Intensified PA counseling supported with an option for monthly thematic meetings with group exercise proved feasible among pregnant women at risk for gestational diabetes and was able to reduce the decrease in their weekly frequency of total and moderate-to-vigorous- intensity LTPA until the end of the second trimester better than conventional counseling.
To the best of our knowledge this is the first study to report on the safety of PA counseling integrated into routine maternity care practices and involving pregnant women at risk for gestational diabetes. The findings indicated no more miscarriages or adverse events listed as warning signs to terminate exercise during pregnancy
[3, 4] in INT than in UC. Neither was there any difference in neonatal safety issues, which have been reported in more detail elsewhere
. The results are in line with studies reporting no harmful effects of moderate-intensity PA on pregnancy outcomes
[13, 14, 47]. However, outside the list of warning signs, INT experienced throughout the pregnancy more musculoskeletal symptoms and painless contractions during or immediately after PA than UC. It is unlikely that the more frequent occurrences of these symptoms were exclusively due to the intervention because there was no between-group difference in change in the duration of total or moderate-intensity LTPA at either of the follow-ups and the only difference between the groups in the frequency of total and moderate-intensity LTPA was discovered at 26-28 weeks’ gestation. It is possible that some of the symptoms remained undetected in UC because the response rate to the questions related to the adverse events was lower in UC (on average 62.8%) than in INT (on average 86.0%). Nevertheless, the more profound reasons for the higher occurrence of musculoskeletal symptoms and painless contractions in INT as well as for the higher occurrence of dyspnea and painful contractions in UC remain unclear.
The high realization of counseling in terms of timing, duration and compliance shows that the intervention was carried out as intended, which improves the likelihood that the results are indeed due to the intervention. Assessing realization is essential especially in interventions delivered by a third party, implemented in a real world setting less controllable than laboratory surroundings and requiring a significant contribution from the deliverers to carry out the actions. To date no other studies have been reported on the realization of PA counseling in this specific target group.
Attendance at monthly thematic meetings on PA was disappointingly low since only a third of the participants attended at least three of the five meetings. A higher participation rate was expected because the meetings were arranged only once a month and considerable effort was made with SMS reminders and telephone feedback to encourage the women to attend the meetings. However, the percentage is similar to our pilot study, where 28% of the first time pregnant women participated in at least half of the group-exercise sessions arranged once a week during pregnancy
. Undoubtedly one reason for the low participation in the present study was the fact that many of the participants already had children making it difficult to find time for extra activities. In the study by Hui et al.
 two thirds of the women attended at least three exercise sessions during pregnancy but more detailed information was not provided to enable comparison with our participation rate. To conclude, the attendance at group-exercise sessions offered in addition to maternal care was quite low. This could be improved after gaining more information about the reasons for non-participation and tailoring the exercise sessions accordingly.
The nurses’ views on the applicability of the counseling protocol were generally positive. The findings are similar to those of our pilot study
 indicating that this kind of counseling on PA may be transferable to routine maternity care visits. The nurses’ views on applicability may also reflect their satisfaction at having a systematic tool for PA promotion. In a recent study by Stotland et al.
 concerning counseling approaches in preventing excessive weight gain in prenatal care, health care providers felt unsure about the effectiveness of their counseling efforts and reported lack of training and tools for implementing counseling. Providers have also indicated a need for more information about the benefits and risks of PA during pregnancy
. According to our pilot study similar issues seem to apply to Finnish maternity care, where the counseling practices at baseline were surveyed (unpublished data). However, with the struggle of continuously diminishing resources, there may also be a need in maternity care for simpler and lighter approaches than the one examined in this study such as PA prescription, step-log monitoring, web-based programs or peer-support systems.
Regarding LTPA during pregnancy the findings concur with those of earlier studies indicating that women’s PA decreases during the course of pregnancy and tends to shift from moderate to lighter activities as the pregnancy proceeds
[5, 8]. The present study indicates that intensified counseling with an option for a monthly thematic meeting on PA can reduce the decrease in the weekly frequency of total and moderate-intensity LTPA compared to the conventional counseling until the end of the second trimester at 26-28 weeks’ gestation.
The finding is similar to that of our pilot study
 with the exception that in the pilot study, where the follow-ups were at 16-18 and 37 weeks’ gestation, the statistical between-group difference was only discovered at 37 weeks’ gestation. However, the pilot study involved only first-time pregnant women not screened for specific health risks, whereas the present study also included multiparous women and exclusively women at risk for gestational diabetes. Therefore the women in the present study were slightly older, heavier and less physically active, presumably limiting their LTPA especially during the last trimester more than in the pilot study.
Nevertheless, in this study, the between-group difference in change at 26-28 weeks’ gestation was less than one day per week, 0.70 days in total LTPA and 0.56 days in moderate-to-vigorous-intensity LTPA, which seems quite modest from the prevention of gestational diabetes point of view especially since it was not sustained until the last trimester. Moreover, no effects were discovered in the weekly duration of total or intensity-specific weekly LTPA, which may be more crucial from the health perspective than the weekly frequency of LTPA. In this respect it may be that the changes in PA due to intervention had only limited potential to affect the incidence of gestational diabetes and the birthweight of the newborns, which were the two primary health outcomes of the whole lifestyle intervention
. However, the more precise timing as well as the amount and intensity of PA sufficient to prevent gestational diabetes need still to be determined in future studies.
Strengths and limitations
The most important strength of the study was that the counseling was integrated into real maternity visits and implemented by the providers themselves. Together with feasibility evaluation this increases the pragmatic value of the findings and improves their transferability to practical maternity work. Counseling was also based on a behaviorally grounded model and the nurses were carefully trained and supported for implementation. In the analysis stage, multilevel models were used to reduce the bias related to intra-level variation within clusters, municipalities, nurses and individual participants.
The study nevertheless has some weaknesses to be taken into consideration when interpreting the results. Firstly, the representativeness of the study sample may have been impaired since there were fewer women whose BMI exceeded 25 and who were smokers in the final sample (N = 399) than among all eligible women (N = 726). The representativeness may also have been slightly although not notably hampered through the 29 dropouts, who had statistically lower education than the women in the final sample.
Secondly, the comparability of INT and UC may have suffered from the non-blinded allocation procedure, meaning that the participants were informed about the study group of their maternity clinic in the consent form. As shown in Table
4, the participants’ LTPA level at baseline was generally higher in UC than INT. This may have resulted from the higher refusal rate of women with low LTPA level in UC since the study offered them “no extra benefit”. In INT the situation may have been the opposite.
Thirdly, the power of the study had been calculated for the incidence of gestational diabetes, not for change in PA behavior. According to the calculations performed after the study the sample size was sufficient to discover the between-group difference of 40 min in the weekly duration of moderate-to-vigorous-intensity LTPA (intra-cluster correlation of 0.01, standard deviation of 125, significance level of 0.05 and power of 80%). Then, ideally, the number of participants in each municipality should have been 14. As the number of participants per municipality within the analyzed sample varied from 9 to 59 and the number of municipalities was quite minimal for multilevel models, this may have caused uneven weighing of data in the analysis.
Fourthly, bias may have resulted from using self-report as an outcome measure in PA assessment. It is possible that women in INT, being aware of the expectations related to their PA behavior, were more likely than women in UC to over-report their LTPA at the follow-ups. On the other hand, keeping a record of the realization of the action plans may have helped the participants in INT to recall their LTPA more accurately than their peers in UC. The use of RPE may also have improved the ability of the women in INT to classify the intensity of their LTPA in the follow-up questionnaires, which may have reduced the possible over-reporting of moderate-to-vigorous-intensity LTPA. In addition, recall errors may have occurred in assessing the LTPA prior to pregnancy since it was elicited at 8-12 weeks’ gestation. However, the possible recall errors at baseline were expected to be the same in INT and UC and should not have affected to group comparisons.
It can also be argued that the clinical significance of the findings concerning the weekly frequencies of total and intensity-specific LTPA should have been verified by complementing it with accurate information on between-group differences in the weekly duration of LTPA. However, the sensitivity of the self-report for detecting the differences was impaired in the sample of this size due to the wide variations in the weekly minutes of total and intensity-specific LTPA. In other words, the large individual differences in duration outcomes may have obscured the possible intervention effects. The assumption is, indeed, supported by the study by Aittasalo et al.
, where larger random errors in duration than frequency estimates was discovered with regard to this particular questionnaire. In future studies, using more objective measures such as pedometers or accelerometers could diminish this deficit. Pedometer seems feasible among pregnant women
 although the compliance may be lower in obese women
. In pregnant women the pedometer has also been shown to correlate with the exercise diary
. Studies on the feasibility and validity of using an accelerometer during pregnancy have not yet been published although the accelerometer has been used in pregnant women for validating PA self-reports
 and assessing PA