European Journal of Epidemiology 17: 89±95, 2001. Ó 2001 Kluwer Academic Publishers. Printed in the Netherlands.

Risk factors for and relationship between bacterial vaginosis and cervicitis in a high risk population for cervicitis in Southern Iran H. Keshavarz1, S.W. Du€y2, A. Sadeghi-Hassanabadi3, Z. Zolghadr3 & B. Oboodi3 1

Medical Research Council, Biostatistics Unit, Institute of Public Health, Cambridge; 2Department of Mathematics, Statistics and Epidemiology, Imperial Cancer Research Fund, London, UK; 3Shiraz University of Medical Sciences, Medical School, Shiraz, Iran Accepted in revised form 11 April 2001

Abstract. In 1990, a study of the reproductive habits and cervical pathology in women of the Qashgha'i nomadic tribe, resulted in a high prevalence of cervicitis. This led us to explore the likely infectious agents responsible for a such high prevalence; to assess the di€erence in cervicitis rates between nomadic and non-nomadic populations in the same area; and to determine the risk factors for and the relationship between cervicitis and bacterial vaginosis (BV). In 1996±1997 a study was carried out of 839 married women of the Qashgha'i, 274 of the Lor nomadic tribes, and 388 non-nomadic urban women. A gynaecological examination, Pap smear and vaginal secretion for assessing BV by gram staining were performed. Data was analysed by logistic regression. Backward stepwise regression was used to assess

multivariable e€ects on risk of cervicitis. The prevalence of cervicitis was 88% in the Qashgha'i, 85% in the Lor and 71% in the urban population. In the multivariate backward stepwise regression analysis for predictors of cervicitis in the Qashgha'i, signi®cant factors were decreased association with those over 40 (p < 0.004) and postpartum bleeding. In the Lor tribe the predictors were an increased risk after more than four pregnancies (p < 0.01) and the use of contraception. Among the urban population, the risk of cervicitis was increased with the use of oral contraceptive previously (p < 0.03) or currently (p < 0.01). BV was strongly associated with cervicitis, with a relatively high attributable risk. Both sexual and childbirth exposures may be associated with cervicitis in these populations.

Key words: Bacterial vaginosis, Cervicitis, Iran, Qashgha'i, Tribal women Introduction Gynaecologic infections are common worldwide. The problem is particularly serious in developing countries, where the health services are inadequately equipped and thus unable to detect and treat these infections. This problem is further complicated by the absence of subsidised diagnostic means to determine etiology [1]. Routine testing for these conditions is recommended in all sexually active patients and screening tests for gonococcal and chlamydial cervicitis have been successfully implemented in many clinics in developed countries. However, clinics in areas with limited resources may have diculty in implementing universal screening for these infections [2±5]. A previous study of the Qashgha'i nomadic population showed a high prevalence of cervicitis among the women of this population [6]. Another study conducted in a population with similarly high risk of cervicitis, resulted in ®ndings of independent associations of cervicitis with chlamydia and bacterial vaginosis (BV) [3]. Bacterial vaginosis is a polymicrobial condition in which a decrease in vaginal acidity and in the con-

centration of lactobacilli is accompanied by an increase of more than hundred fold in the concentration of other micro-organisms. No single micro-organism is detected in all women with BV. In clinics for sexually transmitted diseases the reported prevalence of BV has been between 24 and 37%. The epidemiology of the condition has some of the characteristics of a sexually transmitted disease and it is associated with the presence of Neisseria gonorrhoea and Chlamydia trachomatis [4]. So far, there has been no information on the prevalence of BV in nomadic or even in urban populations in Iran, where there is a Muslim population with the corresponding traditionally strict sexual lifestyle habits in both men and women with serial or multiple partners being rare, with the exception of occasional male polygamy in the form of more than one life partner. A very high population rate of cervicitis is likely to be due at least partly to infectious agents [3]. This high-risk population gives an opportunity to identify some of them. The aim of the present paper was to investigate the association of BV with cervicitis, which in turn will give further clues to the aetiology of the latter.

90 In this paper we report the prevalence of cervicitis and BV and also the relationship between them in the three high cervicitis risk populations among the married women of the Qashgha'i nomadic tribe, the Lor nomadic tribe and the non-nomadic urban population in Southern Iran. The Qashgha'i and the Lor are pastoral tribes, native to the Fars province in Southern Iran. The movements of the Qashgha'i and Lor peoples follow a regular, yearly pattern, moving from summer pastures in mountainous regions to winter pastures in the lowlands. Their main source of subsistence is domestic livestock (sheep and goats), that graze on natural pastures. They live for part of the year (all of the year in the case of some Qashgha'i tribespeople) in tents and have a tribal social structure. The Lor population mainly live in stone houses in the winter. The women of these tribes have a lifestyle characterised by hard physical labour [6]. Their access to health care, and obstetric care in particular, is very limited. The nomadic lifestyle also entails a lack of general hygienic and sanitary facilities. Traditionally, this is a society of high parity, and families of seven or more children are common. Our previous study showed a high prevalence of cervicitis in Qashgha'i women. The third population is the urban married women from Shiraz in Southern Iran. These women have a lifestyle closer to western populations in social and economic terms but they still have generally high parity and high rates of cervicitis [7]. The motivation for the present study was to determine the major risk factors of cervicitis, BV and the relationship between the two conditions. We examined reproductive and non-reproductive factors in the study questionnaire to ascertain which, if any, were associated with cervicitis and BV or their severity. These might have implications for both aetiology and prevention. Material and methods Subjects and data collection We carried out a study of married women members of the Qashgha'i tribe, another of married women members of the Mamasani Lor tribe and a third of consecutive urban married women attending a large central OB-GYN clinic in Fars province in Southern Iran. Recruitment took place in 1996 and 1997. The women of the Qashgha'i and the Lor population were selected by a strati®ed random sample, in two tiers to yield numbers of women in proportions representative of the six tribes of the Qashgha'i, of clans within the tribes, and of the four tribes of the Lor population. The sample units were the subclans within the clans (locally known as Bonko). The subclans were selected randomly. The study team invited all married women in each selected bonko. These covered a range

of 450 km. All married women in each selected bonko who were not pregnant or menstruating participated in the study. Number of participant women in the selected bonkos ranged from 7 to 50. The Qashgha'i sample size was planned to be at least 700, to give 95% power to detect a doubling of risk of cervicitis associated with BV, on the basis of anticipated prevalences of 30% for BV and 80% for cervicitis. In the other two studies, the urban and Lor women, the main objective was the estimation of the rate of cervicitis. We therefore planned to recruit 300 women in each of the urban and Lor studies to give an anticipated con®dence interval of no more than 75±85% around an estimated prevalence rate of 80%. In the time available, we recruited 839 Qashgha'i 270 Lor and 388 urban women. Those with readable cervical smear data totalled 822 Qashqua'i, 258 Lor and 315 urban. Our team consisted of one general practitioner, one midwife and one epidemiologist. The women of the selected bonko were announced by the chief of the bonko at least 1 day before the team settled. The study questionnaire was completed by interview by the epidemiologist. The questionnaire consisted of identity and demographic information, family, obstetric, gynaecological and menstrual history, contraceptive use, the place of child bearing, smoking habit and economic history. The women were given a vaginal speculum examination without lubrication. Using Ayer's (wooden) spatula cervical scrapes were collected for Papanicolaou test. All Papanicolaou smears were processed and interpreted by the Shiraz University of Medical Sciences laboratory, known as a reference laboratory in Shiraz province. All reports were checked by an obstetrician and gynaecologist. Cervicitis was classi®ed by degree of in¯ammation, using the criteria of Kiviat et al. [8] as used in previous studies by ourselves [6, 7] and others [9, 10]. The criteria used by the cytologists±pathologists for reporting cervical in¯ammation were: polymorphonuclear leukocytes (PMNLs) recorded as 0 if none was seen, 1‡ (mild) if the cellular morphology of less than one fourth of the squamous cells was obscured by PMNLs, 2‡ (moderate) if one fourth to one half of the squamous cells were obscured, and 3‡ (severe) if more than one half were obscured. The vaginal smear for subsequent gram staining was obtained after the assessment of clinical signs and before the Papanicolaou smears were taken. A vaginal smear was obtained by rolling a swab across the vaginal wall and then onto a glass slide. The smears were ®xed by air drying and gram stained by using safranin as counterstain. Each gram-stained smear was evaluated for the following morphotypes under oil immersion (1000 magni®cation): large gram-positive rods (lactobacillus morphotypes), small gram-variable rods (G. vaginalis morphotypes), small gram-negative rods (Bacteroides spp. morphotypes), curved gram-variable rods (Mo-

91 biluncus spp. morphotypes), and gram-positive cocci. The quantitated morphotype was assessed for the diagnosis of BV. We used the Nugent score for BV. The scoring criteria summed the weighted quantitation of the following morphotypes to yield a score of 0±10 for each person: large gram-positive rods (lactobacillus morphotypes-weighted such that absence yielded the highest score), small gram-negative to -variable rods (G. vaginalis and Bacteroides spp. morphotypes), and curved gram-negative rods (Mobiluncus spp. morphotypes). A score of 0±3 was considered normal, a score of 4±6 was considered intermediate, and a score of 7 or higher was considered as severe BV. Typically a score in excess of 6 would lead to further investigation and intervention with drug therapy, whereas a score of 4±6 would be regarded as `abnormal vaginal ¯ora' or borderline BV and would lead to follow-up and repeat testing [11]. In this study, we were interested not only in BV as a clinical entity in itself but also in the association of borderline BV or worse with cervicitis (as de®ned by in¯ammation above), and so a score of 4 or more was classi®ed as a case of BV. Statistical methods Data were analysed by univariate and multiple logistic regression ®rst, with cervicitis as the outcome, then with BV as the outcome. Backward stepwise regression was used to assess multivariate e€ects on risk, after ®rst removing all those variables not univariately signi®cant at 10% level. We also calculated the attributable risk of cervicitis associated with BV [12]. Results Table 1 shows the basic characteristics of the study subjects. The prevalence of cervicitis was 88% in the

Qashgha'i, 83% in the Lor and 71% in the urban population. No dysplasia was observed. Cytology results including presence and severity of cervicitis, were available for 1395 women in the three populations. All three populations had high parity, with the highest being observed in the Qashgha'i. The relationship between BV, cervicitis and population is shown in Table 2. The prevalence of BV in the Qashgha'i tribe was 50%, in the Lor tribe 49% and in the urban population 40%. More cases with BV had moderate to severe cervicitis. Table 3 shows e€ects of non-reproductive factors on risk of cervicitis. There was a suggestive association of a lower risk of cervicitis associated with age more than 40 in the Qashgha'i tribe. No other signi®cant e€ects were observed. Table 4 shows the e€ects of reproductive factors on cervicitis. Signi®cant or borderline signi®cant increases in risk were observed due to the use of contraceptive pill at the time of study (p ˆ 0.06), use of any contraception (p ˆ 0.05) and birth in maternity hospital (p ˆ 0.006) and a signi®cantly lower risk was noted with postmenopausal status in the Qashgha'i tribe (p ˆ 0.01). The Lor population showed signi®cant increases in risk with high parity, in terms of numbers or either pregnancies (p ˆ 0.003) or deliveries (p ˆ 0.004), previous use of contraception pill (p ˆ 0.03), use of any contraception (p ˆ 0.01), and history of ever using a traditional midwife or neighbour as a birth assistant (p ˆ 0.01). In the urban population, signi®cant increases in risk were observed due to the use of contraceptive pill either before (p ˆ 0.01) or during (p ˆ 0.04) the study. There was a suggestive association of cervicitis with frequent sexual intercourse (p ˆ 0.09). Results of the multivariate backward stepwise regression analysis for cervicitis are given in Table 5. In

Table 1. Basic characteristics of the study populations Population Variable

Category

Qashgha'i

Lor

Age

<20 20±29 30±39 40±49 50‡ Premenopausal Postmenopausal 0 1±3 4±6 7‡ None Mild Moderate Severe

8 237 254 195 142 663 176 33 180 250 376 96 250 429 47

4 69 95 55 43 253 17 15 48 96 111 45 78 128 7

Menopausal status Parity

Cervicitis status

(1) (28) (30) (23) (17) (79) (21) (4) (21) (30) (45) (12) (30) (52) (6)

Urban (1) (26) (36) (21) (16) (94) (6) (5) (18) (36) (41) (17) (30) (50) (3)

8 124 150 71 34 386 2 26 201 120 41 92 25 190 8

(2) (32) (39) (18) (9) (99.5) (0.5) (7) (52) (31) (10) (29) (8) (60) (3)

92 Table 2. Cervicitis status by BV in the study populations BV in population Qashgha'i

Lor

Cervicitis

Present (%)

Absent (%)

None Mild Moderate Severe

31 106 190 32

42 116 185 7

Total

359

(8) (29) (53) (10)

(12) (33) (53) (2)

Present (%) 18 37 57 5

350

Urban Absent (%)

(15) (32) (49) (4)

22 36 62 1

117

(18) (30) (51) (1)

121

Present (%) 17 3 45 4

Absent (%)

(25) (4) (65) (6)

41 9 54 1

69

(39) (9) (51) (1)

105

Table 3. Univariate OR's-e€ect of non-reproductive variables on risk of cervicitis in each study population Population Qashgha'i Variable

OR (95% CI)

Age 40‡ Household size 8‡ Formal education Smoke cigarette Smoke hookah High economic score Age at marriage 21‡ Married for 30‡ years

0.52 1.20 1.18 0.70 0.75 1.06 0.66 0.75

a b

a

(0.34±0.81) (0.77±1.85) (0.66±2.12) (0.20±2.47) (0.47±1.19) (0.64±1.74) (0.39±1.13) (0.46±1.29)

Lor

Urban

OR (95% CI)

OR (95% CI)

0.68 (0.35±1.31) 1.15 (0.57±2.16) 1.69 (0.67±4.25) ±b 2.01 (0.79±5.10) 0.82 (0.31±2.15) 0.83 (0.26±2.61) 0.70 (0.33±1.45)

0.81 0.76 1.32 1.21 0.58

(0.46±1.40) (0.40±1.45) (0.81±2.15) (0.12±11.80) (0.24±1.41) ±b 075 (0.42±1.34) 0.62 (0.31±1.23)

Crude OR for any cervicitis as opposed to none. Not estimated due to sparse data.

Table 4. Univariate ORs-e€ect of reproductive variables on risk of cervicitis in each study population Population Qashgha'i

Lor

Urban

Variable

OR (95% CI)

OR (95% CI)

OR (95% CI)

Age at ®rst pregnancy 21‡ Four or more pregnancy History of abortion Four or more deliveries Postmenopausal Polygamy Intercourse 3‡ times/week Intercourse in menses Pill (before) Pill (now) Contraception (any±now) Birth in hospital Birth in maternity hospital Birth assistant*

0.96 1.80 0.76 1.29 0.54 0.64 1.04 1.03 1.26 1.75 1.52 1.00 1.91 0.99

0.83 2.79 0.81 2.63 0.98 0.41 0.83 1.01 2.06 1.10 2.19 3.13 1.76 2.57

0.86 0.82 0.76 0.91

(0.58±1.61) (0.72±1.93) (0.49±1.28) (0.81±2.06) (0.34±0.87) (0.33±1.24) (0.59±1.85) (0.65±1.64) (0.78±2.04) (0.94±3.16) (0.98±2.34) (0.48±2.09) (1.19±3.08) (0.60±1.67)

(0.33±1.61) (1.40±5.55) (0.36±1.83) (1.33±5.22) (0.27±3.59) (0.16±1.09) (0.40±1.75) (0.50±2.06) (1.06±4.03) (0.47±2.58) (1.14±4.23) (0.71±13.7) (0.86±3.62) (1.21±5.48)

0.60 1.56 0.71 1.87 2.48 1.35 0.92 1.57

(0.49±1.53) (0.50±1.34) (0.44±1.30) (0.56±1.50) ± (0.16±2.19) (0.92±2.66) (0.42±1.20) (1.11±3.15) (1.00±6.16) (0.78±2.34) (0.53±1.58) (0.86±2.84) ±

* Traditional midwife or neighbour.

the Qashgha'i, signi®cant or borderline signi®cant factors were a decreased risk associated with age group above 40 (p ˆ 0.004) and of postpartum

bleeding (p ˆ 0.09). In the Lor tribe, the remaining factors after stepwise regression were increased risk with more than four pregnancies (p ˆ 0.01) and use

93 Table 5. Backward stepwise regression results for predictors of cervicitis in each study population Cervicitis sample

Risk factor

Category

No

Yes

Qashgha'i

Postpartum bleeding

No Yes No Yes

85 11 44 51

(11) (20) (9) (16)

677 45 449 275

No Yes No Yes

18 27 28 18

(30) (14) (24) (12)

No Yes No Yes

86 6 64 27

(31) (15) (34) (22)

Age 40‡ Lor

Pregnancy 4‡ Oral contraceptive (any)

Urban

Oral contraceptive (now) Oral contraceptive (before)

of contraception (p ˆ 0.06). Among urban women, the risk of cervicitis was increased with prior (p ˆ 0.03) and current (p ˆ 0.01) use of oral contraceptives. The same univariate analyses were performed for risk of BV (data not shown for brevity but available from the authors). In the Qashgha'i tribe, signi®cant increases in risk were observed for age 40 or more (p < 0.001), being married for more than 30 years (p < 0.001), high parity, in terms of numbers of either pregnancies (p ˆ 0.004) or deliveries (p < 0.001), postmenopausal status (p < 0.001), birth in maternity hospital (p ˆ 0.006) and traditional midwife or neighbour as birth assistant (p < 0.001). Signi®cant decreases in risk were observed for use of contraceptive pill before (p ˆ 0.02) or during (p ˆ 0.009) the study, formal education (p < 0.001) and use of any contraceptive (p < 0.001). There was a suggestive increase in risk in case of a history of abortion (p ˆ 0.09). In the Lor population, signi®cant increases in risk were observed for postmenopausal status (p ˆ 0.02), lower risk for use of any contraception (p ˆ 0.05) and a suggestive increase in risk in case of a history of abortion (p ˆ 0.09). For the urban population, the only signi®cant association was a decreased risk with a history of birth in hospital (p ˆ 0.002). Backward stepwise regression for predictors of BV resulted in increased risks with postmenopausal status (p < 0.001) and traditional midwife or neighbour as birth assistant (p ˆ 0.02), and a decreased risk with use of any contraceptive (p ˆ 0.02) in the Qashgha'i tribe. In the Lor tribe an increased risk with having been married for more than 30 years (p ˆ 0.002) was the only signi®cant predictor. The risk of BV in the urban population was decreased for birth in hospital (p ˆ 0.01). Detailed results are available from the authors. Noting from Table 2 that the presence of BV was similarly associated with an increased prevalence of cervicitis in all three populations, we tested for hetr-

OR

95% CI

(89) (80) (91) (84)

1.00 0.55 1.00 0.48

± (0.27±1.11) ± (0.29±0.79)

41 172 86 126

(70) (86) (76) (83)

1.00 2.51 1.00 1.88

± (1.24±5.08) ± (0.95±3.69)

190 33 124 98

(69) (85) (66) (78)

1.00 1.96 1.00 2.68

± (1.15±3.33) ± (1.07±6.70)

ogeneity between populations of the e€ect of BV on risk of cervicitis. There was no (p ˆ 0.6) signi®cant heterogeneity. OR's for cervicitis associated with the BV were 1.44 (95% CI: 0.88±2.35), 1.22 (0.61±2.42) and 1.96 (0.99±3.85) in the Qashgha'i, Lor and urban populations respectively. We therefore assessed the e€ect of BV on risk of cervicitis in all three populations combined with and without adjustment for all variables which were signi®cant predictors of cervicitis for any population in the multivariate analysis (Table 5). Adjusted only for study population, the OR for any cervicitis associated with BV was 1.50 (95% CI: 1.06±2.11). Adjusted for the factors in Table 5, the OR was 1.61 (95% CI: 1.13±2.28). For severe cervicitis, the OR adjusted only for study population was 5.02 (95% CI: 2.41±10.46). Adjusted for the factors in Table 5, the OR was 4.15 (95%CI: 1.96±8.78). Attributable risk calculations suggest that BV may account for 19% of total cervicitis cases and 60% of severe cases. Discussion In this study, we found a signi®cant increase in the risk of cervicitis in association with contraceptive usage in the non-nomadic urban population and the semi-sedentary Lor population. Although there were univariate e€ects of contraception in the Qashqua'i population, these were not signi®cant when adjusted for age and postpartum bleeding. In the Lor and the Qashgha'i, the most commonly used contraceptive method was the oral contraceptive pill (44% in the Qashqua'i and 31% in the Lor), with tubal ligation (22%) coming second in the Qashqua'i and intra-uterine device (20%) in the Lor. In the urban population, the most commonly used method was withdrawal (26%), with tubal ligation (22%) coming second. It may be that in the Lor and urban populations, in¯ammatory changes are more related to

94 infections transmitted in coitus, whereas in the Qashqua'i they are associated with infections contracted during childbirth. This is possible, since among the Qashqua'i, birth usually takes place in a tent without professional medical support. The association of cervicitis with contraception use is consistent with previous results in the USA [13]. In an Indian population with similar cervicitis rates to the Qashqua'i, there was an association of high parity with cervicitis [3], again consistent with infectious agents being transmitted during childbirth (although it does not rule out a sexually transmitted aetiology). It is of interest that no dysplasia cases were observed in this population with a high rate of in¯ammatory changes. Low rates of dysplasia have also been observed in other populations, notably in the developing world, with high rates of in¯ammation [3, 14]. It is likely that there are some cases of dysplasia in such populations which are undiagnosed due to the fact that severe in¯ammation of cervical smears makes the dysplasia more dicult to detect. Analysis of e€ects on risk of BV produced similar results to those for cervicitis but suggested both sexual transmission of infection and birth practices as in the aetiology of BV in other populations [13, 15]. The sexually transmitted element is consistent with the literature, in which BV has been observed to be correlated with sexually transmitted agents such as HPV and chlamydia in both developed [14] and developing countries [3]. We found a negative association of BV with hospital birth in the urban women and positive association of using a neighbour or traditional midwife as birth assistant in nomadic Qashqua'i women. The association of BV with childbirth does not seem to be well researched elsewhere. Finally we found a very strong positive association between moderate BV and cervicitis. This association persisted even after adjustment for other risk factors for cervicitis. Attributable risks were also higher for severe cervicitis, suggesting that BV may be implicated in 60% of cases of severe cervicitis cases. Other studies have found similar results with respect to presence of cervicitis in populations from India [3, 15] and the USA [13]. BV has also been associated with increased risk of pelvic in¯ammatory disease [16, 17]. To our knowledge, this is the ®rst study demonstrating an enhanced correlation of BV with severe cervicitis. Clearly, there are other agents which need to be taken into account in order to reliably quantify the role of BV in cervicitis, notably HPV and Chlamydia trachomatis, both of which have been implicated in the aetiology of in¯ammatory atypia in the past [3, 13]. Facilities for testing for these agents have not been made available and we are currently seeking funds to ®nance the testing of samples abroad. In conclusion, we have found some evidence of contraction of infections during childbirth as associ-

ates of both BV and cervicitis. We have also found evidence that BV is itself strongly associated with cervicitis, particularly severe cervicitis. Acknowledgements The study forms a part of a PhD dissertation in the Medical Research Council±Biostatistics Unit in Cambridge UK with generous support from the Ali-Reza Soudavar foundation. The ®eld work was supported by Vice Chancellor for Research, Dr Mohamad Reza Pangeh-Shahin, in Shiraz University of Medical Sciences and the Director of the Organization for the nomadic people in Fars province, Mr Mansoori. Special thanks are due to Miss Venus Farhadi, Mrs Azar Riahi and Furoogh Dehbozorgian for the contribution during ®eld work and Dr Nilli for reviewing the Pap smears. We thank the women who participated in the study. References 1.

Singh V, Sehgal A, Satynarayana L. Clinical presentation of gynecologic infections among Indian women. Gynecolog Infect 1995; 85: 215±219. 2. Chacko MR, Kozinetz CA, Hill R, et al. Leukocyte esterase dipstick as a rapid screening test for vaginitis and cervicitis. J Pediatr Adolesc gynecol 1996; 9: 185±189. 3. Singh V, Sehgal A, Satynarayana L, et al. Association between reproductive tract infections and cervical in¯ammatory epithelial changes. Sexually Transmitt Dis 1995; 22: 25±30. 4. Hay PE, Taylor-Robinson D, Lamont RF. Diagnosis of bacterial vaginosis in a gynecology clinic. Br J Obstet Gynaecol 1992; 99: 63±66. 5. Nugent RP, Krohn MA, Hillier SH. Reliability of diagnosing bacterial vaginosis is improved by a standardized method of gram stain interpretation. J Clin Microbiol 1991; 29: 297±301. 6. Keshavarz H, Du€y SW, Sotodeh-Maram E, et al. Factors related to cervicitis in Qashgha'i nomadic women of southern Iran. Rev Epidem et Sante Publ 1997; 45: 279±285. 7. Du€y SW, Keshavarz H, Sadeghi-Hassanabadi A, et al. Prevalence of cervicitis in women living in southern Iran. Rev Epidem et Sante Publ 1999; 47: 86. 8. Kiviat NB, Paavonen JA, Brockway J, et al. Cytologic manifestations of cervical and vaginal infections I epithelial and in¯ammatory cellular changes. JAMA 1985: 253: 989±996. 9. Lawley TB, Lee RB, Kapela R. The signi®cance of moderate and severe in¯ammation on class I papanicolaou smear. Obstet Gynecol 1990; 76: 997±999. 10. Rantala I, Kivinen S. Demonstration of Chlamydia trachomatis in Papanicolaou-stained gynecological smears. Eur J Clin Microbiol Infect Dis 1998: 17: 46±48. 11. Martin HL, Richardson BA, Nyange PM, et al. Vaginal lactobacilli, microbial ¯ora, and risk of human immunode®ciency virus type 1 and sexually transmitted disease acquisition. J Infect Dis 1999; 180: 1863±1868.

95 12. Breslow NE, Day NE. Statistical methods in cancer research. Lyon: International Agency for Research on Cancer, 1980. 13. Paavonen J, Critchlow CW, DeRouen T, et al. Etiology of cervical in¯ammation. Am J Obstet Gynecol 1986; 134(3): 556±564. 14. Parashari A, Singh V, Gupta MM, et al. Signi®cance of in¯ammatory cervical smears. APMIS 1995; 103: 273± 278. 15. Larsson PG, Platz-Christensen JJ. Bacterial vaginosis and the vaginal leucocyte/epithelial cell ratio in women attending an outpatient gynaecology clinic. Eur J Obstet Gynecol Reproduct Biol 1991; 42: 217±220.

16. Peipert JF, Montagono AB, Cooper AS, et al. Bacterial vaginosis as a risk factor for upper genital tract infection. Am J Obstet Gynecol 1997; 177: 1184±1187. 17. Eschenbach A. Bacterial vaginosis and anaerobes in obstetric infection. Clin Infectious Dis 1993; 16: S282± S287. Address for correspondence: S.W. Du€y, Department of Mathematics, Statistics and Epidemiology, Imperial Cancer Research Fund, PO Box 123, Lincoln's Inn Fields, WC2A 3PX London, UK Phone: +44-207-269-2826; Fax: +44-207-269-3429 E-mail: s.du€[email protected]