SOCIOCULTURAL FACTORS IN PUERPERAL INFECTIOUSMORBIDITY AMONG NAVAJO WOMEN
W. THOMAS BOYCE,1 CATHERINE SCHAEFER,1 H. ROBERT HARRISON,'WILLIAM H. J. HAFFNER,4 MARGUERITE LEWIS," AND ANNE L. WRIGHT6
Boyce, W. T. (Dept of Pediatrics, U. of California, San Francisco, CA 94143),C. Schaefer, H. R. Harrison, W. H. J. Haffner, M. Lewis, and A. L WrightSoctocultural factors in puerperal infectious morbidity among Navajo women. AmJ Epidemiol 1989; 129:604-15.
From 1980 to 1982, a sample of 968 pregnant Navajo women in New Mexicowas enrolled in a prospective study of biologic and sodocultural factors inpuerperal infectious morbidity. Past studies have independently implicated bothgenital Infection and psychosodal stressors in perinatal complications, but, tothe authors' knowledge, no previous work has concurrently investigated theinteractive effects of genital pathogens and psychosocial processes. Endocer-vlcal cultures for Mycoplasma homtos and Chlamydia trachomatls were obtainedduring prenatal visits, and structured interviews were conducted assessing socialsupport and the degree of cultural tradttionalrty, in this context a proxy measureof accutturative stress. The incidences of postpartum fever, endometritis, andpremature rupture of membranes were significantly associated with the concur-rence of two factors: the presence of genital tract M. homlnls and a highlytraditional cultural orientation. When demographic and conventional obstetric riskfactors were controlled for, women with both M. homlnls and high tradttionalityexperienced Infectious complications at a rate twice that of women with eitherfactor alone. Among the plausible explanations for this result Is the possibilitythat accurturative stress undermines physiologic resistance to infectious genitaltract disease.
The role of genital organisms in abnor- and investigated for nearly two decades,mal pregnancy outcomes has been proposed Studies such as those by Jones (1) and
Received for publication January 14, 1988, arid in 400 Parnassus Avenue, San Francisco, CA 94143.final form September 7, 1988. Supported by a grant from the Thrasher Research
1 Division of Behavioral and Developmental Pedi- Fund. Dr. Harrison is the recipient of a John A. andatrics, Department of Pediatrics, University of Cali- George L. Hartford Fellowship.fornia, San Francisco, CA Presented in part at the May 1987 Annual Meeting
2 Department of Epidemiology and Public Health, of the Society for Behavioral Pediatrics, Anaheim,College of Medicine, Yale University, New Haven, CT. CA.
3 Centers for Disease Control, Atlanta, GA. The authors gratefully acknowledge the assistance4 Department of Obstetrics and Gynecology, Uni- of Dr. S. Leonard Syme and Dr. Stanislav V. Kasl in
formed Services University of the Health Sciences, reading previous drafts of this paper. Data were col-Bethesda, MD. lected through the efforts of Julia Arthur, with the
'Seattle-King County Department of Public collaboration of the medical staff from the Gallup andHealth, Seattle, WA. Crownpoint Indian Health Service Hospitals. Finally,
* Department of Pediatrics, University of Arizona, the principal author is particularly grateful for theTucson, AZ. insights and encouragement of Dr. Frank Loda and
Reprint requests to Dr. W. Thomas Boyce, Division the late Dr. John Cassel, who together provided a firstof Behavioral and Developmental Pediatrics, Depart- and memorable exposure to the ideas on which thement of Pediatrics, University of California, A-203, study was based.
604
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
others (2-9) have implicated genital colo-nization or infection with Mycoplasmahominis in a variety of puerperal compli-cations, including premature rupture ofmembranes (5), spontaneous abortion (2,3), low birth weight (4, 5, 9), endometritis(6, 7, 9), and postpartum fever (8). In ad-dition, other genital pathogens, such asChlamydia trachomatis and the group BStreptococcus, have been similarly proposedas etiologic agents in the infectious compli-cations of pregnancy (10-14).
However, as summarized by Harrison forstudies of mycoplasma infection, epidemi-ologic investigations have often yieldedconflicting results, suggesting a "need formore information on . . . the roles of non-infectious factors . . . in determining abnor-mal (infectious) outcomes of pregnancy"(15, p. 311). Because psychosocial processesmay alter host resistance to infections (16,17), factors such as psychologic stress andthe availability of social support may con-tribute to the observed variability in rela-tions between genital infections and preg-nancy complications.
Independent of biologic factors, a varietyof psychosocial phenomena have beenshown to influence the course and out-comes of pregnancy (18, 19). Early studies,such as those by Berle and Javert (20),Kapp et al. (21), and Drillien (22), noteddisproportionate numbers of emotionallydistressing events during the pregnanciesof women sustaining spontaneous abor-tions, prolonged labors, or prematurebirths. More recent work, employing bothretrospective and prospective designs, hassimilarly documented a modest but signif-icant association between various measuresof stressful experience and complicationsof pregnancy (23-25).
However, as noted by Chalmers (26),variables describing psychologically stress-ful events have never accounted for a majorproportion of the variance in pregnancycomplications. Partially in response to suchcriticism, investigators have become in-creasingly interested in the supportive, aswell as the stressful, aspects of social ex-
perience and have studied the capacity ofsocial support to buffer or ameliorate thehealth consequences of stress (27, 28).Nuckolls et al. (29), for example, found thatamong pregnant women with high levels ofstressful life change, those with strong so-cial support developed complications at arate only one third that of those with weakor nonexistent supports. Others (30, 31)have similarly reported beneficial effects ofsocial support on the course and outcomesof pregnancy.
Finally, in a previous paper, Boyce et al.(32) reported that both adherence to tra-ditional cultural practices and low levels ofsupportive personal interaction were inde-pendently associated with obstetric compli-cations in a population of pregnant Navajowomen. In that study, both traditionalityand lack of social support were interpretedas markers for social isolation, and preg-nancy complications were regarded as a setof potential health consequences stemmingfrom the stress of marginality.
A broad range of previous work thus im-plicates both biologic factors (i.e., infec-tious agents) and psychosocial factors (i.e.,stress, social support, and social marginal-ity) in the occurrence of abnormal preg-nancy outcomes. However, to our knowl-edge, no past studies have concurrently in-vestigated the separate and interactiveeffects of biologic pathogens and psycho-social processes. Because of the known in-fluences of such processes on the individ-ual's susceptibility to infectious disease,this study was designed to examine theinteraction of biologic and socioculturalrisk factors in the course and outcomes ofpregnancy. Navajo women were chosen asthe subjects of study because of the knowncultural heterogeneity among contempo-rary Navajo persons and the potential im-plications of such cultural differences forexperiences of stress and social isolation.In addition, prior work (33) with a Navajopopulation has demonstrated a high prev-alence of genital tract organisms known tobe associated with complications of preg-nancy.
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
A sample of 968 pregnant Navajo womenwas enrolled during the first prenatal visitto the obstetrics clinics at two IndianHealth Service hospitals in Gallup andCrownpoint, New Mexico. Enrollment wascompleted during the two-year period be-tween October 1980 and October 1982. Thestudy population was nonrandomly se-lected as a sample of convenience, compris-ing 40 per cent of the 2,421 women pre-senting for prenatal care during that period.Recruitment was conducted without priorknowledge of variables intrinsic to thestudy design. At the time of enrollment,basic demographic information was col-lected, and a structured interview was com-pleted to assess the availability of socialsupport and the woman's degree of culturaltraditionality. Interviews were conductedby a Navajo research assistant in either theNavajo or the English language, accordingto the subject's preference. In addition, en-docervical cultures for both M. hominis andC. trachomatis were obtained at the enroll-ment visit and again during a prenatal visitin the third trimester of pregnancy.
Within two months after delivery, a re-view of the medical record was completedwithout knowledge of prenatal interviewsor the results of cervical cultures. Data weregathered on the occurrence of four specificobstetric complications: postpartum fever,endometritis, premature rupture of mem-branes, and preeclampsia. These outcomeswere chosen to represent both a set of com-plications in which infection may play arole (postpartum fever, endometritis, andpremature rupture of membranes) and acomplication with no known or biologicallyplausible link with infection (preec-lampsia). Additional information was ob-tained from the medical record on anti-biotic treatment in the course of the preg-nancy, as well as on subjects' past medicaland obstetric histories. Although length ofhospital stay was not assessed, standardIndian Health Service obstetric practice atthe time of the study was hospitalization
for three postpartum days. It is thereforeunlikely that ascertainment of the targetedoutcome variables was biased by a short-ened observation period.
In accordance with standard definitionsof febrile morbidity (34), postpartum feverwas defined as a temperature equal to orgreater than 38 C on two occasions morethan 24 hours after delivery, and endome-tritis was defined as postpartum fever pluspurulent cervical discharge and/or uterinetenderness. Because of the frequent diffi-culty in clinically distinguishing postpar-tum fever from endometritis, these two di-agnoses were combined into a single out-come, designated postpartum fever/endometritis. Premature rupture of mem-branes was defined as that occurring onehour or more prior to the onset of labor.Again in accordance with standard criteria(35), preeclampsia was defined as the de-velopment of hypertension with protein-uria, edema, or both after the twentiethweek of gestation.
At the time of the enrollment interview,subjects reported their age, education, andthe income of the household in which theylived. The presence of medical conditionswith possible effects on the course of preg-nancy was ascertained, including anemia(hematocrit <28 per cent), diabetes, andpreexisting hypertension. Social supportwas measured using a weighted index ofinterview items modified from the instru-ment of Schaefer et al. (36) for use in aNavajo population. The five-item index as-sessed marital status and the availability ofboth emotional support and instrumentalsupport (or aid). For purposes of analysis,scores were collapsed into a dichotomousvariable representing high or low availabil-ity of social support.
Subjects' degree of traditionality wasevaluated using an abbreviated instrumentderived in part from the work of Milliganet al. (37). As previously reported (32), fac-tor analysis of interview responses revealedtwo factors, corresponding to the tradition-ality of cultural practices and the modern-ity of the home environment. Five items in
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
the first factor were retained as the measureof traditionally. These were 1) religiousaffiliation (Navajo Way or Native Ameri-can Church versus Christian or none), 2)use of corn pollen in ritual observances, 3)use of a traditional healer, 4) having had aKinaalda (Navajo puberty ceremony), and5) planning a Blessing Way ceremony forthe baby. In the subsequent analysis, tra-ditionality scores were converted to a two-level variable, representing the most tradi-tional (highest quartile) and the less tradi-tional (lowest three quartiles). Three itemsfrom the second factor were used as mea-sure of the modernity of the home. Theseitems were the presence of 1) electricity, 2)running water, and 3) a telephone.
Endocervical cultures were collected withcalcium alginate-tipped swabs immersed insucrose-phosphate buffer for C. trachoma-tis, and in trypticase soy broth with addedbovine serum albumin for M. hominis. Allspecimens were immediately frozen at —70C until inoculation. Standard culture meth-ods for the identification of C. trachomatisand M. hominis were followed and havebeen described in detail elsewhere (33). Forboth genital organisms, cervical cultureswere counted as positive if the organismwas isolated at the time of either enroll-ment or the final prenatal visit.
Table 1 summarizes the independent andoutcome variables assessed in the course ofthe study. Outcomes were treated as di-chotomous (present/absent) variables inthe analysis of data. Bivariate relationsamong pairs of independent variables andbetween independent and outcome vari-ables were examined using Pearson corre-lations and chi-square analyses, accordingto the continuous or categorical nature ofthe variables. At the multivariate level,stepwise multiple logistic regression wasused, following the approach of Kleinbaumet al. (38), to estimate the approximaterelative risk associated with an individualpredictor variable, while controlling for thepotentially confounding effects of other in-dependent variables. Hypothesized inter-action effects were tested by including
product terms (e.g., M. hominis X tradition-ality) in the regression equations after allmain effects had been added.
RESULTS
A total of 12 per cent of the sampledeveloped postpartum fever or endometri-tis. Premature rupture of membranes andpreeclampsia were found in 8 and 21 percent, respectively. The prevalence of thegenital organisms were 22 per cent for C.trachomatis and 50 per cent for M. hominis.As described in another report (9), the pres-ence of genital organisms was significantlyrelated to abnormal pregnancy outcomes incertain subgroups of the entire population,such as women delivered by cesarean sec-tion or those with a past history of spon-taneous abortion. However, as shown intable 2, no significant bivariate relationswere found between outcome variables andculture results in the study population as awhole.
Other independent variables, however,were significantly associated with abnor-mal outcomes at the bivariate level. Asdisplayed in table 3, postpartum fever/en-dometritis was significantly related to nul-
TABLE l
Independent and outcome variables for Navajowomen, New Mexico, October 1980 to October 1982
Independent variablesAgeEducationIncomeParityPast pregnancy complicationsPresent medical problemsGestational age at first prenatal visitAntibiotic treatment during pregnancyCesarean sectionModernity of homeSocial supportTraditionalityChlamydia trachomatisMycoplasma hominis
Outcome variablesPostpartum fever/endometritisPremature rupture of membranesPreeclampsia
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
Percentages of Navajo women with specified abnormalpregnancy outcomes and results of cultures for
Chlamydia trachomatis and Mycoplasma hominis,New Mexico, October 1980 to October 1982*
Culture results<»>
C. trachomatisPositive (22)Negative (78)
M. hominisPositive (50)Negative (50)
Postpartumfever/endo-
metntis
13.611.3
12.411.6
Prematurerupture of
membranes
9.47.8
8.97.1
Preeclamp-sia (%)
21.420.6
22.019.8
' No differences significant at p < 0.05 by chi-square with 1 df.
liparity, presentation for prenatal careearly in pregnancy, delivery by cesareansection, and low social support. In the caseof premature rupture of membranes, asso-ciations were found for parity alone, withnulliparous women again having higherrates. Finally, preeclampsia was signifi-cantly related to parity, the presence ofchronic medical problems, delivery by ce-sarean section, low social support, and highcultural traditionally. Many of the vari-ables, however, were significantly associ-ated with each other, raising the possibilitythat bivariate associations were con-founded by interrelations among indepen-dent variables. Parity, for example, washighly correlated with age (r = 0.47, p <0.001), suggesting that the relation of nul-liparity to preeclampsia could be due in partto the confounding effect of age.
Because of the potential for confounding,stepwise multiple logistic regression anal-yses were used to examine the relations ofthe predictor variables to the three selectedoutcomes of interest. In addition, since amajor focus of the study was the potentialinteraction between biologic and psycho-social variables, logistic regression analyseswere also used to test for significant effectsof interaction (cross-product) terms. Mul-tivariate analyses therefore employed thefollowing strategy for each of the threeoutcome variables. First, a limited model
was constructed in which the list of poten-tial independent variables included only C.trachomatis, M. hominis, traditionally, so-cial support, and their interaction terms.Forward stepping was used with the inter-action terms admitted to the model onlyafter all main effects had been added. Anyinteraction terms with probability values ofless than 0.20 were then included in a sec-ond equation, which tested the main effectsof all the potential independent variables(table 1) in addition to the interactive ef-fects of the surviving product terms. For-ward stepping and backwards eliminationwere used to arrive at the subset of termsthat represented the best compromise inachieving both validity and precision. Allterms of theoretical importance or withprobability values of 0.20 or less were re-tained in the final models. Results of boththe limited model (step 1) and the finalregression model (step 2) are presented intables 4-6, for the outcomes of postpartumfever/endometritis, premature rupture ofmembranes, and preeclampsia, respec-tively.
For postpartum fever/endometritis (ta-ble 4), a significant effect was found for theinteraction of M. hominis with traditional-ity in the limited regression model (p =0.05). Women with both a positive genitalculture for M. hominis and a highly tradi-tional orientation bore a risk of postpartumfever or endometritis approximately 2.7times that of women with either risk factoralone. In the expanded model, in whichother significant independent variables areincluded, the approximate relative risk as-sociated with the M. hominis X tradition-ality product term fell slightly to 2.4 (p =0.10). This decrement in relative risk indi-cates that the effect of the interaction maybe accounted for in part, but not com-pletely, by the confounding influence ofother significantly associated variables,such as parity, delivery by cesarean section,and gestational age at the first prenatalvisit.
In the case of premature rupture of mem-branes (table 5), a borderline significant
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
Risk of postpartum fever/endometritis among Navajo women by genital infection with Chlamydia trachomatis orMycoplasma hominis and other selected risk factors, New Mexico, October 1980 to October 1982*
Variables
Chlamydia trachomatis (present/absent)Mycoplasma hominis (present/absent)Social support (low/high)Traditionality (high/low)Caesarean sectionParity (1-3 vs.:)
0»4
Gestational age at first prenatal visit(trimester)
C. trachomatis x social supportM. hominis x traditionality
Limited modelt
Oddsratio§
1.00.71.50.9
2.02.7
95%CI|
0.52-1.960.40-1.180.87-2.600.44-1.78
0.73-5.571.04-6.75
Full
Odds ratio
1.00.81.30.96.0
2.91.2
0.61.82.4
1 modelt
95% CI
0.49-2.040.43-1.360.69-2.320.45-1.973.70-9.89
1.76-̂ 1.820.54-2.79
0.40-0.92-0.50-5.59
0.89-6.58
* Adjusted for the effects of other variables in the model.t Includes only C. trachomatis, M. hominis, traditionality, social support, and their interaction terms.t Includes all independent variables and interactions terms of theoretical importance or with probability
values of 0.20 or less.§ Approximate relative risk.| 95% CI, 95% confidence intervals.
TABLE 5
Risk of premature rupture of membranes among Navajo women by genital infection with Chlamydia trachomatisor Mycoplasma hominis and other selected risk factors, New Mexico, October 1980 to October 1982*
Variables
Chlamydia trachomatis (present/absent)Mycoplasma hominis (present/absent)Social support (low/high)Traditionality (high/low)Parity (1-3 vs.:)
0£ 4
Gestational age at first prenatal visit(trimester)
Chlamydia trachomatis x social supportMycoplasma hominis x traditionality
Limited
Oddsratio§
1.30.91.20.7
0.82.9
modelt
95% CI[
0.63-2.580.49-1.690.62-2.370.26-1.68
0.23-3.000.91-9.34
Full
Odds ratio
1.20.91.00.6
2.51.3
1.30.83.3
model}
95% CI
0.59-2.460.47-1.650.48-1.900.25-1.60
1.41-4.280.5O-3.16
0.86-1.990.23-3.061.00-10.59
* Adjusted for the effects of other variables in the model.t Includes only C. trachomatis, M. hominis, traditionality, social support, and their interaction terms,t Includes all independent variables and interaction terms of theoretical importance or with probability
values of 0.20 or less.§ Approximate relative risk.| 95% CI, 95% confidence intervals.
effect of the same interaction term wasfound in both the limited (p = 0.08) andthe full (p = 0.06) regression models. Thefinal model indicates that subjects withboth genital tract M. hominis and a highly
traditional life-style were 3.3 times morelikely to experience premature rupture ofmembranes than their counterparts withonly one of the two risk factors. As shownin table 5, a significant effect of parity was
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
Risk of preedampsia among Nauqjo women by genital infection with Chiamydia trachomatis or MycopUumahominis and other selected risk factors, New Mexico, October 1980 to October 1982*
Limited modelt FullmodeltVariables
Odds ratio! 96% CI| Approximateratiot
95% CI
Chiamydia trachomatis (present/absent) 0.9 0.51-1.44Mycoplasma hominis (present/absent) 1.1 0.72-1.65Social support (low/high) 1.7 1.13-2.62Traditionality (high/low) 1.3 0.79-2.26Parity (1-3 vs.:)
0£4
Present medical problems (yes/no)Modernity of home (no. of conveniences)Education (years) (0-8 years vs.:)
9-1112>12
C. trachomatis X social support 1.3 0.54-2.93M. hominis x traditionality 1.1 0.53-2.29
0.91.11.61.3
1.61.45.50.9
0.60.80.41.21.1
0.49-1.480.74-1.760.99-2.430.72-2.17
1.12-2.420.79-2.562.72-11.000.77-1.07
0.35-1.050.48-1.430.16-0.860.51-2.950.60-2.33
• Adjusted for the effects of other variables in the model.t Includes only C. trachomatis, M. hominis, traditionality, social support, and their interaction terms.t Includes all independent variables and interaction terms of theoretical importance or with probability
values of 0.20 or less.8 Approximate relative risk.| 95% CI, 95% confidence intervals.
also found, such that nulliparous womenhad a risk of premature rupture of mem-branes 2.5 times that of other subjects.
Table 6 shows that for preeclampsia, anabnormal outcome unrelated to infection,no significant main or interactive effectswere found for the genital organisms. Inthe limited model, a significant effect oflow social support was seen, but this wasno longer significant in the final modelcontrolling for the effects of other predic-tors. Significant associations with pre-eclampsia were found only for nulliparity,present medical problems, and maternaleducation. In particular, the interaction ofM. hominis and traditionality was, in thisinstance, unrelated to the outcome.
Figure 1 provides a more graphic pictureof the synergistic effects of M. hominis andtraditionality on the incidence of postpar-tum fever/endometritis and premature rup-ture of membranes. Using women with nei-ther M. hominis colonization nor a highlytraditional life-style as a point of reference,
figure 1 shows the adjusted relative risks ofthe two complications for the three othercategories of subjects: those with M. hom-inis alone, those with a highly traditionalorientation but no M. hominis, and thosewith both. No significant differences in riskwere associated with either risk factor inisolation. On the other hand, women withboth factors bore a risk of postpartum feverendometritis and premature rupture ofmembranes two to three times that ofwomen with only one or neither factor.
Finally, it is important to note that, inthe case of each outcome, tests for the sizeof the interaction effect have been madewhile controlling for the potentially con-founding influences of other, significantlyrelated independent variables. The effectsize of the M. hominis X traditionalityproduct term, for example, has been esti-mated, in the case of postpartum fever/endometritis, controlling for the effect ofcesarean delivery. Similarly, the logisticcoefficient for the interaction term in pre-
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
FIGURE 1. Synergistic effects of Mycopiasma hominis and traditionality on the incidence of postpartumfever/endometritiB and premature rupture of membranes. Mh+, Mycopiasma hominis-poaitive culture.
mature rupture of membranes was derivedfrom a logistic equation that controls forthe confounding influence of parity. Otherindependent variables not included in thefinal regression models were either foundnot to confound the interaction effects orwere not associated with the outcome inquestion.
DISCUSSION
The results indicate that in this sampleof pregnant Navajo women the develop-ment of infectious perinatal complicationswas related to the concurrence of two riskfactors: the presence of M. hominis in thewoman's genital tract and a highly tradi-tional cultural orientation. Women withboth factors experienced postpartum fever/endometritis or premature rupture of mem-branes at a rate two to three times that ofwomen with only one such factor. The termrepresenting the interaction of M. tracho-matis and traditionality attained only aborderline level of significance in the fullydeveloped logistic regression models. How-ever, the magnitude of the relative risksassociated with the interaction suggests aneffect of substantial proportion. It shouldbe noted that the method used in testingfor interaction effects is an inherently con-servative procedure. The independent vari-
ables M. hominis and traditionality are nec-essarily intercorrelated with the productterm M. hominis x traditionality, and theregression procedure assigns to the maineffects of M. hominis and traditionality allthe predictive contribution that cannot beunequivocally attributed to the productterm. The regression model thus exertsstringent criteria on the estimation of in-teraction effects. Finally, the absence ofeven a borderline significant M. hominis xtraditionality interaction in the case of thenoninfectious outcome, preeclampsia, lendsfurther strength to the validity of the inter-action found for infectious complications.
In an earlier report (32), we argued thatsince it is unlikely that traditionality per seinfluences complications of pregnancy, tra-ditionality should be regarded as a markerfor some other variable or condition withmore direct and biologically plausible influ-ences on the course of pregnancy. In thecase of the interactive effects of M. hominisand traditionality on the infectious morbid-ity assessed in this study, several possibleexplanations can be examined. First, thephysical home environment of highly tra-ditional Navajo women is likely to be prim-itive, with little or no access to runningwater and other modem conveniences. Insuch a setting, hygiene may well be com-
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
promised, and susceptibility to infectiousdisease may be augmented by the relativelack of sanitary facilities and conditions.This account of the observed interactioneffect is weakened, however, by the absenceof an association between infectious com-plications and the modernity of the homeenvironment. For both postpartum fever/endometritis and premature rupture ofmembranes, the variable describing themodernity of the home was not signifi-cantly related to the outcome and failed togain entrance into the regression model.The presence of running water (one com-ponent of the modernity scale) was sepa-rately analyzed and was also not signifi-cantly associated with either infectious out-come. It is therefore unlikely that theinteraction of M. hominis and traditionalityreflects simply an underlying effect of hy-giene on the relative incidence of infection.
Second, it is possible that the level oftraditionality is a correlate of some healthbehavior or set of health behaviors thatalters the likelihood of infection. It is con-ceivable, for example, that traditional Na-vajo women would be less likely to presentearly in pregnancy for prenatal care. Suchwomen may also be less likely to complywith antibiotic therapy prescribed duringthe course of pregnancy. Although thisstudy included no direct assessments ofhealth care utilization or compliance be-havior, two study variables offer at leastapproximate measures of related health be-haviors and antibiotic exposure. Gesta-tional age at the first prenatal visit wasincluded in the final regression model forboth postpartum fever/endometritis andpremature rupture of membranes and wasthus controlled for in the estimation of theinteraction effects. Antibiotic treatmentduring pregnancy, on the other hand, wasnot sufficiently associated with any out-come to merit inclusion in the regressionmodels. Neither variable was therefore alikely confounder of the interaction be-tween M. hominis and traditionality.
A final and, we believe, equally plausibleaccount of the observed interaction is the
possibility that extreme traditionality, inthe setting of this study, is associated withpsychologic processes that undermine ordiminish the person's physiologic capacityto resist disease. Our results suggest thatgenital colonization with M. hominisachieved maximum pathogenicity in a spe-cific sociocultural context, one character-ized by the retention of highly traditionalpractices and beliefs within a societyundergoing rapid acculturative change.Within such a context, it is reasonable topropose that highly traditional womenwould be most likely to experience the bi-ologic effects of stress and alienation, astheir families and friends move, both cul-turally and geographically, toward an alter-native way of life.
A growing body of work suggests thatstressful experiences may have importantinfluences on a person's immunologic com-petence and susceptibility to pathogenicagents (16, 17, 39-45). As summarized byPalmblad (17), Kiecolt-Glaser and Glaser(41), and others (43, 44), psychologic stresshas now been associated with a varietyof immunologic and reticuloendothelialchanges, ranging from involution of thethymus and spleen to suppression of inter-feron production and impairment of lym-phocyte cytotoxicity. In one example of theinfectious sequelae that may accompanysuch immunologic impairment, Kasl et al.(46) examined the psychosocial correlatesof clinical infectious mononucleosis in acohort of West Point cadets, all of whomhad serologic evidence of exposure to theetiologic agent, the Epstein-Barr virus.Among seroconvertors, the risk of clinicallyevident mononucleosis was significantly re-lated to the stress-producing combinationof high motivation and relatively poor acad-emic achievement. As in our study, thedevelopment of clinical disease was bestpredicted by the concurrent presence of apathogenic organism and circumstancescharacterized by chronic psychosocialstress.
There is evidence as well that local andsystemic immunity to genital organisms
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
plays an important role in preventing thetransition from genital tract colonizationto clinical infectious disease. Holmes (47)for example, in a review of mycoplasma-related morbidity, noted that lack of bac-teriocidal antibody to M. hominis correlateswith susceptibility to postpartum fever andthat the presence of antibody with comple-ment appears to protect against invasivemycoplasma infection. Osser and Persson(48) found that among patients with posi-tive genital tract cultures for C. trachoma-tis, those who developed salpingitis hadlower mean C. trachomatis-specific anti-body titers than those who did not. Thus,psychologic processes that impair immunecompetence may play significant roles inthe conversion of asymptomatic carrierstates into invasive disease.
This study provides evidence that psy-chologic and sociocultural factors play sig-nificant roles in mediating susceptibility topuerperal infection among women colo-nized with genital organisms. As suggestedby Syme (49) and Depue et al. (50), socio-cultural processes may create a generalphysiologic vulnerability to the agents ofdisease, with the specific category of illnessthen determined largely by the pathogensharbored or encountered by persons at risk.In 1965, Rene Dubos wrote that " . . . theprevalence and severity of microbial dis-eases are conditioned more by the ways oflife of the persons afflicted than by thevirulence and other properties of the etio-logical agents" (51, p. xxi). A view that hasbecome increasingly accepted within thefield of behavioral medicine is that biologicfactors, while necessary for the occurrenceof most diseases, are insufficient to accountfor the usual observed range in individualvulnerability. Until recently, what has notbeen widely held, at least as reflected in thedesign of psychosomatic investigations, isthe reciprocal view: that psychosocial fac-tors are similarly necessary, but not suffi-cient, to account for the occurrence of dis-ease. A fuller, more elegant understandingof illness susceptibility awaits substantial
and truly interdisciplinary collaboration byresearchers representing both biologic andsocial science perspectives. Psychologic andsociocultural processes do not operate inisolation from the biologic origins of dis-ease. By providing further evidence of sig-nificant interactions between biologic andpsychosocial risk factors, we hope that thisstudy will foster even greater interest in therich and complex interplay among deter-minants of human infectious disease.
REFERENCES
1. Jones DM. Mycoplasma hominis in abortion. BrMed J 1967;l:338-^0.
2. Driscoll SG, Kundsin RB, Home HW Jr, et al.Infections and first trimester losses: possible roleof mycoplasmas. Fertil Steril 1969;20:1017-19.
4. Braun P, Lee Y-Hm, Klein JO, et al. Birth weightand genital mycoplasmas in pregnancy. N Engl JMed 1971;284:167-71.
5. DiMusto JC, Bohjalian O, Miller M. Mycoplasmahominis type I infection and pregnancy. ObstetGynecol 1973;41:33-7.
6. Taylor-Robinson D, McCormack WM. The geni-tal mycoplasmas. N Engl J Med 1980;302:1003-10.
7. Thompson S, Lopez B, Wong K-H, et al. A pro-spective study of ChJamydia and Mycoplasma in-fections during pregnancy: relations to pregnancyoutcome and maternal morbidity. In: Mardh P-A,Holmes KK, Oriel JD, et al., eds. Chlamydialinfection. Amsterdam: Elsevier, 1982.
8. McCormack WM, Rosner B, Lee Y-H, et al. Iso-lation of genital mycoplasmas from blood obtainedshortly after vaginal delivery. Lancet 1975;1:596-9.
9. Berman SM, Harrison HR, Boyce WT, et al. Lowbirth weight, prematurity, and postpartum endo-metritis. JAMA 1987;257:1189-94.
10. Harrison HR, Alexander ER, Weinstein L, et al.The epidemiology and effects of cervical C. tra-chomatis and mycoplasmal infections in preg-nancy. JAMA 1983;250:1720-7.
11. Wager GP, Martin DH, Koutsky L, et al. Puer-peral infectious morbidity: relationship to routeof delivery and to antepartum Chlamydia tracho-matis infection. Am J Obstet Gynecol 1980;138:1028-33.
12. Martin DH, Koutsky L, Eschenback DA, et al.Prematurity and perinatal mortality in pregnan-cies complicated by maternal Chlamydia tracho-matis infections. JAMA 1982;247:1585-8.
13. Regan JA, Chao S, James LS. Premature ruptureof membranes, preterm delivery, and group Bstreptococcal colonization of mothers. Am J Ob-
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
21. Kapp FT, Hornstein S, Graham VT. Some psy-chologic factors in prolonged labor due to ineffi-cient uterine action. Compr Psychiatry 1963;4:9-18.
22. Drillien CM. The social and economic factorsaffecting the incidence of premature birth. J Ob-stet Gynecol Br Empire 1957;64:161-83.
23. Williams CC, Williams RA, Griawold MJ, et al.Pregnancy and life change. J Psychosom Res1975;19:123-9.
24. Newton RW, Hunt LP. Psychosocial stress inpregnancy and its relation to low birth weight. BrMed J 1984;288:1191-4.
25. Gorsuch RL, Key MK. Abnormalities of preg-nancy as a function of anxiety and life stress.Psychosom Med 1974;36:352-62.
26. Chalmers B. Psychosocial factors and obstetriccomplications. Psychol Med 1983;13:333-9.
27. Cohen S, Syme SL, eds. Social support and health.New York: Academic Press, Inc., 1985.
28. Broadhead WE, Kaplan BH, James SA, et al. Theepidemiologic evidence for a relationship betweensocial support and health. Am J Epidemiol1983:117:521-37.
29. Nuckolls KB, Cassel J, Kaplan BH. Psychosocialassets, life crisis and the prognosis of pregnancy.Am J Epidemiol 1972^5:431-41.
30. Norbeck JS, Tilden VP. Life stress, social support,and emotional disequilibrium in complications ofpregnancy, a prospective, multivariate study. JHealth Soc Behav l983;24:30-48.
31. Carr-Hill RA, Tahlin M, Johansson S. Pregnancy,social status and health in Sweden. Soc Sci Med1983;17:343-7.
32. Boyce WT, Schaefer C, Harrison HR, et al. Socialand cultural factors in pregnancy complicationsamong Navajo women. Am J Epidemiol 1986;124:242-53.
33. Harrison HR, Boyce WT, Haffner WHJ. Theprevalence of genital C. trachomatis and myco-plasmal infections during pregnancy in an Amer-
ican Indian population. Sex Transm Dis 1983;10:184-6.
34. Sweet RL, Ledger WJ. Puerperal infectious mor-bidity: a two-year review. Am J Obstet Gynecol1973;117:1093-1100.
35. Romeny SL, Gray MJ, Little AB, et al., eds.Gynecology and obstetrics: the health care ofwomen. (2nd ed.) New York: McGraw-HilL1981:688.
36. Schaefer C, Coyne JC, Lazarus RS. The health-related functions of social support. J Behav Med1981;4:381-406.
37. MiUigan BC, Dalton M, Swoboda VC, et al. Nurs-ing care and beliefs of expectant Navajo women.Am Indian Q 1984;8:83-101.
38. Kleinbaum DG, Kupper LL, Morgenstern H. Ep-idemiologic research: principles and methods.London: Lifetime Learning Publications, 1982:448-56.
39. Ader R, ed. Psychoneuroimmunology. New York:Academic Press, Inc., 1981.
40. Jemmott JB, Locke SE. Psychosocial factors, im-munologic mediation, and human susceptibility toinfectious diseases: how much do we know? Psy-chol Bull 1984;95:78-108.
41. Kiecolt-Glaser JK, Glaser R. Psychological influ-ences on immunity. Psychosomatics 1986;27:621-4.
42. Spector NH, ed. Neuroimmunomodulation: Pro-ceedings of the First International Workshop onNeuroimmunomodulation. Bethesda, MD: Inter-national Working Group on Neuroimmunomod-ulation, 1985.
43. Stein M, Keller SE, Schleifer SJ. Stress andimmunomodulation: the role of depression andneuroendocrine function. J Immunol 1985;135(suppl.):827-33.
44. Riley V. Psychoneuroendocrine influences onimmunocompetence and neoplasia. Science 1981;212:1100-9.
45. Lloyd R. Mechanisms of psychoneuroimmuno-logical response. In: Fox, BH, Newberry, BH, eds.Impact of psychoendocrine systems in canc«r andimmunity. Lewiston, NY: CJ Hogrefe, 1984:1-57.
46. Kasl SV, Evans AS, Niederman JC. Psychosocialrisk factors in the development of infectiousmononucleosis. Psychosom Med 1979;41:445-66.
47. Holmes KK. Mycoplasma hominis—a humanpathogen. Sex Transm Dis 1984;ll:159-63.
48. Osser S, Persson K. Postabortal pelvic infectionwith Chlamydia trachomatis and the influence ofhumoral immunity. Am J Obstet Gynecol 1984;150:699-703.
49. Syme SL. Sociocultural factors and disease etiol-ogy. In; Gentry, WD, ed. Handbook of behavioralmedicine. New York: Guilford Press, 1984:13-37.
50. Depue RA, Monroe SM, Shackman SL. The psy-chobiology of human disease: implications for con-ceptualizing the depressive disorders. In: DepueRA, ed. The psychobiology of the depressive dis-orders: implications for the effects of stress. NewYork: Academic Press, Inc., 1979:3-20.
51. Dubos R Man adapting. New Haven, CT: YaleUniversity Press, 1965.
at University of A
rizona on June 25, 2010 http://aje.oxfordjournals.org
