Standard treatment consisted of basic information about PGP and anatomy of the back and pelvis. The purpose of this information was to reduce fear, and to encourage patients to become active in their own treatment. The patients got a pelvic belt and a traditional home training programme including easy performed exercises for global muscles, with aim to increase strength in the abdominal, back, gluteal and shoulder muscles. Patients were told that they were free to call the physiotherapist at any time if they had questions about the home exercises or wanted more help i.e. needed crutches.

The same treatment as in the standard group, but in addition acupuncture. The aim with acupuncture was to establish control of PGP to prevent dysfunction of muscles of the spine and pelvis. The aim with the stimulation was to activate both the segmental pain inhibitory system, involving the so-called gate control mechanism 32 and the central pain inhibitory system, involving secretion of endogenous opioids 33. The number and selection of acupuncture points was based on clinical experience and expert knowledge of acupuncture in pregnant women with PGP. Ten classical acupuncture points were selected individually in the same segments as the location of PGP after diagnostic palpation to identify sensitive spots (Table 1). Two acupuncture points on the medial side of the leg and foot were selected in the same segment as the PGP and extra-segmental points to the lumbosacral area were used to strengthen and lengthen the effect of the central control systems.

The needles (Hegu; Hegu AB, Landsbro, Sweden) were disposable and made of stainless steel (∅ 0.30) and they were inserted intramuscularly to a depth of 15–70 mm. Needles were manipulated manually to evoke needle sensation (De Qi), a characteristic feeling of tension, local pain, soreness or numbness, and often a radiating feeling from the point of insertion. The needles were manually stimulated every ten minutes and left in situ for 30 minutes. Treatment was given twice a week.

The same treatment as in the standard group but, in addition, stabilising exercises and massage 34. The aim was to retain motor control by retaining a co-contraction pattern of the deep trunk muscles 35, to correct posture and avoid negative stress to pelvic structures in daily life 343637. The exercises which caused low load on the pelvic structures were given according to Richardson and Jull 34 but modified because of the pregnancy. All exercises were carefully instructed individually, but performed as home-exercises on a daily basis. Exercises started with isolated low force co-contractions in M. Transversus Abdominis and Mm Multifidii in static postures. These exercises were, followed by increasingly more difficult challenges as the patient's skills improved. The aim with the massage was to increase blood-flow in the gluteal muscles and hip extensors. It was given once or a few times by the physiotherapist, the patient and partner was thereafter instructed how to perform the massage at home.

The women receiving acupuncture and stabilising exercises were asked whether they had identified a side effect at each visit, although they were not presented with a list of possible side effects. However, before the first treatment the women in the acupuncture group were informed of possible side effects connected to acupuncture. I.e. they were informed that bleeding or bruising, pain on needling, and aggravation of symptoms occur in some patients and that they could experience drowsiness, especially after the first treatment. Patients were also advised to avoid strenuous physical activity the same day as the treatment. Immediate adverse events connected to acupuncture as hypotension/fainting and/or nausea were recorded. Fetal heart rate, maternal heart rate and blood pressure were monitored before and after all acupuncture sessions to assess the cardiovascular effect of acupuncture and influence on the fetal heart rate was measured with CTG in 43 women. CTG was recorded prior, during and after 43 treatment (30 minutes each time or until the trace was approved as being normal). The CTG traces were assessed both by the computer (Oxford 8000 computerised CTG analyser, Oxford Sonicaid Ltd, Oxford, England) and visually by an experienced obstetrician 31.

The women in the stabilising exercise group were informed that PGP could aggravate if they overloaded their pelvis and they were instructed to lower the intensity of the exercises if this happened.

At follow-up visit to the physiotherapist, (who was not informed of treatment allocation), all women filled out a questionnaire in which they answered a question if they had experienced harm of the treatment they had received (no or yes). If they marked yes, they were asked to list disadvantages and/or adverse events that they had experienced. The women were also asked to indicate their overall experience of the treatments. The options were: no help, some help, good help and very good help.

To assess influence of treatments on the pregnancy, delivery and condition of the child at birth, we recorded antenatal, intrapartum, neonatal and infant data that are normally registered in the Medical Birth Register (MBR) in Sweden. Assessment of the condition of the child at birth was based on weight, cord-artery and vein acid-base status (pH, and base deficit, mmol/L) and Apgar score. In Sweden preterm birth is defined as < 37 weeks gestation, metabolic acidosis is defined as cord artery pH ≤ 7.05 and BDef ≥ -12 and neonatal death is defined as death of a live born infant within 28 days of birth.

Data analyses were performed by a statistician who was blinded to group and treatment. Continuous data were tested for significance with Kruskal-Wallis test. If a significant difference was found between any of the three groups, Mann Whitney U test was used to compare one group against one another. Dichotomous data were tested for significance with Fischer's exact test. Adjustments (multiplication by three) of the p-values due to multiple comparisons were performed by Bonferroni's method when appropriate. An adjusted p-value < 0.05 was considered statistically significant. Calculation of statistical tests and descriptive evaluations were carried out using SAS software package, version V8.

The computerized or visually assessed CTG analyses of antenatal recordings in connection with acupuncture were all normal (Table 3).

There were no serious adverse events associated with either of the treatments. Minor adverse events were most common in the group receiving acupuncture (Table 4).

The acupuncture group and the stabilising exercise group were significantly more likely to be "very satisfied" with their treatment than the standard treatment group (Table 5).

No negative influence on the pregnancy outcomes were found after either of the treatments (Table 2). Acupuncture did not lead to an increased rate of preterm delivery, five women (4%) in the acupuncture group (two in week 34 and 3 in week36) delivered preterm. Frequency of pre-term delivery was similar in the two other treatment groups, seven women (5.4%) in the standard group (two in week 34 and 35 and three in week 36 respectively) and seven women (5%) in the stabilising exercise group (one in week 34, two in week 35 and four in week 36 respectively) delivered preterm. There were no differences between groups with regard to pregnancy complications or adverse clinical conditions (Table 6).

Table 7 shows that the usage of analgesia was similar in the three treatment groups.

Table 8 shows that there were no differences detected between the study groups in neonatal outcomes. Acupuncture did not lead to credible negative influences on the fetuses or neonates. There was one perinatal death related to meconium aspiration combined with infection and pneumothorax reported in the stabilising exercise group. There were four neonates in the acupuncture group, one neonate in the standard group and two neonates in the stabilising group with metabolic acidosis.

We are not surprised that adverse events were rare in the standard group as the prescribed exercises were of low-force. Experience of slight rubbing connected with use of the semi plastic belt have also been reported by others 3839404142.

Our result shows that although minor adverse events were more frequent in the acupuncture group than in the other study groups, they were sparse. Only 64 minor adverse events (5%) out of 1414 acupuncture treatments were reported. These minor adverse events were reported by 43 women showing that some women were more susceptible than others. Also, six out of ten of women who withdrew from acupuncture treatment reported one of these adverse events as reason for not completing their participation. The most frequent adverse events were pain at the site of needling, drowsiness after treatment and headache, which is in line with a postal survey of 9408 prospectively identified acupuncture patients 25. Our data of no serious adverse events in pregnant women during the treatment course of acupuncture are also in agreement with previous studies of PGP and/or LBP during pregnancy 10111214.

We believe that experience of needle pain (n = 27) must be considered unavoidable when treatment involves needling. In addition, the pain may have resulted from the achievement of de qi, a feeling thought to be essential for a successful treatment. Nevertheless, in spite of the minor adverse events, the women rated acupuncture favourably even though this and a majority of them were willing to use the same treatment in the future if needed, suggesting that the pain of the treatment did not negatively impact their overall experience of the treatment.

We think that fear of immediate and long-term serious adverse events of acupuncture in pregnant women have resulted in both a restricted use of acupuncture and a delayed start of this treatment in pregnant women. To reduce development of abnormal patterns of muscle activity 3 we decided to begin treatments as soon as possible after the first symptoms of PGP. As a consequence, 62 women (50%) in the acupuncture group started treatment as early as in gestational week 14 to 25.

In addition, an adequate dose of acupuncture probably is essential for the outcome of treatment and, as no scientific evidence of harm of acupuncture during pregnancy exists we decided not to reduce the stimulation because of the pregnancy. Therefore we used acupuncture points and the same stimulation that is normally used in a non pregnant population. Fortunately, this strategy resulted in significant pain relieving effects 13. It is not easy to find safety data of acupuncture given to women in the second and third trimester of pregnancy because these women are seldom offered acupuncture due to the fact that pregnancy sometimes is regarded as a relative contraindication to acupuncture (because of an old Chinese proposal of an increased risk of uterine contractions leading to abortion or pre-term delivery. There are only four other studies published (171 women in total) of acupuncture for PGP and/or LBP 10111214. Two of them (n = 71) registered Apgar score and infant weight. However these studies used weaker stimulation than in our study. Kvorning et al. 11 started treatment with only two segmental points, stimulated to achieve de-qi twice during the session which only lasted for about three minutes. The needles were than withdrawn and the patient was allowed to rest for ten minutes 11. Guerreiro da Silva et al. 12 used 8 to 12 sessions with 12 needles for 25 minutes. In addition, Smith et al. 17 who treated 583 pregnant women with acupuncture for hyperemesis during the first trimester 17 reported no influences on the pregnancy outcome. But as in our study the sample size in that study was only able to detect large differences in pregnancy outcomes. These authors declared that a sample size of 19,476 women would have been required to detect an increased from 6 to 7% in spontaneous abortions.

There were 22 minor adverse events reported by 22 women in the stabilising exercise group. The most frequent adverse event was pain in the lower back at the beginning of the treatment period (n = 17). Our results of no serious adverse events of physiotherapy are in agreement with earlier research 10114142434445464748495051. In addition, 115/131 (88%) women in our study started treatment after gestational week 18 and a prospective epidemiological study of 92 671 pregnant women have reported no association between exercise performed after 18 weeks of gestation and risk of miscarriage 52.

Our results of no signs of fetal distress in connection with acupuncture is supported by Neri et al. 53 who found no signs of fetal distress or changes in short-or long-term variability on computerized CTG before, during and after minimal and true acupuncture plus moxibustion for breech presentation in 15 pregnant women. In addition, two prior studies of acupuncture have showed beneficial effect on the blood flow velocity ratios in the uterine arteries, which may be beneficial also for the fetus. One study was performed in infertile women and the other in women with uncomplicated pregnancies. The first study 54 showed a significant decrease of the mean pulsatility index, PI, in the uterine arteries after eight electro-acupuncture treatments reduced resistance in the placental vascular bed. The other study 55 showed that insertion of acupuncture needles at LI-4 plus at SP-6 lead to a significantly decrease of the uterine S/D ratios from 2.45 ± 0.3 to 2.22 ± 0.2 after 30 minutes indicating a positive influence on uterine artery waveforms. Possible mechanisms behind these results are a decreased tonic activity in the sympathetic vasoconstrictor fibres to the uterus and involvement of central mechanisms with general inhibition of the sympathetic outflow, in accordance with previously observed acupuncture effects 54. However, the mechanism which leads to an alteration of blood flow velocity and vascular tonus after acupuncture treatments is not well understood and to our knowledge the clinical benefit of a moderate decrease of the uterine S/D ratios from 2.45 ± 0.3 to 2.22 ± 0.2 in uncomplicated pregnancies is questionable.

As described in the method section, we used the same antenatal, intrapartum, neonatal and infant data that has been complied on all births in Sweden by the Swedish MBR since 1973. This register offers a unique possibility to obtain reliable prevalence figures regarding obstetric and neonatal outcome. Unfortunately this register has no data of frequency of acupuncture during the antenatal period. If that had been the case it would have been easy to collect data on safety from all pregnant women in Sweden that had got acupuncture for PGP and/or LBP.

The three different treatments were not found to affect the use of pain relief in labour, parturition or neonatal outcome. Effects on labour or parturition or cord blood acid base/gas balance after acupuncture treatment for PGP have not been reported earlier. Analysis showed that four neonates in the acupuncture group and two neonates in the stabilising exercise group had metabolic acidosis in the arterial cord blood. However, only one of the neonates in the standard group, one neonate in the acupuncture group and two of the neonates in the stabilising exercise group had Apgar score ≤ 7 at 5 minutes indicating metabolic acidosis of any importance. One limitation of the study is that cord blood was analysed in only 70% of the neonates. We assume that the generalizability of our results is not biased by selection effects as the frequency of cord blood analysis did not differ from all deliveries at Sahlgrenska University hospital during the same period.

The frequency of induction of delivery, oxytocin augmentation, preterm delivery, and postmaternity were similar in the treatment groups. Nor did acupuncture lead to observable negative influences on the neonates in our study, yet, the study was not powered to fully exclude such influences. In Sweden, the frequency of neonatal death is 0.3% and the frequency of preterm delivery is 5.6%, thus a very large number of patients would have been required to fully investigate these issues.

Our results are in line with a previous published review 56 of acupuncture for labour pain, which showed no negative effects on the delivery or the newborns. However, the women in our study received treatments that was considered "strong" (17 needles during 12 sessions), which can not be comparable to acupuncture given only during labour. A secondary finding in our study was that the distribution between male and female fetuses. One previous epidemiological study has reported that male sex of the fetus was related to back pain symptoms during pregnancy 57 However, this was not seen in our study, of the 383 neonates 48.7% were female and 51.3% were male as expected.

When trying to determine the clinical implications of the present study we have to consider that

1. The data emerged from a relatively large randomised trial, although we are aware that an efficacy randomised controlled trial is not an appropriate design to provide evidence of safety. Studies of more than 97 000 participants' have been used in the past 26.

2. A substantial number of women started treatment already during the second trimester of pregnancy.

3. The acupuncture may be considered to be strong compared to earlier studies. We used 17 needles, 10 of them were located in the lumbosacral region (i.e. somatic segments according to the innervation of the uterus that is, Th 12-L2, S2–S3) 18. They were stimulated to de qi 3 times during 30 minutes and, acupuncture points were used that have been stated by others 1920 to be avoided during pregnancy (LI 4, BL 31 and 33).

4. The investigation of the complications during pregnancy, pregnancy outcomes and the neonates were quite extensive compared to earlier published studies.

5. But the study was of insufficient size to exclude negative effects on preterm delivery, perinatal morbidity and mortality as well as on CTG.