The current study is an extension of a previous questionnaire study investigating low back and pelvic girdle pain during pregnancy. The initial recruitment procedure has been reported elsewhere in detail 171819. Data were gathered between January 1, 2002 through April 30, 2002 from Umeå University Hospital (UUH) and Sunderby Hospital (SH), both located in the northern part of Sweden. The initial data set for this study consisted of 891 women. Inclusion criteria for participation were delivery after 23 gestational weeks and proficiency in the Swedish language. All women participating in the initial study were contacted through a letter sent by mail with information of the extension of the study. It was explained in the letter that they could withdraw from further participation by returning a prepaid envelope with a note saying that they declined further participation. After the informed consent of the participants, comprehensive information on the outcomes of pregnancy and childbirth was collected from medical records and added to the previously collected questionnaire data. The selection of the final sample is presented in Figure 1. The participants in this study consisted of 822 women, constituting 73.8% of the total eligible population of 1114 women who had delivered within the two hospitals during the same period.

The questionnaire covered different aspects of gynaecologic and obstetric history of the participating women as well as questions concerning life style before and during pregnancy. Parts of these data have been presented in relation to low back and pelvic girdle pain 171819. From each medical record, 75 items presenting maternal characteristics and pregnancy outcomes were collected and added to the questionnaire data. These items included the presence of risk factors according to the local guidelines of GDM screening, medical outcomes of pregnancy and birth, and medical diagnosis using International Classification of Diseases (ICD) version 10.

The local guidelines obtained for both catchment areas, addresses the risk factors which indicates screening for GDM. Compilation of local guidelines for each catchment area is presented in Table 1. The majority of the criteria were identical in the catchment areas. According to the guidelines, the OGTT should be performed in early pregnancy in women with prior GDM and repeated at 28 gestational weeks if the first OGTT were normal. Other risk factors indicated that OGTT should be performed at 28 gestational weeks. Blood glucose levels were measured independent of last meal of the pregnant woman at 4 times regularly during pregnancy in the SH catchment area and at 5-6 times regularly during pregnancy in the UUH area. When a high blood glucose value would be detected, the OGTT should be performed immediately independent of gestational age.

Some available data from the questionnaires allowed comparisons between participants and non-participants. The non-participants (n = 69) did not differ significantly from participants (n = 822) concerning maternal age, parity, educational level, gestational age at birth, mode of delivery, or birth experience. A comprehensive description of non-participants in the initial data collection has been reported elsewhere 18.

To compare the characteristics and outcomes of women with and without risk factors of GDM, the women were categorized into groups based on presence of risk factors as defined in the local guidelines (Figure 2). Firstly, all women were dichotomized with respect the presence or absence of GDM risk factors in their medical history. Secondly, within these two groups, all women were dichotomized with respect to developing GDM risk factors or not during pregnancy. Based on these categorizations, three groups of women with risk factors were identified.

1. Risk factor group1, R1: women with risk factors in medical history and additional risk factors developing during pregnancy.

2. Risk factor group 2, R2: women with risk factors in their medical history.

3. Risk factor group 3, R3: women with risk factors developing during pregnancy.

A fourth group of women with absence of any risk factors in their medical history or during pregnancy (normal group, NG) was also identified.

Statistical analyses were performed using the statistical software package (SPSS, version 16.0). For analyses of categorical variables, the Chi square-test was performed. For small samples, the Fischer's exact test was applied. For normally distributed continuous variables, the Student's t-test and analysis of variance [ANOVA] were performed. The Bonferroni procedure was used to control for multiple testing. For variables with a skewed distribution, the Kruskal-Wallis and Mann-Whitney tests were applied. The level of significance was set at p = 0.05.

Univariate and stepwise multiple logistic regression analyses were used to investigate the association between exposures before and during pregnancy and outcomes of pregnancy and birth. The analyses of risk factors were conducted in two steps. All exposure variables were tested one by one in separate, univariate analysis. In a second step, all statistically significant variables in the univariate analysis were tested using multivariate logistic regression analysis. Significant variables were entered in a stepwise manner. Results from the final model are presented as odds ratios (ORs) with confidence interval (CI) 95%.

Three models were tested in the analysis. Model 1 consisted of the group of women fulfilling the criteria of OGTT being correctly exposed to the test and the normal group of women with absence of GDM risk factors in their medical history and during pregnancy. Model 2 was based on the group of women fulfilling the criteria of OGTT and not being exposed to the test, and the normal group of women with the absence of GDM risk factors in their medical history and during pregnancy. Finally, the third model consisted of the women who fulfilled the criteria, but were not exposed to the test and the women fulfilling the criteria for OGTT and correctly exposed to the test.

The internal and external validity of the questionnaire has been discussed elsewhere. Twenty-five women completed an identical questionnaire two to three weeks after the first questionnaire and the consistency between the two sets was high, with total agreement in majority of questions 171819.

To test the internal validity of data from the medical records, a control of 50 randomly selected medical journals was performed. Of the 3 750 items controlled, 5 items (0.1%) of incorrect data excerption were found and corrected.

At the first antenatal visit, at least one risk factor for GDM in the medical history as stipulated in the local guidelines was reported by 170 (20.7%) women. There were no significant differences in prevalence of risk factors in medical history between the two catchment areas. Among the women with risk factors in their medical history diabetes mellitus in family was the most frequent risk factor (61%), followed by maternal body weight of 90 kg or more and/or BMI ≥33 (38%), previous infant with birth weight of 4500 g or more (14%) and previous pregnancy with GDM (1%). As presented in Additional file 1, women with risk factors in their medical history reported more problems related to life style and had lower educational level. Most women were married or cohabiting with their partner. The women with risk factors in their medical history described their work as being less intellectually stimulating (71.3% vs. 81.5%, p = 0.034). No differences were found regarding other aspects of the work such as the work being sedentary or psychologically demanding. Furthermore, there were no significant differences concerning the evaluation of the relationship with spouse/partner and satisfaction with sex-life.

Risk factors developing during pregnancy were reported in 132 pregnancies (15.9%). Additional file 1 presents maternal characteristics of pregnant women with risk factors for GDM in their medical history and/or developing during pregnancy and pregnant women with no risk factors for GDM in their medical history. Of the 132 women, 45 women (33.6%) already had at least one risk factor for GDM in their medical history. The most frequent risk factor developed during pregnancy was accelerated foetal growth (59%), followed by random high blood glucose value (42%) and polyhydramniosis (1%). The women developing risk factors during pregnancy were significantly heavier, 74.6 (± 16.9) kg vs. 65.5 (± 11.2) kg, p < 0.001 and correspondingly presenting a higher BMI 27.8 (± 5.9) vs. 24.3 (± 4.0), p < 0.001 compared to women not developing risk factors of GDM during pregnancy. No other significant differences were observed between the women developing risk factors during pregnancy and those who did not.

The 257 women included in the categories R1, R2 and R3 (ie. 31.3% of the total sample) fulfilled criteria for performing OGTT as stipulated in the local clinical guidelines. However, only 84 women were actually exposed to OGTT according to their medical records (Additional file 1). Five of the 84 women had no indication of OGTT presented in their medical records (although performed), leaving 79 out of 257 (30.7%) women correctly exposed to OGTT. Seven women (8.8%) of the women correctly exposed of OGTT were diagnosed with GDM. Women in R1 underwent an OGTT significantly more often compared to the women in R2 or R3. Obesity was the most frequent risk factor (in the medical history) in combination with accelerated foetal growth during pregnancy in the R1. Family history of diabetes mellitus was the most prevalent risk factor for R2. Among the women in R3, accelerated foetal growth was the most prevalent risk factor. As presented in Additional file 1, the women included in the R1 were significantly more obese than the women included in the other risk factor groups, even when a body weight ≥90 kg was one of the risk factors.

The outcomes of pregnancy and birth in relation to risk factor groups are presented in Table 2. The mean birth weight was significantly higher in R1 and R3 as well as the mean placenta weight in relation to NG. The correlation coefficient between birth weight and placental weight (subjects with twin pregnancies and/or preterm birth were excluded from the analysis) was r² = 0.41, p < 0.001. Similar results for the four subgroups were r² (R1) = 0. 51, r² (R2) = 0. 42, r² (R3) = 0.39 and r²(NG) = 0.34. There were no significant differences between the subgroups concerning mode of delivery, preterm delivery, induction of labour, haemorrhage, Apgar score, pH of the umbilical cord blood, and proportion of children referred to neonatal intensive care unit. Nor were there any significant differences concerning jaundice or hypoglycaemia of the infant observed. Women developing risk factors during pregnancy (R1 and R3) had significantly more often been exposed to induced labour (19% vs. 12%, p = 0.037) compared to women with no risk factors during pregnancy (R2 and NG). The birth experience did not show significant differences between the subgroups, although women with risk factors in their medical history (R1 and R2) expressed significantly more negative birth experience scores; 7.6 (± 2.4) vs. 8.0 (± 2.0), p = 0.026 compared to women with no risk factors in their medical history (R3 and NG).

In univariate logistic regression analysis, delivery of macrosomic infant was significantly associated with the determinants: multiparity (≥3 children), education lower than university level, previous infant with macrosomia, maternal bodyweight ≥90 kg, accelerated foetal growth, being correctly exposed to OGTT and weight gain of 16 kg or more during pregnancy. In the stepwise multiple logistic regression model, the determinants previous infant with macrosomia (OR 17.77, 95% CI 1.38-229.15) and maternal body weight ≥90 kg at entrance of pregnancy (OR 17.80, 95% CI 1.06-299.40) remained significant.

R1 demonstrated an increased risk of a macrosomic infant (OR 5.65, 95% CI 2.64-12.11) as did R3 (OR 3.48, 95% CI 1.27-9.57) in relation to NG. Adjusting for multiparity (≥3 children), education less than university level, maternal body weigh ≥90 kg at entrance of pregnancy and weight gain of ≥16 kg, R1 had a substantially increased risk (OR 9.33, 95% CI 2.65-32.87) of giving birth to a macrosomic infant as did women in R3 (OR 12.96, 95% CI 1.33-126.63) compared to NG.

R2 and R3 had an increased risk of non-exposure to OGTT in relation to R1 (R2: OR 7.01, 95% CI 3.30-14.89 and R3: OR 8.49, 95% CI 3.75-19.21). As presented in Additional files 2 and 3, women fulfilling criteria of OGTT and correctly exposed to OGTT and the women fulfilling criteria but not exposed, gave birth to infants with macrosomia significantly more frequent compared to NG. In the stepwise multiple regression, the risk of an infant with macrosomia was more than 4-folded for the women correctly exposed to OGTT and more than 5-folded for the women not exposed to OGTT. As presented in Additional file 4, in the univariate analysis, women fulfilling criteria, however not exposed to OGTT presented decreased risk of obesity, diastolic blood pressure of ≥140 mm Hg, proteinuria and macrosomia. Further, a more negative birth experience was found among these women compared to women correctly exposed to OGTT. However, in the stepwise multiple regression analysis, only the outcome of a decreased risk of developing proteinuria during pregnancy remained significant. Adjusting for all risk factors of GDM (as defined in the local guidelines), women with the lowest education level had an increased risk (OR 6.19, 95% CI 1.58-24.27) of not being correctly exposed to OGTT when fulfilling the criteria for performing the test in relation to women with university education.

In Sweden, almost all deliveries take place in hospitals. Since less than 0.5/1000 are planned home births 31, the population-based design with data collection before discharge from hospital after childbirth accurately reflect the population. The validity of the data from the questionnaires has been extensively discussed elsewhere 171819. The current study combined data from medical records with self-reported results from a questionnaire. Furthermore, the data were collected within two separate regions in the northern part of Sweden.

A retrospective data collection procedure may induce recall-bias. However, the accuracy of perinatal information has been reported as very good even four to six years after childbirth 32. The data from the women were collected within the first days after delivery and birth in our study, which most probably will account for accurate memories and statements by the participants. However, the accuracy of midwives' and physicians' documentation may be questioned. The documentation in medical records by midwives and physicians show that midwives have higher accuracy of recorded medical conditions, pregnancy complications, intrapartum and postpartum events than physicians in a study in the USA 33. A validation of the excerpted data in the current study was performed showing that only a few items were incorrect.

In order to improve the compliance to the local guidelines, the OGTTs could preferably be performed within the midwifery service. Furthermore, there is a need of national consensus of the design of a screening program for GDM. Further studies should address the issue of compliance with local guidelines to verify whether these findings are general. Future studies may address midwives' attitudes toward obesity and women with low educational levels. Furthermore, investigation of whether interventions regarding diet and blood glucose levels may influence the prevalence of macrosomia among women developing risk factors during pregnancy should be conducted.