Childbirth and prolapse: long-term associations with the symptoms and objective measurement of pelvic organ prolapse

Authors


Correspondence: Prof C Glazener, Health Services Research Unit, University of Aberdeen, Foresterhill, Aberdeen AB25 2ZD, UK. Email c.glazener@abdn.ac.uk

Abstract

Objectives

To investigate prolapse symptoms and objectively measured pelvic organ prolapse, 12 years after childbirth, and association with delivery mode history.

Design

Twelve-year longitudinal study.

Setting

Maternity units in Aberdeen, Birmingham and Dunedin.

Population

Women dwelling in the community.

Methods

Data from women were collected 12 years after an index birth and women were invited for examination. Logistic regression investigated associations between risk factors and prolapse symptoms and signs.

Main outcome measures

Prolapse symptom score (POP-SS); objectively measured prolapse (POP-Q).

Results

Of 7725 continuing women, 3763 (49%) returned questionnaires at 12 years. The median POP-SS score was 2 (IQR 0–4). One or more forceps deliveries (OR 1.20, 95% CI 1.04–1.38) and a body mass index (BMI) over 25 were associated with higher (worse) POP-SS scores, but age over 25 years at first delivery was associated with lower (better) scores. There was no protective effect if all deliveries were by caesarean section (OR 0.84, 95% CI 0.69–1.02). Objective prolapse was found in 182/762 (24%) women. Women aged over 30 years when having their first baby and parity were significantly associated with prolapse. Compared with women whose births were all spontaneous vaginal deliveries, women who had all births by caesarean section were the least likely to have prolapse (OR 0.11, 95% CI 0.03–0.38), and there was a reduced risk after forceps or a mixture of spontaneous vaginal delivery and caesarean section.

Conclusions

These findings are at odds with each other, suggesting that prolapse symptoms and objective prolapse may not be in concordance, or are associated with different antecedent factors. Further follow-up is planned.

Introduction

Pelvic organ prolapse affects over half of all women, and some degree of prolapse is nearly ubiquitous in older women (32.2% stage 1, 65.5% stages 2 or 3).[1] Although it often follows childbirth, it only becomes symptomatic later in life: the mean age for surgery for prolapse is 60 years.[2] It has a profound effect on quality of life, including not only the classic symptom of ‘something coming down’ (bulge or discomfort in the vagina), but often involving bladder, bowel and sexual function. Unlike urinary incontinence, its diagnosis is problematic: there are often discrepancies between the ‘objective’ measurement of descent of the vaginal walls or apex and the women's own report of prolapse symptoms.[3] Not all women with measurable descent have symptoms, and vice versa.

It is also possible that classic prolapse symptoms are not necessarily caused by classic anatomical findings.[3] If so, it may be that the two have different antecedent causes. In either scenario, the need for treatment will be based more on the symptoms than the signs, but curing the ‘objective prolapse’ may not necessarily cure the symptoms. A fuller understanding of the causes and effect on symptoms may provide guidance to clinicians and women grappling with the problems posed by pelvic organ prolapse.[4]

We present new data at 12 years after an index delivery from a longitudinal study of women after childbirth,[5-9] including the findings from a subsample examined using the pelvic organ prolapse quantification (POP-Q) system. The main research question was to investigate the relationship between prolapse and delivery mode history, and with other maternal factors.

Methods

The population included all of the women (= 10 989) who delivered over a 12-month period (1993–94) in three maternity units: in Aberdeen (UK), Birmingham (UK) and Dunedin (New Zealand). These deliveries are referred to as ‘index births’. The 7883 women who responded to an invitation to participate in this research 3 months after the index birth have been followed up during the subsequent 12 years, and form the source population for the current study, excluding those known to have died and those who requested no further contact in the intervening period (= 158). Baseline data on maternal and obstetric characteristics were obtained for all index deliveries to allow a comparison of responders and non-responders.

The current study involved a further questionnaire survey of the 7725 women remaining in the cohort, sent around 12 years after their index delivery. Women were also invited to a clinical examination to assess any degree of pelvic organ prolapse using the POP-Q system.[10]

Data collection and outcome measures

Outcome measures were women's reports of prolapse symptoms measured using the Pelvic Organ Prolapse Symptom Score (POP-SS),[11] and objectively measured prolapse using the POP-Q system on a subsample of women who agreed to be examined.[10] Obstetric and maternal data relating to the index delivery were obtained from routine hospital case notes or databases at the time of recruitment. The follow-up questionnaire at 12 years obtained date and mode of delivery from each woman for all of her births, which enabled delivery mode history to be obtained. Self-reported height and weight were used to determine body mass index (BMI). Valid weight and height measurements were given by 96.5% and 99.7% of respondents, respectively. Where BMI data were missing, mean imputation was used to estimate the BMI.

The POP-SS consists of seven items related to prolapse symptoms.[11] Each item is scored from 0 (never) to 4 (all of the time), and the seven item responses are summed to give a total score ranging from 0 to 28, where higher scores indicate more frequent symptoms. Other variables reported by women in the postal questionnaire included: urinary incontinence; faecal incontinence; having had a prolapse or continence operation; and ethnic origin (categorised as Indian subcontinent or not).

The POP-Q measurement system measures each prolapse compartment separately (anterior, posterior and apical).[10] We chose the leading edge to dichotomise stage-2 prolapse into two categories: measurements above the hymen (<0 cm), to indicate no prolapse, and measurements at the hymen or beyond (≥0 cm), to indicate significant prolapse. The latter measure was also chosen to define objective anterior or posterior vaginal wall prolapse, or apical prolapse.

Research question

The main research question was whether delivery mode history was associated with either prolapse symptoms or prolapse signs at 12 years after the index birth. Predefined questions probed the relative effects on these associations of: delivering exclusively by caesarean section; ever having a forceps delivery; or (separately) a ventouse delivery; or having both caesarean and spontaneous births. Other potential explanatory variables explored included: maternal age at first birth; parity (total number of births); BMI; and prolapse surgery.

Analysis

Ordinal logistic regression was used to assess the independent effects of delivery mode history on symptom outcome, and to adjust for and report on other independent predictors. In all models, adjustment was made for age in years at first birth, total number of births and BMI. Interaction terms were not included because of multicollinearity. All regressor variables were included and retained in the models.

The variable delivery mode history, created from the reported birth histories, categorised all of a woman's deliveries into: spontaneous vaginal delivery only (reference); caesarean section only; one or more forceps deliveries; one or more vacuum extractions, but no forceps; and the remainder as a combination of only spontaneous vaginal deliveries and caesarean sections. The forceps and vacuum categories could include women who also had spontaneous vaginal deliveries or caesarean sections. Replies with missing values in the mode of delivery history were omitted from the analysis. We were also able to conduct subsidiary regression models to explore two of the individual prolapse symptom questions (‘feeling of something coming down’ and ‘uncomfortable feeling in vagina’), and the effect of removing the women who had already had prolapse surgery.

In a parallel analysis using the same variables, binary logistic regression was used to assess their effects on the presence of objectively measured prolapse (i.e. at hymen or beyond). The analysis was based on overall stage (the most severe compartment stage).

Results

This study is one of a series exploring the long-term effects of childbirth on pelvic floor dysfunction. At the 12-year follow-up, we enquired for the first time about prolapse symptoms using the POP-SS, and examined a subsample of women to determine who had physical evidence of prolapse, as measured using the POP-Q system.

Completed questionnaires were returned by 3763/7725 women (a response rate of 49%). Of these, 1450 women consented to an examination and 762 women (20%) were subsequently examined. Baseline data on maternal and obstetric characteristics were obtained for all index deliveries and used to compare responders and non-responders. They differed in terms of baseline characteristics (Table 1): age (respondents older); ethnic origin (respondents less likely to be of Indian subcontinent ethnicity); parity (more respondents were having their first baby); and delivery mode (fewer respondents had spontaneous vaginal deliveries and more had assisted vaginal deliveries). Onset of labour and length of second stage of index birth were similar. The prevalence of faecal incontinence at 3 months was a little lower among respondents, but the prevalence of urinary incontinence was the same.

Table 1. Baseline characteristics of respondents and non-respondents at 12 years, and respondent women who were examined, at index delivery
n = 7883Non-respondent at 12 years n = 4120Respondent at 12 years n = 3763
  1. a

    This includes non-responders, those who had requested no further contact at 6 years and 41 women who had died.

  2. b

    Excludes caesarean section.

Died, moved away or withdrew from follow upa 158/7883 
Age at index birth P < 0.001 
Mean years (SD)27.7 (5.4) 29.2 (4.9)
Parity at index delivery (%) P = 0.006 
Primiparous1782 (43.6)1760 (46.8)
Multiparous2304 (56.4)2003 (53.2)
Not Known340
Mode of index delivery (%) P = 0.020 
Caesarean section645 (15.9)600 (16.3)
Forceps/breech382 (9.4)392 (10.6)
Vacuum166 (4.1)190 (5.2)
Spontaneous vaginal delivery2867 (70.6)2506 (68.0)|
Not known6075
Induced at index delivery (%) P = 0.657 
Yes695/4048 (17.2)646/3675 (17.6)
Not known7288
Second stage (%)b = 0.065 
≥1 hour986/2818 (35.0)1018/2724 (37.4)
Not known597364
UI at 3 months postpartum (%) = 0.654 
Incontinent1386/4120 (33.6)1247/3763 (33.1)
Not known00
FI at 3 months postpartum (%) = 0.038 
Incontinent392/4120 (9.5)307/3763 (8.2)
Not known00

Of the respondents at 12 years, the 762 women who were examined differed from those who were not examined in terms of: age (older); parity (fewer children); ethnic origin (less likely to be of Indian subcontinent ethnicity); current prolapse symptoms (higher symptom scores on POP-SS); other current symptoms of pelvic floor dysfunction (more urinary incontinence and faecal incontinence); and mode of delivery history (fewer caesarean sections only, more any forceps) (Table 2). Incontinence surgery was more prevalent in the women examined, although not significantly so. They were similar with respect to prolapse surgery and BMI.

Table 2. Characteristics of women and pelvic floor dysfunction at 12 years after an index delivery
 Respondent at 12 yearsNot examined at 12 yearsExamined at 12 years
  1. a

    No forceps.

n 37633001762
Centre (%)
Dunedin434 (11.5)269 (9.0)165 (21.7)
Birmingham1490 (39.6)1287 (42.9)203 (26.6)
Aberdeen1839 (48.9)1445 (48.2)394 (51.7)
Age at first birth  P = 0.001 
Mean (SD) 26.5 (4.9)26.3 (4.9)27.2 (4.8)
Number of births (%)  = 0.033 
1410 (10.9)337 (11.2)73 (9.6)
21836 (48.8)1428 (47.6)408 (53.5)
31016 (27.0)827 (27.6)189 (24.8)
4 or more500 (13.3)408 (13.6)92 (12.1)
Missing110
Mode of delivery history (%)  = 0.039 
Only spontaneous vaginal delivery1855 (49.4)1481 (49.5)374 (49.1)
Only caesarean section403 (10.7)342 (11.4)61 (8.0)
Any forceps956 (25.5)739 (24.7)217 (28.5)
Any vacuuma248 (6.6)197 (6.6)51 (6.7)
Spontaneous vaginal delivery and caesarean section297 (7.8)235 (7.8)59 (7.7)
Missing770
Current BMI  = 0.587 
Mean (SD) 26.0 (5.4)26.0 (5.4)26.1 (5.4)
Ethnic group (%)  < 0.001 
Non-Asian3600 (95.7)2845 (94.8)755 (99.1)
Indian subcontinent 163 (4.3)156 (5.2)7 (0.9)
POP-SS < 0.001 
Mean (SD) 2.66 (3.46) 2.41 (3.30)3.64 (3.88)
Current UI (any) (%) 1983/3763 (52.7)

< 0.001

1477/3001 (49.2)

506/762 (66.4)
Current FI (any) (%) 487/3763 (12.9)

< 0.001

356/3001 (11.9)

131/762 (17.2)
Prolapse operation by 12 years (%) 57/3681 (1.5)

< 0.054

39/2927 (2.4)

18/754 (2.4)
Continence (%) operation by 12 years (%) 52/3591 (1.4%)

< 0.320

38/2857 (1.3)

14/734 (1.9)

Prolapse symptoms (POP-SS)

The mean symptom score of respondents was 2.66 (SD 3.46, with scores ranging from 0 to 28; median POP-SS score of 2, IQR 0–4; = 3763).

Ordinal logistic regression of the prolapse score, adjusting for a range of independent antecedent variables, showed statistically significant associations: one or more forceps deliveries (OR 1.20, 95% CI 1.04–1.38) and a BMI greater than 25 (OR 1.31, 95% CI 1.15–1.50 for overweight; OR 1.59, 95% CI 1.36–1.87 for obese) were associated with higher (worse) POP-SS scores, and being aged over 25 years at first delivery was associated with lower (better) POP-SS scores for each of the three age categories (Table 3).

Table 3. Ordinal logistic regression of prolapse symptoms measured using POP-SS
VariableNumber

POP-SS

Mean (SD)

OR (95% CI) P
  1. a

    No forceps.

All women 37632.7 (3.5)  
Age at first birth
<25 years12793.4 (4.1)Reference 
25–29 years14972.4 (3.1)0.68 (0.60–0.78) <0.001
30–34 years7852.2 (2.9)0.63 (0.53–0.74) <0.001
≥35 years1972.0 (2.7)0.60 (0.46–0.80) <0.001
Missing5   
Number of births
14102.6 (3.2)Reference 
218362.5 (3.2)0.84 (0.69–1.02) 0.075
310162.6 (3.6)0.76 (0.61–0.93) 0.010
4 or more5003.4 (4.1)0.92 (0.71–1.18) 0.506
Missing1   
Delivery mode history
Only spontaneous vaginal delivery18552.7 (3.5)Reference 
Only caesarean section 4032.1 (2.8)0.84 (0.69–1.02) 0.076
Any forceps9562.9 (3.6)1.20 (1.04–1.38) 0.012
Any vacuuma2482.4 (3.2)0.93 (0.73–1.18) 0.547
Spontaneous vaginal delivery and caesarean section2972.9 (3.5)1.13 (0.90–1.41) 0.282
Missing7   
Current body mass index
18.5–24.918432.3 (3.1)Reference 
<18.5613.3 (4.2)1.36 (0.84–2.21) 0.215
25–29.911842.9 (3.6)1.31 (1.15–1.50) <0.001
≥306753.3 (3.9)1.59 (1.36–1.87) <0.001
Missing0  

The findings were substantially unchanged when the analysis was restricted to women who had not previously had prolapse surgery, and if either one of the first two questions of the POP-SS score were used as the dependent variable (for ‘a feeling of something coming down’ and for ‘vaginal discomfort with standing’). We also explored the associations restricted to women who were examined. The POP-SS score in the group as a whole was higher (Table 2), but the associations were similar with respect to the other confounding factors, particularly age.

Objectively measured prolapse (POP-Q)

Around 20% of the women who responded at 12 years were examined using the POP-Q system to measure prolapse. Of the 762 women examined, just under half had no prolapse (stage 0, n = 49, 6.4%, or stage 1, n = 286, 37.5%). A further 415 women (54.5%) had stage 2 prolapse. Of these, 245 women had prolapse above the hymen, and the remainder had prolapse at the hymen or beyond. 9 (1.2%) and 3 (0.4%) women had stage 3 and 4 prolapse, respectively. If one takes prolapse at the hymen or beyond as a working definition of significant prolapse, 182 (24%) women could be diagnosed as having a prolapse. Of these women, 149 had prolapse in only one compartment, 24 had prolapse in two compartments and nine had prolapse in all three compartments. Multivariate analysis showed that women whose age was over 30 years at first delivery, and who had had more than one baby, were independently significantly associated with prolapse (Table 4). In addition, delivery history was associated with the presence of prolapse: relative to women who had all their deliveries by spontaneous vaginal delivery, those who had exclusive caesarean section deliveries were less likely to have prolapse (OR 0.11, 95% CI 0.03–0.38), and to a lesser extent so were women who had at least one forceps delivery (OR 0.64, 95% CI 0.42–0.96) or at least one each of spontaneous vaginal delivery and caesarean section deliveries (OR 0.48, 95% CI 0.22–0.97). Current BMI, however, was not associated with objectively measured prolapse (Table 4).

Table 4. Logistic regression of prolapse staging measured using POP-Q to define leading edge of prolapse (adjusted, i.e. leading edge at hymen or beyond)
VariableNumberProlapse (%)OR (95% CI) P
  1. a

    No forceps.

All women 182/76224  
Age at first birth (P = 0.099)
≤24 years40/20719Reference 
25–29 years71/312231.46 (0.92–2.31) 0.108
30–34 years56/191292.49 (1.49–4.18) 0.001
35+ years15/52293.08 (1.43–6.61) 0.004
Number of births (= 0.048)
18/7311Reference 
2102/408253.30 (1.49–7.32) 0.003
347/189253.93 (1.69–9.18) 0.002
4 or more25/92275.23 (2.04–13.39) 0.001
Delivery mode history (= 0.005)
Only spontaneous vaginal delivery107/37429Reference 
Only caesarean section 3/6150.11 (0.03–0.38) <0.001
Any forceps48/217220.64 (0.42–0.96) 0.031
Any vacuuma13/51250.71 (0.35–1.42) 0.338
Spontaneous vaginal delivery and caesarean section11/59190.48 (0.22–0.97) 0.041
Current BMI (= 0.826)
Normal81/36222Reference 
Underweight3/10301.19 (0.28–5.01) 0.812
Overweight64/248261.33 (0.90–1.96) 0.150
Obese34/142241.48 (0.91–2.40) 0.111

Discussion

Principal findings

Our primary research question was the relationship between childbirth and subsequent prolapse, both symptoms and signs. The mean symptom score (POP-SS) was low (median 2), as would be expected in a relatively asymptomatic population of community-dwelling women. The proportion with a leading edge at the hymen or beyond (24%) was very similar to that in an older population (25.2%, mean age of 68.3 years) of 270 women examined at one centre of the Women's Health Initiative Hormone Replacement Therapy randomised controlled trial.[1] In another study, 22% of 1004 women with a mean age of 42 years had a prolapse at −0.5 cm or beyond.[12]

For prolapse symptoms measured in nearly 4000 women using a prolapse symptom score (POP-SS), only BMI over 25 and any forceps delivery, compared with women who had all their deliveries by spontaneous vaginal delivery, were associated with more frequent symptoms (Table 3). Although women had fewer symptoms if all their deliveries were by caesarean section, this did not reach statistical significance. The relationship between prolapse symptoms and parity was unclear when other factors were taken into account in multivariate analysis. Women who were under the age of 25 years when they had their first birth had significantly more prolapse symptoms (higher POP-SS scores) than women who had their first baby when they were older.

In contrast, in the 762 women who were examined, women were more likely to have prolapse at the hymen or beyond (24%) if they were older (over 30 years) when they had their first birth, if they had more than one baby and if they had all their deliveries by spontaneous vaginal delivery. Although women were less likely to have prolapse if all their deliveries were by caesarean section, this was also true for women who had at least one forceps delivery or mixed spontaneous vaginal delivery and caesarean section deliveries. Surprisingly, current BMI was not associated with prolapse. However, the women who were examined were slightly more likely to have symptoms of pelvic floor dysfunction than those who were not examined in terms of higher POP-SS scores, and greater numbers reported urinary or faecal incontinence.

Strengths and weaknesses

This large longitudinal cohort is being followed prospectively from the time of an index delivery in 1993–94. The women were recruited from three different centres and represent the largest prospective, longitudinal postpartum cohort, with the most detailed data concerning mode of delivery history, in the literature.[8] Although causality cannot be assumed from statistical associations, other studies have also identified childbirth and parity as significant antecedent factors.[1, 12] Although we do not have any nulliparous women in our study, it is well recognised that the rates of pelvic floor dysfunction (although not non-existent) are considerably less amongst this population.[1, 13]

We have been able to take account of all the deliveries for each woman using a classification scheme that was developed at an earlier stage of this research. However, around half the women have dropped out over the 12-year time period. Although there are some systematic differences between the responders and non-responders (the latter were younger, were more likely to be multiparous and delivered by spontaneous vaginal delivery when first recruited, with less perineal trauma and slightly less faecal incontinence), the differences were small, and would, if anything, tend to be associated with less risk of subsequent pelvic floor dysfunction. Thus, our data might represent at worst a small overestimate of the prevalence of the problems, but this should have little impact on the relationships between the variables being examined. Another weakness is that a validated questionnaire was not used to assess urinary or faecal incontinence, but at the time of initial recruitment (1993–94) there were none available. For consistency, we used exactly the same questions at each follow-up time point.

A further discussion of risk of bias is given in our previous publication from this cohort.[9]

Although we were only able to examine a relatively small number of women (762), this is comparable with other similar studies that involved examining women and recording PO-Q scores.[1, 12, 14] In addition, we had unique prospective data collection, virtually complete mode of delivery histories and extensive symptom data, recorded where possible using validated questionnaires.

Those who came forward to be examined were more likely to be symptomatic than those who did not, in respect of older age, higher POP-SS scores, current urinary and faecal incontinence, parity and mode of delivery history. It is likely that some of these women chose to be examined because they had already started to have symptoms of pelvic floor dysfunction, leading to some selection bias.

Meaning of the study

Our population comprised substantially asymptomatic women living in the community. The mean age of the women was 42 years (range 26–58 years) at this follow-up point. Prolapse symptoms normally become severe enough to require treatment after the menopause, so the majority of our women were possibly too young to be symptomatic; nevertheless, just under 2% had already had prolapse or incontinence surgery. However, the findings of our analyses were the same when these women were excluded.

On the whole, we feel that these findings raise more questions than they answer. There is clearly a relationship between childbirth and subsequent pelvic floor dysfunction, but this varies subtly depending on the aspect of dysfunction that is of interest. In this paper we have explored separately the effects of a range of confounding variables on both prolapse symptoms and prolapse signs, with some consistent and some conflicting findings. In previous work we have demonstrated similar issues when exploring the antecedents of urinary and faecal incontinence.[5-9]

There are a number of reasons that may explain the inconsistencies. Our population was largely asymptomatic for prolapse symptoms, as demonstrated by the low POP-SS score (median POP-SS score of 2, IQR 0–4, from a total possible score of 28). The mean age of the women was 42 years, which is substantially younger than the age group who seek treatment. Despite this, our estimate of 24% of women having a prolapse at the hymen or beyond accords well with the results of a multicentre observational study by Swift[12] of 1004 women of a similar age (mean age 42.7 years) having an annual gynaecological examination (22%), and with the 25.2% rate reported by Nygaard et al.[1] from a survey of 270 older women (mean age 68.3 years) enrolled in a randomised controlled trial of hormone replacement therapy. Both these studies subdivided stage 2 prolapse using the POP-Q method around the hymen in a similar way to our analysis.

Another explanation may be the lack of correlation between measurable prolapse and its symptoms in women.[3] This may explain why different variables are selected by the two models as being associated with the symptoms (Table 3) or the objective measurements (Table 4) of prolapse.

Another inconsistency between the associations of the prolapse symptoms and the objective measures concerned maternal age at first birth. We were surprised to observe that prolapse symptoms were higher in women who were younger when they had their first delivery, whereas prolapse signs were more prevalent in women who were older when they had their first delivery. Further exploratory analyses showed similar trends with current age. We postulated that symptoms might improve with time after delivery as women recovered from the trauma of childbirth. Therefore, we also explored the relationship between prolapse symptoms and the number of years since the last delivery: there was some evidence that symptoms appeared higher only in the women who had started their families at a younger age regardless of time since last delivery. An alternative explanation might be that younger women have higher expectations of their health, and hence report minor symptoms more readily than older women. We are unable to explain this finding without further analysis.

Conclusion

Childbirth clearly has an influence on subsequent pelvic floor dysfunction. The association between women's prolapse symptoms and mode of delivery history (more symptoms after at least one forceps delivery) suggests that forceps delivery has an adverse effect on pelvic floor function, whereas delivering exclusively by caesarean section does not appear to be protective. However, the lack of association between subjectively reported prolapse symptoms and objectively measured prolapse suggests that this is not necessarily mediated through the mechanical changes of pelvic organ descent.

This study also suggests that if women wish to reduce their risk of developing objective prolapse, they might consider having their first child before the age of 30 and have fewer children. Women who have all their deliveries by caesarean section have a reduced risk of developing prolapse, but the effect of different vaginal modes of delivery is less clear.

Unanswered questions and future research

We plan further long-term follow-up of this cohort when the women reach 20 years after recruitment. We are particularly interested in the fluctuation of symptoms of pelvic floor dysfunction, in women's need to access treatment and its outcomes.

Disclosure of interests

All authors declare that they have no interests to declare.

Contribution to authorship

CG, CM, DW, RL, PH and AMcD contributed to the design and analysis of the entire ProLong study. AE, GMcP, SH, CB, KR, ND and PT-H joined the study group at 6 or 12 years, and contributed to this part of the study. GMcP was responsible for the database design. CG drafted the article, AE and RL analysed the data, and all authors contributed to writing and commented. A previous member of the ProLong study group was Adrian Grant.

Details of ethics approval

Ethical approval from UK centres for the 12-year follow-up was obtained from the Multicentre Research Ethics Committee Edinburgh, ref. no. RG 819/06, in November 2007, and from the New Zealand National Ethics Committee, ref. no. LRS/05/04/009 in March 2005.

Funding

Grants were obtained from Wellbeing of Women/Royal College of Obstetricians and Gynaecologists for the UK follow-up and from the Health Research Council of New Zealand for the follow-up in New Zealand.

Acknowledgements

The authors would like to thank the women who took part in the study and Anne-Marie Rennie, Jane Cook and Jane Harvey who were involved in assembling the original cohort. Margaret MacNeil was involved in the administration of the follow-up study. We are also grateful to our funders.

Ancillary