Pelvic girdle pain (PGP) is common in pregnancy. Recent studies have shown that about 33-50% of pregnant women report PGP before 20 weeks of gestation, and that the prevalence may reach 60-70% in late pregnancy [1–3]. Despite these high prevalence estimates, we have little knowledge about the risk factors for PGP in pregnancy. Previous studies have reported that strenuous work, a pre-pregnancy history of low back pain (LBP), previous PGP, and multipara are associated with PGP in pregnancy [4–9]. Associations between PGP and psychological variables such as catastrophizing, fear-avoidance beliefs and distress have also been reported [8, 10]. However, the number of studies are limited and often hampered by either being retrospective or cross-sectional [7, 8], or by lack of multivariable analyses in the prospective studies [4–6].

Moreover, the response variables used in previous studies have most often been dichotomous, such as presence of PGP or not, and did not necessarily reflect the severity of the condition. The importance of also using graded scales has recently been pointed out by Croft [11]. In a recent cross-sectional study we used graded scales and showed that women with combined symphysis pain and bilateral posterior pelvic pain in late pregnancy reported more disability than women with fewer pain sites in the pelvis [3]. Others have shown that women with this combination of pain locations were also less likely to recover postpartum than those with more limited pain distribution [12, 13].

Clinical management would probably benefit from an early identification of women at risk for developing disabling symptoms later in pregnancy. A number of tests for pain provocation of different tissues and locations in the pelvis are commonly used and recommended [14]. Although both pain provocation tests and functional tests have most often been used for diagnostic purposes [13, 15–17], they might also detect processes at an early stage. Previous studies of PGP during and after pregnancy have reported that positive scores on the posterior pelvic pain provocation (P4) test and the functional Active Straight Leg Raise (ASLR) test were associated with disability [3, 18, 19] and pain [16, 18, 19]. Furthermore, when blinded assessors were used, relative high frequencies of positive responses to the tests were also reported for pregnant women without pain in the pelvic area [3]. These results could either indicate low specificity or alternatively that the tests could detect subclinical afflictions and thus be valuable in early identification of those at risk for more severe afflictions.

We established a cohort of pregnant women to study the associations between socio-demographical, psychological and clinical factors measured at inclusion in early pregnancy and disability or pain intensity in gestation week 30.

This is a prospective cohort study following pregnant women in Norway from early pregnancy to gestation week 30.

Procedure

The Norwegian public health system offers all women free health services during pregnancy and most women seek special maternity care units (MCUs) for this purpose. We collaborated with four public MCUs in this study, one was located in central Oslo (capital, about 580 000 inhabitants), and the other three covered one entire community (about 24 000 inhabitants) just outside Oslo. Eligible participants were Norwegian-speaking women, who registered at these four MCUs between January 2006 and June 2007. Women not expected to have a normal pregnancy (as determined by the midwives) were excluded. Out of 385 eligible women, 326 gave their informed consent for participation. Out of these 326 women, 280 were included before they reached 20 weeks of gestation, and were thus defined as being in early pregnancy (figure 1). From the time of inclusion to gestation week 30, there were 3 drop-outs and 9 miscarriages among the 280 women included early, thus 268 women participated in gestation week 30 and these constituted our study sample.

After inclusion all answered a comprehensive questionnaire assessing sociodemographic variables, pain locations, pain intensity and disability, distress, and fear-avoidance beliefs. The questionnaire also included questions on general health, health-related quality of life, health locus of control, use of contraceptives, other complaints, and physical activities, variables that were not used in this part of the study. The registered gestation week refers to the week the women were included to the study and completed the questionnaire.

All women were clinically examined in early pregnancy by one of two physiotherapists with post-gradual education in manual therapy. This examination was performed as closely to the inclusion date as possible. Mean time difference between answering the questionnaire and being examined was 1.1 week (SD 1.7 weeks). The clinical examination included six pain provocation tests for the pelvic joints as well as the functional ASLR test and Beighton score for hyper mobility. Other clinical tests were also included, but were not used in this part of the study. The examiner was blinded for all questionnaire data. In gestation week 30, the women filled in a new questionnaire assessing the same elements as at inclusion and underwent a corresponding clinical examination. Data from the clinical examination in gestation week 30 was not used in this part of the study. The Regional Committee for Medical Research Ethics and the Norwegian Social Science Data Services gave formal approval for the study.

Pre-pregnancy BMI, smoking, physical activity before pregnancy, full time work and Beighton score for hypermobility were not significantly associated with DRI in gestation week 30 in the bivariate analysis (0.16 ≤ p ≤ 0.64). Physical activity before pregnancy, full time work and Beighton score for hypermobility were not significantly associated with pain intensity in gestation week 30 in the bivariate analysis (0.38 ≤ p ≤ 0.98). These variables were not entered in to the respective multivariable models. Age, gestation week, pre-pregnancy LBP, and work condition were not significantly associated with the response variables (0.11 ≤ p ≤ 0.65), but were entered into the multivariate models based on associations reported in previous studies [4–9].

The effect estimates of each response variable were relatively large in both models, although the 95% confidence intervals were wide. Yet the effect estimates seemed to be higher for pain intensity compared with DRI. For instance our data shows that pain intensity in late pregnancy is 40.4 (95% CI: 24.4, 56.6) higher when pain was present in the symphysis only, compared with having no pain in early pregnancy and adjusted for P4 test and sum of pain provocation tests (table 5, model 1).

The main results from this study were that pain locations in the pelvis, positive P4 test and sum of pain provocation tests in early pregnancy were significantly associated with disability and pain intensity in late pregnancy. In addition, distress was significantly associated with disability. The functional test ASLR, fear avoidance beliefs and the number of pain sites were not significantly associated with neither disability nor pain intensity.

The risk factors identified in this study differ from those that have been reported before. Strenuous work, pre-pregnancy history of LBP and parity have previously been identified as risk factors for PGP in studies applying bivariate statistics [4–6] and multivariable models [9]. In our bivariate correlation analyses, the first two variables were not significantly associated to neither disability nor pain intensity in gestation week 30, while parity was. None of the variables were significant in the multivariable analyses. This could be due to difference in design or to the use of different levels of statistical methods. One possible explanation for the difference could be that previous studies have often recorded the risk factors retrospectively, late in pregnancy and after the onset of symptoms. Hence, the women's reporting of these factors might be biased by pain [7, 8]. The prospective design of the present study ensured that this possible bias was avoided. At the time of inclusion and measurement of the risk factors, none of the women had defined their symptoms as a problem and they were not seeking treatment.

It is also noteworthy that the results of the functional ASLR test measured in early pregnancy, used with a distinction between those with strong affliction and those with none or lesser affliction, was not significantly associated with disability. This might indicate that severe impairment of motor control and movement of the legs relative to the pelvis was not important for the development of PGP. On the other hand, the response to the P4 test was identified as a risk factor for both pain intensity and disability. Since this test is supposed to elicit a distinct located pain deep in the gluteal area [17], it seems that affliction in the posterior pelvis has an impact on the course. This is, however, partly contradicted by the data from pain locations. Self-reported pain only in the symphysis in early pregnancy had about the same impact on disability and pain intensity in gestation week 30 as did combined symphysis pain and posterior pain. Moreover posterior pain (without symphysis pain) in early pregnancy was not significantly associated with disability and pain intensity in gestation week 30. Since this group was the largest, the lack of effect can hardly be explained by lower test power than the other pain locations. Although the confidence intervals were wide, our data indicate that subclinical afflictions in both anterior and posterior part of the pelvis are of importance for development of pain and disability. Hence, our data suggest that symphysis pain can be an early indicator or precursor for pain development in other areas of the pelvis. Interestingly, the association seems to disappear when pain location and disability are measured simultaneously in late pregnancy [3].

When we included disability or pain intensity assessed in early pregnancy in the multivariable models, some explanatory variables were no longer significant. This means that these variables were not risk factors for the change in disability or pain intensity. However, from a clinical point of view it is more important to identify risk factors for disability and pain intensity late in pregnancy than the change from early pregnancy. This is supported by the data showing an increased DRI already in early pregnancy compared with healthy non-pregnant women [20].

Several of the previously identified risk factors for PGP in pregnancy are similar to those reported for LBP and for other musculoskeletal disorders and are not specific for PGP [36, 37]. These comprise socio-demographical factors, previous history of LBP, strenuous work and high level of distress. In contrast, positive response to the P4 test has been shown to be sensitive and specific for PGP [17]. Also the pattern of pain locations within the pelvis is probably specific for PGP, and one might therefore hypothesize that both the P4 test and pain locations are "condition specific" risk factors for PGP.

The response variables used in this study were measured as scale values whereas previous studies have used dichotomous responses for example reporting PGP or not. From the large variation in responses shown when using scales in the present study, one might question to what extent the dichotomous response variables actually reflects important affliction. The dichotomous response variables have resulted in very high prevalence rates for PGP in pregnancy [1, 2, 10, 25]. We have recently found that the variability in DRI was large both for women reporting and not reporting PGP [3]. In order to capture associations to this large range of affliction, the used scales seem to provide additional information than the dichotomous responses.

Previous studies have shown associations between distress, fear avoidance beliefs and activity limitations in patients with LBP [38–42], and also that distress contributed to physical activity and work loss in an acute sample of LBP patients [36]. Our results showed that distress contributed into the model for disability but not for pain intensity. Interestingly the effect of HSCL-25 on disability in gestation week 30 disappeared when we controlled for disability at inclusion. As in the study from Grotle and co-workers of acute LBP [43], fear avoidance beliefs was not identified as a risk factor for either disability or pain intensity.

Over the years, there has been a growing evidence for predictive effect of widespread pain on long term changes in work disability [44]. Furthermore, it has also been reported that the risk of long-term work disability was lower for persons with localized LBP compared with persons with LBP combined with pain in other bodily areas. The risk for long-term work disability increased with the latter [45, 46]. We included number of pain sites (excluding low back and pelvic area) in the multivariable analyses, and found that it did not contribute in any of the models. The lack of effects may be due to the small number of possible pain sites. However, it is also possible that PGP in pregnancy is a specific condition characterized by a rather short course compared with other musculoskeletal pain conditions. Most of the women recover shortly after delivery. One might thus speculate that multiple pain sites are not of importance for development of PGP in pregnancy, but could still be of importance for non-recovery from PGP postpartum.

The present study has several strengths, including the use of a prospective design, continuous response variables and multivariable statistics. Furthermore the implementation of clinical risk factors, use of blinded examiners and the follow-up of all pregnant women in the cohort independent of having PGP or not also strengthen the study.

A limitation that should be considered when interpreting the results is the limited numbers of women in some of the groups. However, even though the confidence intervals are wide, the findings indicate that the risk factors are of importance. On the other hand, lack of significant results should be interpreted with caution.

Another possible weakness could be the representativeness. The women participating in the cohort were about the same age and in the same gestation week as women declining participation. Women who were excluded from analyses due to late inclusion were also about the same age. The average age of women giving birth in Norway have been 30.3 years (2006 - 2007) [47] i.e., almost similar as in our cohort. There were some differences in the percentage of nulliparous women in the non-participant group, the excluded group and the participants (44%, 77% and 59% respectively). The number of nulliparous women in the cohort was also slightly higher than among Norwegian women (59% vs 42%). We cannot exclude the possibility that another cohort of pregnant women in Norway, would result in somewhat different results with regard to prevalence of pain locations and positive clinical tests. However, the associations between them are expected to be similar.

In conclusion, we have found that pain locations in the pelvis, bilateral positive P4 test, and sum of pain provocation tests in early pregnancy were significantly associated with disability and pain intensity in gestation week 30. The effect estimates were relatively large. Furthermore distress was significantly associated with disability, but not with pain intensity. Fear avoidance beliefs were not significantly associated with any of the responses. These results thus suggest that a clinical examination including a few tests performed in early pregnancy may identify women at risk of a more severe PGP late in pregnancy. The identification of clinical risk factors may provide a foundation for development of targeted prevention strategies.

HSR (RPT and MSC) is doctoral student and manual therapist, MBV (PhD) is associated professor in biostatistics, AMM (RPT and PhD) is professor, NKV (PhD) is professor and Head of institute