ratio; THM, trihalomethane.
also apparent in the analysis of source by aAdjusted as needed for potential confounders listed in text; if no confounders identified, the crude OR is
amount and possibly reflected in the presented.category.
reduced risk in the highest tertile of THM Referent
dose (THM concentration x amount).
THM concentration was not associated Table 3. Preterm delivery in relation to drinking water characteristics: Alamance, Durham, and Orange
with miscarriage risk in the categorical counties, North Carolina, 1988-1991
analysis, yet the continuous measure pre- Cases Controls Crude OR Adjusted ORa 95% Cl
dicted a rather substantial association, Water source
with an odds ratio of 1.7 per 50 ppb Private well 95 114 1.0 1.0
increment. This was attributable to a Community 294 388 0.9 0.9 0.7-1.2
much higher risk associated in the highest Bottled 24 43 0.7 0.8 0.4-1.4
sextile of exposure (adjusted OR = 2.8, Water amount (glasses per day)
95% CI = 1.2-6.1) with an anomalously 0 44 40 1.4 1.4 0.9-2.2
low risk in the second to highest sextile 1-3b 212 261 1.0 1.0 -
4+ 157 244 0.8 0.8 0.6-1.0
(adjusted OR = 0.2, 95% CI = 0.0-0.5) Water source x amount
(not shown). Private well/1-3b 46 50 1.0 1.0
Preterm delivery showed virtually no Private well/4+ 34 48 0.8 0.7 0.4-1.3
association with water source, THM con- Community/1-3 153 189 0.9 0.9 0.6-1.4
centration, or THM dose; all adjusted ORs Community/4+ 111 173 0.7 0.8 0.5-1.3
Bottled/l+ 24 43 0.6 0.6 0.3-1.3
were between 0.8 and 1.2 (Table 3). The THM concentration (ppb)
number of glasses of water consumed per 40.8-63.3b 80 110 1.0 1.0 -
day showed the same pattern as for miscar- 63.4-82.7 102 118 1.2 1.2 0.8-1.8
riage, with decreasing risk with increasing 82.8-168.8 62 105 0.8 0.9 0.6-1.5
amount. The estimates were much more Per 50 ppb change 244 333 0.8 0.8 0.6-1.2
precise than for miscarriage, due to a case THM dose (gpb x glasses/day)
44.0-169.9 78 108 1.0 1.0
group nearly twice as large and a control 170.0-330.8 97 115 1.2 1.2 0.8-1.7
group nearly three times as large. 330.9-1171.0 69 110 0.9 0.9 0.6-1.3
Analysis of low birth weight (Table 4) Per 250 unit change 244 333 0.9 0.9 0.8-1.1
indicated no association with water source Abbreviations: OR, odds ratio; THM, trihalomethane.
and a decreased risk with increasing num- aAdjusted as needed for potential confounders listed in text; if no confounders identified, the crude OR is
ber of glasses per day. Categorical analysis resented.
of THM concentration indicated the low- Referent category.
594 Environmental Health Perspectives

4. e0
A -- 9 9
and bottled water use may reflect a system- water with THM levels in the range of the strongest findings of Kramer et al. (5)
atic tendency for bottled water users to 100 ppb and above, which is the federal due to our method of selecting cases. Term
more comprehensively seek medical care standard. Preterm delivery was unrelated births who weighed >2500 g, a large pro-
for miscarriages or for women with height- to THM concentration but low birth portion of all SGA deliveries, were not
ened health concerns to drink bottled weight risk was reduced among women in selected as cases in our study.
water. Socioeconomic status may influence the lowest tertile of exposure with no The limitations in our study should be
both miscarriage identification and bottled increase in risk above that exposure level. noted. A sizable fraction of nonrespon-
water use, but the associations we reported Total dose of THM (incorporating THM dents (due to refusal, being untraceable, or
were adjusted for mother's education and concentration and amount consumed) having key water information unavailable)
family income. yielded little association with any of the raises the question of whether the partici-
A consistent pattern of decreasing risk outcomes. pants differed from nonparticipants in a
with increasing consumption of water sug- The miscarriage results have few prior manner that would distort exposure-dis-
gests either a genuine beneficial effect of studies to which they can be compared, but ease associations. The interview itself may
such consumption or a reporting artifact. appear not to support the previous observa- generate erroneous reports, particularly of
There are potential benefits of increased tion of decreased risk among bottled water the amount of water consumed. We did
fluid consumption during pregnancy since users (9). The absence of association with not ask about preparation of cold beverages
the plasma volume must expand markedly water source is consistent with the report of from tap water, such as frozen orange
in that period (12), but it is not clear that Aschengrau et al. (8), that risk was similar juice. Furthermore, we did not ask where
restricted fluid intake would influence the in Massachusetts communities served by the water was consumed (home, work, or
outcomes we addressed. With the available chlorinated versus chloraminated supplies. elsewhere), or whether home filters were
data, we could not examine total fluid con- To our knowledge, no previous study has used. Thus, inferences about chlorination
sumption or consider beverages prepared explicitly evaluated THM concentration in by-product exposure based on available
from tap water. Analysis of water source by relation to miscarriage. data are subject to error.
amount added little additional insight to Preterm delivery and low birth weight The link to water suppliers is likely to
this pattern. results may be compared to those from be accurate, but the assignment of a par-
Analysis of THM concentrations yield- Iowa (5) and New Jersey (6). The absence ticular THM score based on the nearest
ed some indication of an association with of association with preterm delivery in our measurement day is certain to contain
miscarriage, with a notably increased risk study is consistent with the lack of associa- error relative to the true THM values over
in the most highly exposed subset driving tion found in Iowa (5), but is not notably the etiologic period of interest. The THM
a linear dose-response pattern. Analysis by discrepant with the small associations sample is taken from an approximately
tertiles yielded little evidence of increased (ORs <1.5) found in New Jersey (6). The appropriate point in time at locations other
risk, whereas isolation of the most highly small increase in risk of low birth weight than the occupant's home. Furthermore,
exposed sextile generated a pronounced for the upper two tertiles in the present the pattern of home water use and ventila-
association, with an aberrantly low risk in study is likewise compatible with small tion could produce rather different expo-
the next to highest sextile. Although limit- increases reported in each of the other two sures even for a given tap water THM
ed by imprecision, these data encourage studies (5,6). We were not able to exam- concentration. Changes in water con-
further examination of women who drink ine risk of SGA births for comparison to sumption during the course of pregnancy
were not ascertained, requiring respon-
Table 4. Low birth weight in relation to drinking water characteristics: Alamance, Durham, and Orange dents to provide an average value for the
counties, North Carolina, 1988-1991 entire pregnancy that may differ from the
Cases Controls Crude OR Adjusted OR" 95% Cl consumption in the etiologically relevant
Water source time period. Most of these sources of error
Private wellb 63 114 1.0 1.0 are likely to be similar for cases and con-
Community 225 388 1.0 1.0 0.7-1.4 trols, yielding relative risk estimates that
Bottled 13 43 0.5 0.8 0.4-1.6 are biased toward the null value (13).
Water amount (glasses per day) Given the seasonal patterns in THM lev-
0 35 40 1.5 1.5 0.9-2.5 els, matching controls on time of birth
1-3b 150 261 1.0 1.0
could have further biased the results
4+ 116 244 0.8 0.6 0.6-1.1
Water source x amount towards the null value (13), but analyses
Private well/1-3b 32 50 1.0 1.0 adjusted for season (summer versus other)
Private well/4+ 22 48 0.7 0.8 0.4-1.6 did not differ substantially from those
Community/1-3 12 189 0.9 0.8 0.4-1.3 reported.
Community/4+ 86 173 0.8 0.8 0.5-1.4 On the other hand, along with Bove et
Bottled/l+ 13 43 0.5 0.6 0.3-1.6
THM concentration (ppb) al. (6), our study was among the first to
40.8-63.3b 48 110 1.0 1.0 try to link THM measurements in both
63.4-82.7 74 118 1.5 1.5 1.0-2.3 time and space to study subjects.
82.8-168.8 57 105 1.3 1.3 0.8-2.1 Availability of data on water ingestion and
Per 50 ppb change 178 333 1.1 0.9 0.6-1.4 a wide array of potential confounders dis-
THM doses (ppb x glasses/day) tinguishes this study from previous record-
44.0-169.9b 60 108 1.0 1.0
based investigations (5,6). Accuracy of
170.0-330.8 63 115 1.0 1.0 0.6-1.5
330.9-1171.0 55 110 0.9 0.8 0.5-1.3 identifying pregnancy outcomes is also
Per 250 unit change 178 333 1.0 1.0 0.8-1.2 certain to be improved using hospital data
Abbreviations: OR, odds ratio; THM, trihalomethane. as opposed to birth certificate informa-
aAdjusted as needed for potential confounders listed in text; if no confounders identified, the crude OR is tion. Finally, because of the extensive
resented. array of information we obtained through
Referent category. the interview, we were able to examine
Volume 103, Number 6, June 1995 595

5. i I ...
and control for confounding much more REFERENCES Jersey Department of Health, 1992.
effectively than studies based on birth cer- 7. U.S. Environmental Protection Agency.
tificates. 1. Rook JJ. Formation of haloforms during chlo- National interim primary drinking water regu-
The challenge in interpreting our rination of natural waters. J Soc Water Treat lations: control of trihalomethanes in drinking
Exam 23:234-243 (1974). water. Fed Reg 44 (231): 68624 (1979).
results and the literature as a whole is that 2. Morris RD, Audet A-M, Angelillo IF, 8. Aschengrau A, Zierler S, Cohen A. Quality of
we would like to distinguish between the Chalmers TC, Mosteller F. Chlorination, chlo- community drinking water and the occurrence
absence of association and the presence of rination by-products, and cancer: a meta-analy- of late adverse pregnancy outcomes. Arch
a modest association. To do so with confi- sis. Am J Public Health 82:955-963 (1992). Environ Health 48:105-113 (1993).
dence requires large studies with refined 3. [ARC. [ARC monographs on the evaluation of 9. Windham GC, Swan SH, Fenster L, Neutra
exposure assessment. Subject to some carcinogenic risk to humans, vol 52. Cobalt RR. Tap or bottled water consumption and
and cobalt compounds, chlorination of drink- spontaneous abortion: a 1986 case-control study
uncertainty, literature suggests that there is ing water and some compounds found in in California. Epidemiology 3:113-119 (1992).
not a strong association between THM drinking water. Lyon:International Agency for 10. Fenster L, Windham GC, Swan SH, Epstein
exposure and adverse pregnancy outcome Research on Cancer, 1991. DM, Neutra RR. Tap or bottled water con-
but provides some tentative suggestions 4. EPA/ILSI. A review of evidence on reproduc- sumption and spontaneous abortion in a case-
that risk of miscarriage (based on the pre- tive and developmental effects of disinfection control study of reporting consistency.
sent study) and low birth weight or small- byproducts in drinking water. Washington, Epidemiology 3:120-124 (1992).
DC:U.S. Environmental Protection Agency 11. Savitz DA, Brett KM, Evans LE, Bowes W.
for-gestational-age births (based on previ- and International Life Sciences Institute, 1993. Medically treated miscarriage in Alamance
ous studies) may be affected. More sophis- 5. Kramer MD, Lynch CF, Isacson P, Hanson County, North Carolina, 1988-1991. Am J
ticated approaches to exposure assessment JW. The association of waterborne chloroform Epidemiol 139:1100-1106 (1994).
through water quality models in the con- with intrauterine growth retardation. 12. Hytten FE. Weight gain in pregnancy. In:
text of rigorous epidemiologic study Epidemiology 3:407-413 (1992). Clinical physiology in obstetrics (Hytten F,
designs can be expected to help reduce the 6. Bove FJ, Fulcomer MC, Klotz JB, Esmart J, Chamberlain G, eds). Oxford:Blackwell
Dufficy EM, Zagraniski RT. Report on phase Scientific Publications, 1980;193-233.
uncertainty. IV-A: public drinking water contamination 13. Rothman KJ. Modern epidemiology.
and birthweight, fetal deaths, and birth defects. Boston:Little, Brown, and Company, 1986.
A cross-sectional study. Trenton, NJ:New
1995 Swine in Biomedical Research
THE INTERNATIONAL SYMPOSIUM
October 22-25, 1995
The Inn and Conference Center
University of Maryland, University College, College Park, MD
SPONSORED BY
THE UNIVERSITY OF MINNESOTA &
THE UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN
CONFERENCE AGENDA
Abstracts accepted relating to:
Transplantation Pharmacology
Nutrition Transgenics
Genetic Models Toxicology
Behavior Infectious Diseases
Immunology Physiology
Obesity Dermatology
Other Subjects
ABSTRACT DEADLINE IS JULY 15,1995
For Registration and Abstract Information contact:
SWINE IN BIOMEDICAL RESEARCH
Secretariat International Symposium
College of Veterinary Medicine
295 AS/VM Bldg., 1988 Fitch Avenue
St. Paul, MN 55108-6009
Internet: pigmodel~gold.tc.umn.edu
596 Environmental Health Perspectives