Smoking and pregnancy-related pelvic pain

Transcription

1 DOI: /j x Epidemiology Smoking and pregnancy-related pelvic pain K Biering, a E Aagaard Nohr, b J Olsen, c NH Hjollund, d,e A-M Nybo Andersen, f M Juhl g a Department of Occupational Medicine, Herning Regional Hospital, Herning, Denmark b Department of Epidemiology, Institute of Public Health, University of Aarhus, Aarhus, Denmark c Department of Epidemiology, School of Public Health, University of California, Los Angeles, California, USA d Department of Occupational Medicine, Herning Regional Hospital, Herning, Denmark e Department of Clinical Epidemiology, Aarhus University Hospital, Aarhus, Denmark f Division of Epidemiology, Institute of Public Health, University of Southern Denmark, Odense, Denmark g National Institute of Public Health, University of Southern Denmark, Copenhagen, Denmark Correspondence: K Biering MHSc, Department of Occupational Medicine, Herning Regional Hospital, DK-7400 Herning, Denmark. Karin.Biering@vest.rm.dk Accepted 7 April Published Online 26 May Objective To investigate possible associations between smoking and pregnancy-related pelvic pain. Design Nested case control study. Setting Denmark Population The Danish National Birth Cohort. Methods The women were interviewed twice in pregnancy and twice after childbirth. The first pregnancy interview provided information on smoking and possible confounding factors, whereas the first interview after birth addressed case identification. Cases (n = 2302) were defined on the basis of self-reported pelvic pain, and controls were selected among women who did not report pelvic pain (n = 2692). Logistic regression analysis was used to estimate associations between smoking and pelvic pain. Main outcome mreasue Pregnancy-related pelvic pain. Results Compared with non-smokers, women who smoked during pregnancy had an adjusted odds ratio of 1.2 ( ) for overall pelvic pain, similar to women who stopped smoking in early pregnancy 1.3 ( ). The equivalent adjusted odds ratio for severe pelvic pain was 1.2 ( ) for smokers, and 1.5 ( ) for women who stopped smoking. Smoking intensity, measured as number of cigarettes smoked per day, was associated with pelvic pain in a dose response pattern. Information about smoking was collected prospectively, which makes it unlikely that differential recall alone explains the results. Conclusions Smoking was associated with pregnancy-related pelvic pain, with a dose response pattern between reported smoking intensity and pelvic pain. These findings suggest a possible new risk factor for a common ailment during pregnancy. Keywords Epidemiology, pelvic pain, pregnancy, risk factors, smoking. Please cite this paper as: Biering K, Aagaard Nohr E, Olsen J, Hjollund N, Nybo Andersen A-M, Juhl M. Smoking and pregnancy-related pelvic pain. BJOG 2010;117: Introduction Pelvic pain during pregnancy may affect the ability to perform normal daily activities, such as walking, turning over in bed and getting up from a chair. 1 4 In Denmark, it is the most frequent reason for sick leave during pregnancy. 5,6 Pelvic pain normally starts in the second half of pregnancy, with the highest intensity between 24 and 36 weeks of gestation. In most cases the pain disappears spontaneously after birth, but for some women the pain persists, and may even become chronic. 4 Whether pelvic pain is an early marker of other joint diseases or an isolated discomfort in pregnancy is not known. Only little is known about the aetiology of this common pregnancy complaint, and most studies have been small, with retrospectively collected exposure data. Strenuous work, previous low back pain, previous pregnancyrelated pelvic pain, or low back pain have been suggested as risk factors. 4,7 Smoking has been examined as a risk factor, 3,8,9 with negative findings except in one study in a subgroup of women with symphysiolysis (OR 2.2; P < 0.05). 9 However, all studies had limited power to detect a possible low-to-moderate effect of smoking. Smoking has been associated with low back pain in nonpregnant populations, perhaps mediated by vasoconstriction and local ischaemia. 10,11 Others have questioned these findings and argued that lifestyle factors associated with smoking may confound the results A review concluded that smoking is associated with musculoskeletal diseases such as shoulder pain, low back pain and fibromyalgia/chronic widespread pain, but no underlying mechanism was suggested. 15 ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology 1019

2 Biering et al. Although smoking causes several serious problems related to fetal health, it is still a common exposure in pregnancy in some countries. The number of Danish primiparous women who reported smoking during pregnancy was 22% in 1997, and 16% in While some women stop smoking when they plan to become pregnant, most smoking women either postpone this decision to the first trimester of pregnancy, or try to reduce smoking intensity while pregnant. 21 The aim of this study was to estimate the association between smoking and pelvic pain in pregnancy using a case control study, nested in a large population-based pregnancy cohort with 2302 cases of pelvic pain, and with prospectively collected comprehensive information about smoking and potential confounding factors. Methods Material The Danish National Birth Cohort (DNBC) holds data from more than pregnancies, recruited from 1996 to 2002 by their general practitioner at the first antenatal visit. About half of the general practitioners in Denmark took part in the study, and about 60% of the women invited accepted the invitation. The participating women were interviewed by telephone twice in pregnancy (around 16 and 30 weeks of gestation), and when the child was 6 and 18 months old. The cohort is described in detail elsewhere, 22 and at This study is based upon responses from the first pregnancy interview, and the interview at 6 months after delivery. Methods A nested case control study on pregnancy-related pain was conducted within the cohort from April 2000 to November During this period, specific questions regarding pain localisation and pain severity in daily activities were added to the interview conducted 6 months after delivery, to identify cases and controls. Only women with singleton births were included. Information about smoking and possible confounding factors was collected in the first pregnancy interview. The average duration for the interviews were approximately 15 minutes. A version of the interview guide in English is available at Selection of cases and controls, and assessment of outcome Only women who completed both the first pregnancy interview and the first postpartum interview were eligible for this study. In the postpartum interview, all women were asked: Did you feel pelvic pain to an extent that affected your ability to walk during pregnancy or shortly after delivery? During the entire data collection period of the DNBC, women gave a positive answer to this question. During the recruitment period for the case control study, this question served as a screening question. A positive answer to this question indicated a potential case, and prompted additional questions about pain localisation. Potential controls were selected over 5 weeks, spread equally over the case recruitment period, among women who gave a negative response to the screening question. To validate the case/control status, both potential cases and controls were asked about pain intensity when performing five daily activities (turning over in bed, walking, standing up from a dining room chair, standing up from a sofa and walking on stairs). Mild cases were defined as women who reported some pain in at least one daily function and/or strong pain in not more than one daily function. Severe cases reported strong pain in at least two daily functions. The procedure for this categorisation is illustrated in Box 1. Because of logistic failures, 327 potential cases were not asked the additional questions, and were therefore excluded. Among the remaining 2546 potential cases, 178 reported pain outside the pelvic region, and 66 did not report pain in any daily activity, and were therefore excluded, resulting in 2302 cases for further analysis. Among 2854 potential controls, 162 reported pain equivalent to the severe case definition, and were excluded. Hence, the final study population comprised 739 mild cases, 1563 severe cases and 2692 controls. As 41% of the Box 1. Case definition according to the questions asked in the interview Questions: 1 Self-reported pelvic pain (screening question, addressed to all at the time of recruitment). Did you feel pelvic pain to an extent that affected your ability to walk during pregnancy or shortly after delivery? 2 Pain localisation (posed to the women who responded yes to question 1). Where did you feel pain? 3 Pain level in five daily functions (diagnostic questions, merged together here, posed to cases and controls). Did you feel pain when turning over in bed, when walking, when getting up from a dining room chair, when getting up from a sofa, when walking on stairs? (no pain, some pain, strong pain) Mild cases: 1 Yes to question 1. 2 Pain localised around the symphysis or in the lower back region, from the hip or below (question 2). 3 Some pain in at least one of the daily functions (question 3), and/or strong pain in not more than one of the functions. Severe cases: 1 Yes to question 1. 2 Pain localised around the symphysis or in the lower back region, from the hip or below (question 2). 3 Strong pain in at least two of the daily functions (question 3) ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

3 Smoking and pregnancy-related pelvic pain controls indicated pain in daily activities equivalent to the definition of mild cases, there was a considerable overlap of symptoms between mild cases and controls, where the only difference was the answer to the initial question. For this reason we present results for all cases and severe cases separately. Measurement of exposure In the first pregnancy interview the women were asked about smoking during early pregnancy and about current smoking status at the time of the interview. Current smokers and women who stopped smoking in early pregnancy were asked about the average number of cigarettes smoked per week or day, and were analysed separately. Ten women did not report on the intensity of smoking, but did report to be smokers/former smokers, and they were placed in the group with the lowest smoking intensity. Information on potential confounding variables was collected in the first interview, except for information about diabetes and urinary tract infection (UTI) during pregnancy, which were based on information from the first, second, and third interview. The selection of potential confounding factors was made a priori, based on a literature review. Statistical analysis We used multiple logistic regression to estimate crude and adjusted odds ratios for pelvic pain and severe pelvic pain, respectively, according to smoking status. Age, parity, prepregnancy body mass index (BMI), age at menarche, sociooccupational status, physically strenuous work, self-rated health, and concerns about giving birth and about the health of the child were included in the model as potential confounding factors. A priori we suspected age to interact with the association between parity and pregnancy-related pelvic pain. Furthermore, we expected that women with pelvic pain could avoid having more children, and as smoking affects fertility, parity could be considered to be a so-called collider. 23 We repeated the analysis without adjusting for parity, and subsequently by excluding all multiparous women. Testing for trends was examined using the score test for trend of odds. Analyses were performed in stata/ic 9.0 (StataCorp, College Station, TX, USA). Results Table 1 shows the distribution of controls, mild cases and severe cases of pelvic pain according to smoking and potential confounding factors. Compared with controls, women with pregnancy-related pelvic pain were more often smokers (19 versus 16%), or had stopped smoking in early pregnancy (11 versus 10%) (Table 1). Cases were more often multiparous, overweight or obese, of low sociooccupational status, and did more physically strenuous work. Also, they more often reported concern about giving birth and about the health of the unborn child. On average, women with pelvic pain gave birth to larger newborns than women without pelvic pain. Table 2 shows the odds ratios for pregnancy-related pelvic pain according to smoking habits. Both for continued smoking and for smoking only in early pregnancy, the odds ratios indicated an increased prevalence of pelvic pain (Table 2). There were no substantial differences between ex-smokers compared with continuing smokers, but the associations were slightly stronger for severe pelvic pain than for overall pelvic pain. Table 3 shows odds ratios for pregnancy-related pelvic pain according to smoking intensity. The odds for pelvic pain increased, as indicated by the trend test, with an increasing number of cigarettes smoked per day, both for severe pelvic pain and overall pelvic pain. This finding applied to all smokers, irrespective of whether they stopped smoking in early pregnancy or not (Table 3). No interaction on the multiplicative scale was found between age and parity (P = 0.06), so both variables remained in the model as possible confounding factors. In supplementary analyses we omitted adjusting for parity, and restricted the analyses to primiparous women, and this did not influence the estimates. Additional adjustment for birthweight marginally increased the odds ratios for smoking. We also included the 244 cases that were excluded from the main analyses because they did not report any pain in daily activities in a subsequent analysis. In another further analysis, we included the 162 controls that were excluded because of reporting pain equivalent to the severe case definition. In both these supplementary analyses, there were no significant changes in the estimates generated. Discussion We found an association between smoking and pelvic pain that persisted after adjustment for several background and lifestyle factors. The association was stronger for severe pelvic pain than for overall pelvic pain, and increased with an increasing intensity of smoking. This is the first study to indicate that smoking is a risk factor for pelvic pain, except in a subgroup of women with symphysiolysis, but previous studies may have been too small to detect an association of the magnitude observed in this report. 3,8,9 To our knowledge, our study contains the largest number of cases studied so far. Information about smoking and confounding factors was collected early in pregnancy before the expected onset of pregnancy-related pelvic pain. Pelvic pain was self-reported and was found to correlate well with clinical findings, as less than 5% of ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology 1021

4 Biering et al. Table 1. Maternal characteristics according to pregnancy-related pelvic pain in the DNBC, Variable Controls (n = 2692) Mild cases (n = 739) Severe cases (n = 1563) % n % n % n Smoking habits at first interview Non-smokers Stopped in early pregnancy Smokers Smoking intensity Non-smokers Fewer than 5 cig./day cig./day cig./day cig./day Age Under 25 years years years years Parity No previous births previous birth previous births or more previous births Body mass index(bmi) Under 18.5 kg/m kg/m kg/m kg/m kg/m Mean BMI (95% CI) 23.4( ) 23.9( ) 24.9( ) Height Up to 1.60 m m m m Socio-occupational status* High Medium Low Physically strenuous work Rarely Some times Often Not working/missing Concerns about giving birth Not at all Some A lot Concerns about the child s health Not at all Some ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

5 Smoking and pregnancy-related pelvic pain Table 1. (Continued) Variable Controls (n = 2692) Mild cases (n = 739) Severe cases (n = 1563) % n % n % n A lot Mean birthweigth in grams (95% CI) 3563 ( ) 3726 ( ) 3736 ( ) *Socio-occupational status was based on the current or most recent job within 6 months, or, if the woman was attending school, on the type of education: the high category included women in management or in jobs requiring higher education, generally more than 4 years beyond high school; office workers, service workers, skilled manual workers, and women in the military constituted the middle category; whereas unskilled workers and unemployed women were classified in the low category. Women who could not be categorised from their own information were categorised according to their husband s socio-occupational status. A total of 25 women could not be assigned a socio-occupational status. Table 2. Odds ratios for pregnancy-related pelvic pain according to smoking habits in the DNBC, Cases versus controls in brackets Smoking habits Crude Adjusted* All pelvic pain (2302/2692) (2072/2360) (n, cases/controls) Non-smokers Ref. Ref. Stopped in early pregnancy 1.2 ( ) 1.3 ( ) Smokers 1.3 ( ) 1.2 ( ) Severe pelvic pain (1563/2692) (1396/2360) (n, cases/controls) Non-smokers Ref. Ref. Stopped in early pregnancy 1.2 ( ) 1.5 ( ) Smokers 1.4 ( ) 1.3 ( ) *Adjusted for age, parity, BMI, age at menarche, socio-occupational status, physically strenuous work, self-rated health, concern about giving birth and about the health of the child, diabetes and urinary tract infections. women who reported pelvic pain did not fulfil the clinical diagnostic criteria. 1 Six months after delivery, the cohort of study participants answered the pelvic pain question that was used for screening in this nested study. The overall prevalence of self-reported pelvic pain in the DNBC was 18.5% ( out of women answered positively). This corresponds well with previous studies based on clinical examination, which reported on incidences of pelvic pain ranging from 14 to 33%. 3,24,25 The prevalence of pregnancy-related pelvic pain varies substantially from 7 to 72% in the literature because of differences in the definitions used to ascertain this outcome and methods of data collection, as there is no international consensus on the diagnostic criteria. Some studies include both pelvic pain and low back pain as a composite outcome measure, whereas others consider them to be two different outcomes. 4 In this study, the aim was to only include women with pain in the pelvic area, using a specific set of questions on pain localisation. However, we are aware that a number of women with low back pain (which may not be related to pregnancy) were probably also included, which will attenuate the association reported in this study if smoking is not related to this variable. On the other hand, only women who reported pelvic pain to an extent that affected their ability to walk in the initial screening question were included as cases. Pregnant women may under-report smoking. England et al. 26 found that 21.4% of pregnant women who reported that they stopped smoking in early pregnancy had evidence of active smoking based on urinary cotinine concentration. Windsor et al. 27 suggest a deception rate of at least 25%. We presume that this misclassification was not associated with case status, as information on smoking was collected in the first interview. Furthermore, a study by Nohr et al. 28 indicated that although smoking prevalence was lower in the DNBC cohort than in the source population, this did not bias the selected risk estimates when smokers were compared with non-smokers within the cohort. The reported smoking intensity may depend on whether the woman stopped smoking in early pregnancy, and therefore was more likely to report a dose similar to her prepregnancy behaviour, or the opposite, to remain a smoker, who reported her new reduced smoking habits. If smoking prior to pregnancy is a causal factor it could explain why we found a stronger association in women who stopped smoking early in pregnancy. We therefore report results separately for smokers and for women who stopped smoking. The dose response pattern was present in both of these strata, despite this possible incongruence in reporting. Regrettably, we do not have information about smoking habits before pregnancy, and cannot take lifetime smoking dose into account. Also, women who stopped smoking when the pregnancy was planned were categorised ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology 1023

6 Biering et al. Table 3. Odds ratios for pregnancy-related pelvic pain according to smoking intensity in tdnbc Cases versus controls in brackets Average smoking intensity All pelvic pain Severe pelvic pain Crude Adjusted* Crude Adjusted* All women (2302/2692) (2252/2633) (1563/2692) (1525/2633) Non-smokers Ref. Ref. Ref. Ref. <5 cig./day 1.1 ( ) 1.1 ( ) 1.1 ( ) 1.2 ( ) 5 9 cig./day 1.3 ( ) 1.3 ( ) 1.3 ( ) 1.4 ( ) cig./day 1.3 ( ) 1.1 ( ) 1.4 ( ) 1.3 ( ) 20 cig./day 2.0 ( ) 1.9 ( ) 2.4 ( ) 2.2 ( ) P = 0.001** P = 0.001** Stopped in early pregnancy (1855/2270) (1819/2226) (1242/2270) (1213/2225) Non-smokers Ref. Ref. Ref. Ref. <5 cig./day 0.9 ( ) 0.9 ( ) 0.9 ( ) 1.0 ( ) 5 9 cig./day 1.2 ( ) 1.5 ( ) 1.3 ( ) 1.5 ( ) cig./day 1.1 ( ) 1.3 ( ) 1.3 ( ) 1.6 ( ) 20 cig./day 2.0 ( ) 2.2 ( ) 2.3 ( ) 2.6 ( ) P = 0.02** P = 0.005** Current smokers at first interview (2061/2433) (2016/2385) (1395/2384) (1361/2384) Non-smokers Ref. Ref. Ref. Ref. <5 cig./day 1.2 ( ) 1.3 ( ) 1.3 ( ) 1.4 ( ) 5 9 cig./day 1.3 ( ) 1.2 ( ) 1.3 ( ) 1.3 ( ) cig./day 1.4 ( ) 1.0 ( ) 1.5 ( ) 1.1 ( ) 20 cig./day 2.1 ( ) 1.5 ( ) 2.5 ( ) 1.8 ( ) P = 0.006** P = 0.03** *Adjusted for age, parity, BMI, age at menarche, socio-occupational status, physically strenuous work, self-rated health, concern about giving birth and about the health of the child, diabetes and urinary tract infections. **Score test for trend of odds. as never having smoked, and this may attenuate the association if previous smoking plays a role. In the DNBC, about 30% of pregnant woman in Denmark participated during the recruitment period, and some attrition was present, as women of the who gave approval participated in the first interview, participated in the second interview, and participated in the third interview. This may cause selection bias if attrition is related to both exposure and case status. It is likely that women severely impaired by pelvic pain may be less prone to participate in the third interview because of disability caused by the pain, whereas an association with reported exposure is unlikely because information on smoking was collected in the beginning of pregnancy, before the onset of pelvic pain. The exclusion of potential cases who did not report pain during daily functions, and of potential controls who reported pain equivalent to severe cases, could bias the results, but including these subjects into a subsequent analysis did not lead to altered associations. A total of 327 cases were not included because of a logistic error, which we do not believe has biased the results. Information about pelvic pain was obtained 6 months after delivery, and mild cases may have forgotten their pain by then. This may introduce a risk of misclassification of outcome, which will attenuate the association if the reporting error is unrelated to their smoking status. We adjusted the association for lifestyle factors, but the results may still be confounded by unknown risk factors. Birthweight may be an intermediate factor in a causal path between smoking and pelvic pain if smoking women have smaller babies, leading to less tissue damage and thus less pelvic pain. Consequently, we did not include birthweight as a confounding factor. In fact, we found that a high birthweight was associated with pelvic pain (Table 1), but as we do not know the causal mechanism, birthweight may act as a confounding factor. In a supplementary analysis we adjusted for birthweight, which only marginally changed the risk estimates. The possible causal pathway behind smoking and pregnancy-related pelvic pain is left to speculation, as is the relation between smoking and musculoskeletal diseases in general, 15 but a reduced blood flow to tissue surrounding the pelvic girdle joints has been suggested ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology

7 Smoking and pregnancy-related pelvic pain Conclusion Our findings indicate that smoking is a risk factor for pregnancy-related pelvic pain, and adds another potential health consequence to the long list of complications related to smoking. Smoking before pregnancy may be as important as smoking during pregnancy, but these findings are new and need to be replicated. Disclosure of interests None. Contribution to authorship KB conceived the study area, performed the statistical analysis and wrote the first draft of the paper. EAN took part in planning the study, in the statistical analysis and co-wrote the paper. JO planned the initial DNBC data collection, took part in data analysis and co-wrote the paper. NH took part in planning the initial DNBC data collection and co-wrote the paper. AMNA and MJ initiated, planned and designed the initial case control study, and co-wrote the paper. Details of ethics approval The DNBC was approved under no. (KF) /94 of the regional scientific ethical committee for the municipalities of Copenhagen and Frederiksberg, Denmark. Funding The Danish National Research Foundation has established the Danish Epidemiology Science Centre that initiated and created the DNBC. The cohort is furthermore a result of a major grant from this foundation. Additional support for the DNBC was obtained from the Pharmacy Foundation, the Egmont Foundation, the March of Dimes Birth Defects Foundation, the Augustinus Foundation, and the Health Foundation. This specific study was supported by grants from The Health Insurance Foundation in Denmark. The Health Insurance Foundation supported the study financially and had no role in any other parts of the study. j References 1 Hansen A, Jensen DV, Wormslev M, Minck H, Johansen S, Larsen EC, et al. Symptom-giving pelvic girdle relaxation in pregnancy. II: Symptoms and clinical signs. Acta Obstet Gynecol Scand 1999;78: Aslan E, Fynes M. Symphysial pelvic dysfunction. Curr Opin Obstet Gynecol 2007;19: Larsen EC, Wilken-Jensen C, Hansen A, Jensen DV, Johansen S, Minck H, et al. Symptom-giving pelvic girdle relaxation in pregnancy. I: Prevalence and risk factors. Acta Obstet Gynecol Scand 1999; 78: Wu WH, Meijer OG, Uegaki K, Mens JM, van Dieen JH, Wuisman PI, et al. Pregnancy-related pelvic girdle pain (PPP), I: Terminology, clinical presentation, and prevalence. Eur Spine J 2004;13: Tophoj A. Pregnancy and sick leave. Ugeskr Laeg 1994;156: Rasmussen OB, Sorensen AU, Nielsen CV. Sick leave for pregnant women in the municipality of Fjend Ugeskr Laeger 1996;158: Juhl M, Andersen PK, Olsen J, Andersen AM. Psychosocial and physical work environment, and risk of pelvic pain in pregnancy. A study within the Danish national birth cohort. J Epidemiol Community Health 2005;59: Wang SM, Dezinno P, Maranets I, Berman MR, Caldwell-Andrews AA, Kain ZN. Low back pain during pregnancy: prevalence, risk factors, and outcomes. Obstet Gynecol 2004;104: Albert HB, Godskesen M, Korsholm L, Westergaard JG. Risk factors in developing pregnancy-related pelvic girdle pain. Acta Obstet Gynecol Scand 2006;85: Kauppila LI, Penttila A. Postmortem angiographic study of degenerative vascular changes in arteries supplying the cervicobrachial region. Ann Rheum Dis 1994;53: Kauppila LI, Penttila A, Karhunen PJ, Lalu K, Hannikainen P. Lumbar disc degeneration and atherosclerosis of the abdominal aorta. Spine 1994;19: Leboeuf-Yde C. Smoking and low back pain. A systematic literature review of 41 journal articles reporting 47 epidemiologic studies. Spine 1999;24: Heliovaara M, Makela M, Aromaa A, Impivaara O, Knekt P, Reunanen A. Low back pain and subsequent cardiovascular mortality. Spine 1995;20: Mikkonen P, Leino-Arjas P, Remes J, Zitting P, Taimela S, Karppinen J. Is smoking a risk factor for low back pain in adolescents? A prospective cohort study. Spine 2008;33: McBeth J, Jones K. Epidemiology of chronic musculoskeletal pain. Best Pract Res Clin Rheumatol. 2007;21: Wisborg K, Henriksen TB, Hedegaard M, Secher NJ. Smoking during pregnancy and preterm birth. Br J Obstet Gynaecol 1996;103: Wisborg K, Henriksen TB, Obel C, Skajaa E, Ostergaard JR. Smoking during pregnancy and hospitalization of the child. Pediatrics 1999;104:e Wisborg K, Kesmodel U, Henriksen TB, Olsen SF, Secher NJ. Exposure to tobacco smoke in utero and the risk of stillbirth and death in the first year of life. Am J Epidemiol 2001;154: Cnattingius S. The epidemiology of smoking during pregnancy: smoking prevalence, maternal characteristics, and pregnancy outcomes. Nicotine Tob Res 2004;6(Suppl 2):S Egebjerg Jensen K, Jensen A, Nohr B, Kruger Kjaer S. Do pregnant women still smoke? A study of smoking patterns among 261,029 primiparous women in Denmark Acta Obstet Gynecol Scand 2008;87: Wisborg K, Henriksen TB, Hedegaard M, Secher NJ. Smoking habits among Danish pregnant women from 1989 to 1996 in relation to sociodemographic and lifestyle factors. Acta Obstet Gynecol Scand 1998;77: Olsen J, Melbye M, Olsen SF, Sorensen TI, Aaby P, Andersen AM, et al. The Danish National Birth Cohort its background, structure and aim. Scand J Public Health 2001;29: Greenland S. Quantifying biases in causal models: classical confounding vs collider-stratification bias. Epidemiology 2003;14: Albert H, Godskesen M, Westergaard J. Prognosis in four syndromes of pregnancy-related pelvic pain. Acta Obstet Gynecol Scand 2001;80: ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology 1025

8 Biering et al. 25 Gutke A, Ostgaard HC, Oberg B. Pelvic girdle pain and lumbar pain in pregnancy: a cohort study of the consequences in terms of health and functioning. Spine 2006;31:E England LJ, Grauman A, Qian C, Wilkins DG, Schisterman EF, Yu KF, et al. Misclassification of maternal smoking status and its effects on an epidemiologic study of pregnancy outcomes. Nicotine Tob Res 2007;9: Windsor RA, Boyd NR, Orleans CT. A meta-evaluation of smoking cessation intervention research among pregnant women: improving the science and art. Health Educ Res 1998;13: Nohr EA, Frydenberg M, Henriksen TB, Olsen J. Does low participation in cohort studies induce bias? Epidemiology 2006;17: ª 2010 The Authors Journal compilation ª RCOG 2010 BJOG An International Journal of Obstetrics and Gynaecology