The PPIS trial was designed at the same time as a number of other very different pregnancy interventions designed to increase mean birth weight, and to reduce low birth weight and preterm birth through the provision of social support or enhanced nutrition [17–23]. None of the others implemented their intervention package before pregnancy but most aimed to recruit women relatively early in pregnancy. Four of these five trials and a recently reported additional trial  identified unexpected adverse outcomes, particularly extremely preterm births, or late terminations of pregnancy, i.e. findings similar to those in the PPIS trial. One factor which all of those trials had in common was that the intervention was provided by someone who was not the usual provider of antenatal care. In the one trial where there were no such unexpected adverse outcomes the supportive intervention was provided at home by midwives, a pattern within usual maternity care in the UK, where the trial was carried out . Midwives provided the intervention in the PPIS trial at home but this was not usual care in Australia.
Three hypotheses to explain the possible adverse outcomes of the PPIS trial came out of discussions with colleagues in Australia and Canada. The first was that the intervention itself might have heightened parental stress and anxiety about perinatal outcomes, with stress and anxiety contributing to very preterm birth [25, 26]. The second was that there might have been an increased risk of conception at the extremes of the fertile period, giving rise to unplanned conception or less than ideal implantation, if women and their partners had given up their usual method of contraception for an alternative one. The shorter intervention to conception interval associated with the very preterm births in the intervention arm might be a marker for this risk factor. The third hypothesis was that if women and their partners took a very positive approach to the full intervention package it might have improved their overall health to a point where fetuses with a birth anomaly or with poor placentation were actually sustained in utero for a longer period of time, with a shift from miscarriage to very preterm birth.
Alternative explanations are that detailed discussions of family health increased perceptions of risk, or led to difficulties between the parents or difficulties with their families of origin. Recommendations to stop smoking, alcohol and over the counter drugs might have been another problematic piece of advice. It is also possible that the 'medicalisation' of such a human and family-focussed event as pregnancy and birth was in itself harmful. We could not totally rule out a contribution from selective participation in the study at the initial home visit: 65 women in the intervention arm, compared with 44 in standard care, opted out of the study. Overall, there were no significant differences in participation between the two arms.
Informed consent of participants was sought through the three-fold process outlined in the Methods: discussion of the trial by the MCHN, verbal agreement by the woman to have her details given to the PPIS midwife and subsequent contact by the midwife to make an appointment and follow-up with a visit. Consent of the Community Health Centre, and the Maternal and Child Health service was based on a series of consultations held with each group, before the trial was implemented to describe the study and respond to queries, comments and criticisms. The Community Health Centre did not have a Human Research Ethics Committee and there was no equivalent body at that time in the then Health Commission of Victoria. When the project was awarded National Health and Medical Research Council research funding to complete the follow-up of second births (1992) we were required to provide written information about the study to all subsequent participants and to seek formal Ethics approval by the Board of the Community Health Centre. The Board co-opted an Ethics Committee Member from one of the nearby teaching hospitals to assist in this process.
Potential limitations of the intervention included the absence of advice to take periconceptional folate. In 1981–2 there was still marked disagreement about the benefits and risks of vitamins and folate before conception and in the first months of pregnancy. Publication of the proceedings of a workshop held in late 1982 for experts involved in recent or planned research on the prevention of neural tube defects, discussed the available evidence and its shortcomings and we were influenced by the evident uncertainty and disagreement . By the time the four trials of periconceptional supplementation were published [28–31] recruitment to the PPIS trial was virtually complete.
Another potential limitation was the quality of the evidence on the effect of alcohol use before pregnancy. Several US papers had reported adverse effects on the fetus of even extremely low maternal consumption of alcohol, in early pregnancy or immediately prior to pregnancy, as well as effects of paternal alcohol use around the time of conception [32, 33]. Despite the biological implausibility of some of these effects and the relatively small numbers of people in the studies we did advise stopping alcohol consumption before and during pregnancy. Contemporary population-based data on alcohol and tobacco use, collected in early pregnancy in another Australian state (Tasmania) did not show adverse effects of alcohol on fetal growth or preterm birth until maternal consumption was reported at more than two standard drinks a day , a level of drinking reported by 0.4% of Tasmanian women. Subsequent research in large cohorts has not confirmed the early US studies [35, 36].
Limitations of the trial included slow recruitment. It occurred at half the rate planned, extending the duration of the trial and making it harder to maintain local interest and Maternal and Child Health Nurses' enthusiasm. Another factor slowing recruitment was staff turnover . A common work pattern at the time was moving to a new position after two years, to get a broad experience in community settings. Each new staff member needed a familiarization and training period: many were unfamiliar with randomised trials at the time of recruitment.
A greater limitation was the research team's lack of recognition that the women participating were of lower, rather than higher risk of adverse pregnancy outcomes. The decision to recruit women through existing services for mothers and infants failed to take into account the fact that women who had adverse outcomes in their first pregnancy (a perinatal death, an infant with a major congenital anomaly or a very preterm infant) would be under-represented. Mothers whose infants had died would not be part of MCH services and mothers of infants requiring on-going care from hospitals or specialist services were likely to have been using those services, or visits to private paediatricians, for routine health advice and support as an alternative to their local MCHN. None of the first-born infants included in the PPIS trial was born before 30 weeks gestation.
More than 50% of participants moved house before the birth of their second child, making follow-up a much larger part of the midwives' role than originally planned . The average interval between the first and second births was a mean of 39 months compared with the predicted two years. This extended the time required for follow-up and made follow-up more difficult.
Other pre-pregnancy interventions are sparse. A recent systematic review of preconception interventions  found only one randomised trial  which used an opportunistic strategy, recruiting women who had a negative pregnancy test, explaining the trial, inviting participation, and randomising by tossing a coin. All participants were given a preconception risk survey. Although an average of over eight risks per woman were identified, and more than half the women made another visit to the service in the next year, there was no difference between intervention and usual care arms in the proportion of risks addressed. A recently reported trial recruited women who had given birth spontaneously before 34 weeks gestation, to be randomised four months after that birth to receive either oral azithromycin 1 g twice (4 days apart) plus 750 g of sustained-release metronidazole daily for seven days, or identical – appearing placebos, repeated every 4 months until the subsequent pregnancy. The intervention did not reduce subsequent preterm birth with these authors also concluding that the intervention might be associated with a shorter gestational age and lower birth weight .
The PPIS study strengths are that it remains the only completed trial of a broad pre-pregnancy intervention. It demonstrated that it was feasible to test a pre-pregnancy intervention using a 'gold standard' design within socially disadvantaged communities, including women on low incomes and culturally diverse women, incorporating the intervention into an existing population-wide service. It provided a model for joint action between four community health services. It also identified issues relevant to the planning of future trials. Data on the high proportion of participants who turned out to be ineligible or lost to the study are likely to be relevant to other population based pre-pregnancy interventions.