The authors have declared that no competing interests exist.
Conceived and designed the experiments: MJH SS CG NM. Performed the experiments: KD MJH SS CG. Analyzed the data: KD MJH SS CG. Wrote the paper: KD MJH SS CG NM.
It has been suggested that mode of delivery, a potentially powerful influence upon long-term health, may affect later life body mass index (BMI). We conducted a systematic review and meta-analysis of the effect of Caesarean section (CS) and vaginal delivery (VD) on offspring BMI, overweight (BMI>25) and obesity (BMI>30) in adulthood. Secondary outcomes were subgroup analyses by gender and type of CS (in-labour/emergency, pre-labour/elective).
Using a predefined search strategy, Pubmed, Google Scholar and Web of Science were searched for any article published before 31st March 2012, along with references of any studies deemed relevant. Studies were selected if they reported birth characteristics and long-term offspring follow-up into adulthood. Aggregate data from relevant studies were extracted onto a pre-piloted data table. A random-effects meta-analysis was carried out in RevMan5. Results are illustrated using forest plots and funnel plots, and presented as mean differences or odds ratios (OR) and 95% confidence intervals.
Thirty-five studies were identified through the search, and 15 studies with a combined population of 163,753 were suitable for inclusion in the meta-analysis. Comparing all CS to VD in pooled-gender unadjusted analyses, mean BMI difference was 0·44 kg·m-2 (0·17, 0·72;
There is a strong association between CS and increased offspring BMI, overweight and obesity in adulthood. Given the rising CS rate worldwide there is a need to determine whether this is causal, or reflective of confounding influences.
An
The last twenty years have seen worldwide increases in obesity prevalence in children and adults, with the highest incidences reported in the USA and Scotland (33·8% and 30% respectively)
Concurrently, between 1990 and 2008, there has been a 100% increase in Caesarean Section (CS) births in England
Adverse effects of CS on the neonate immediately post-partum are widely recognised. CS is associated with the highest rates of neonatal morbidity and mortality of all modes of delivery
Controversially, it has been suggested that birth by CS predispose offspring to adverse health outcomes in childhood
We aimed to perform a systematic review and meta-analysis to identify any association between mode of delivery and offspring body mass index (BMI), and risk of overweight, and obesity in adulthood. We also sought to determine if offspring age, gender and type of CS had an effect on outcomes. In contrast to Li
A systematic review of studies reporting adult anthropometry (BMI, height, weight, incidence of overweight/obesity) by mode of delivery was conducted using an
Outcomes studied were offspring BMI, overweight and obesity in adulthood (≥18 years). Overweight and obesity were classified according to the National Institute of Clinical Excellence categories
A search was conducted in PubMed (
Titles and abstracts of identified studies were independently screened by two reviewers [KD and MJH]. If abstracts were unavailable, the full text was obtained. The full texts of relevant abstracts were appraised for inclusion [KD and CG]. Any disagreement over eligibility of a study was referred to all authors. For inclusion a study must have reported either: (1) both mode of delivery and adult offspring BMI; (2) mode of delivery with long-term offspring follow-up (into adulthood) or (3) adult offspring BMI with birth characteristics. References of included studies were hand searched for relevant publications. Review articles and letters to editors were excluded after reference lists were searched. If multiple papers reported data from the same cohort, the study reporting BMI at an age closest to the median age across all studies was included.
A basic dataset was extracted from each study using a pre-piloted data collection form [KD and SS]. In studies where required data were not reported or only adjusted results provided, authors were contacted to request these [KD and MJH]. If no response was received after two emails or relevant data were unavailable, the study was excluded from the meta-analysis. Where authors were not contactable, the principal investigator of the cohort was approached. Where mean and SD for BMI was provided by gender only, combined means and SD for both genders together were calculated.
Methodological quality of each study was assessed using a modified Newcastle-Ottawa Scale
Meta-analyses were carried out using the inverse-variance method for mean BMI difference and the Mantel-Haenszel method for overweight and obesity odds ratios (OR) in RevMan 5 (5.0.23) to identify any association between mode of delivery and adult BMI, overweight and obesity separately. Forest plots were created using RevMan 5 (5.0.24). Funnel plots were created and Egger's Test
Random-effects models were used throughout as it was considered unlikely that the effect of interest was the same across all studies, invalidating the main assumption for fixed-effect models. Heterogeneity (between study variation) was assessed using the chi-squared test for Cochrane's Q statistic and by calculating I2, the estimated proportion of variance in the study outcome due to heterogeneity
However, where heterogeneity was low (p>0·05 from the chi-squared test and I2<50%), a fixed-effects model was carried out to check the sensitivity of findings as a random effects analysis can give greater weight to smaller studies. As heterogeneity tests have low power when study numbers are small
For all random effects analyses 95% prediction intervals (PI) were calculated as this provides a useful measure of the effects we would expect to see in future studies
Pre-specified subgroup analyses were carried out by gender and type of CS (in-labour-CS/pre-labour-CS) for studies providing relevant data. Where a substantial difference in the magnitude of BMI difference was shown between subgroups, statistical significance was tested using meta-regression. Meta-regression was also used to determine whether study results varied with age of offspring.
If type of CS data were available for only a subgroup of studies, to confirm that any difference in results was not due to a subgroup effect, the mean BMI difference, overweight OR and obesity OR was calculated for the subgroup comparing all CS deliveries (i.e. not separated into pre-labour-CS or in-labour-CS) to VD.
The search yielded 3292 abstracts (PubMed: 1413; Google Scholar: 1850; Web of Science: 29). Google Scholar imposes a 1000 result limit, so only the first 1000 abstracts were screened. Consequently, 2442 titles and abstracts were screened for inclusion. Seventy-nine abstracts appeared relevant; through hand-searching their references a further 90 abstracts were screened of which 57 appeared relevant. A further paper was identified through external means. Duplicate datasets, and papers which on full examination evidently had not collected mode of delivery data, were removed, leaving 35 studies with apparently relevant data (
The relevant number of papers at each point is given.
The authors of 30 studies were contacted
The 1958 British Birth Cohort dataset included implausible BMI values, so limits were imposed (excluding subjects with BMI>200 kg·m−2 and BMI<10 kg·m−2). Mean (SD) BMI was calculated following exclusion of these implausible data.
The 1924 Helsinki Birth Cohort
In total, 15 studies were included (combined population 163,753). Study descriptions and data are shown in
Authors | Study Details | Factors adjusted for | NOS | Follow-up Rate | Population Size | Mean Age | Sex | MoD | Mean + SD BMI | N | % of Sample BMI > 25 (n) | % of Sample BMI > 30 (n) | |
PC; Finland; Persons born in Helsinki during 1934–1944 at University Central Hospital that still lived in Finland in 1971. All living members were sent a questionnaire in 2000, and a random sample of 2901 was invited for clinical examination. Of this sample data were obtained on 1999. Investigators were blinded to birth data. | N/A | 5* | 68.9 | 1999 | 61.6 | 27.5 (4.3) | 899 | 73.53 (661) | 22.25 (200) | ||||
27.7 (2.8) | 28 | 85.71 (24) | 25 (7) | ||||||||||
27.7 (5) | 1056 | 67.71 (715) | 27.46 (290) | ||||||||||
27.7 (5.7) | 18 | 66.67 (12) | 16.67 (3) | ||||||||||
27.6 (4.7) | 1955 | 70.38 (1376) | 25.06 (490) | ||||||||||
27.7 (4.1) | 46 | 78.26 (36) | 21.74 (10) | ||||||||||
PC; UK; All births in England, Scotland and Wales between 3rd–9th March 1958. Height and weight were measured during an interview when subjects were 33 for 10,490 subjects from a target sample of 15,667. Pre-terms, pregnant women and offspring from multiple pregnancies were excluded. Investigators were blinded to birth data. | N/A | 5* | 67 | 10490 | 33 | 25.71 (4.78) | 5011 | 51.65 (2588) | 11.28 (565) | ||||
26.21 (4.17) | 145 | 55.17 (80) | 14.48 (21) | ||||||||||
24.66 (5.56) | 5209 | 35.71 (1860) | 12.06 (628) | ||||||||||
25.61 (5.70) | 125 | 40.8 (51) | 16.8 (21) | ||||||||||
25.18 (5.218) | 10220 | 43.52 (4448) | 11.67 (1193) | ||||||||||
25.93 (4.937) | 270 | 48.52 (131) | 15.56 (42) | ||||||||||
26.24 (5.278) | 154 | 49.35 (76) | 18.18 (28) | ||||||||||
25.53 (4.435) | 116 | 47.41 (55) | 12.06 (14) | ||||||||||
PC; UK; All children born in Great Britain between 5th – 11th April 1970, of which 9316 were followed up at 34 years old. Height and weight data at 34 was obtained by confidential self-reporting. Investigators were blinded to birth data | N/A | 4* | 52.12 | 8634 | 34 | 26.661 (4.579) | 3983 | 61.11 (2434) | 17.9 (712) | ||||
26.624 (4.677) | 189 | 59.26 (112) | 19.6 (37) | ||||||||||
25.170 (5.160) | 4267 | 40.47 (1727) | 15.4 (658) | ||||||||||
25.705 (5.206) | 195 | 46.67 (91) | 15.9 (31) | ||||||||||
25.89 (4.944) | 8250 | 50.44 (4161) | 16.61 (1370) | ||||||||||
26.16 (4.968) | 384 | 52.86 (203) | 17.71 (68) | ||||||||||
PC; Scotland; Survey of all primary school children aged 6-12 in Aberdeen 1962-1964. Follow-up information in 2001-2003 was conducted via postal questionnaire when participants were aged 43-53. Investigators were blinded to birth data. | N/A | 3* | 61.6 | 6948 | 43–53 | 26.83 (4.043) | 3233 | 64.06 (2071) | 18.65 (603) | ||||
26.877 (3.854) | 103 | 67.96 (70) | 19.42 (20) | ||||||||||
26.21 (5.444) | 3505 | 49.61 (1739) | 19.06 (668) | ||||||||||
26.398 (5.507) | 107 | 53.27 (57) | 20.56 (22) | ||||||||||
26.51 (4.832) | 6738 | 56.54 (3810) | 18.86 (1271) | ||||||||||
26.63 (4.763) | 210 | 60.48 (127) | 20 (42) | ||||||||||
26.46 (4.777) | 91 | 57.14 (52) | 17.58 (16) | ||||||||||
26.76 (4.769) | 119 | 63.03 (75) | 21.85 (26) | ||||||||||
PC; Brazil; All live babies born in 1982 in Pelotas were recruited and then traced and in 2004–2005; blinding not stated. | Sex; birthweight; physical activity; schooling at time of assessment; family income at time of assessment; smoking; alcohol consumption; family income at birth; method of payment for the delivery and maternal factors: skin colour; age; parity; smoking; height; pre-pregnancy weight. | 5* | 75.34 | 4288 | 22.8 | 23.6 (4.08) | 1583 | 27.61 (437) | 7.3 (116) | ||||
24.3 (4.01) | 623 | 38.2 (238) | 8.2 (51) | ||||||||||
23.5 (4.66) | 1503 | 26.75 (402) | 8.9 (134) | ||||||||||
23.7 (4.63) | 579 | 27.63 (160) | 9.7 (56) | ||||||||||
23.551 (4.376) | 3086 | 27.19 (839) | 8.1 (250) | ||||||||||
24.011 (4.328) | 1202 | 33.11 (398) | 8.9 (107) | ||||||||||
PC; Brazil; A random sample of singleton-born subjects from a cohort of all births registered in Ribeirão Preto between 1st June 1978 to 31st May 1979 that were still alive and living in the city between 2002–2004; all investigators and personnel were blinded to birth data. Of note there were some differences described between the sample and the whole cohort. | Sex; SES; birthweight; smoking, education; physical activity; maternal factors: smoking; education; | 7* | 97.81 | 2057 | 23.9±0.71 | 24.6 (4.23) | 675 | 39.85 (269) | 10.67 (72) | ||||
25.9 (4.76) | 317 | 50.47 (160) | 17.35 (55) | ||||||||||
23.3 (4.78) | 725 | 27.45 (199) | 10.21 (74) | ||||||||||
24.2 (5.59) | 340 | 32.35 (110) | 13.24 (45) | ||||||||||
23.927 (4.568) | 1400 | 33.43 (468) | 10.43 (146) | ||||||||||
25.02 (5.271) | 657 | 41.1 (270) | 15.22 (100) | ||||||||||
PC; Finland; Control group from the VLBWA cohort; subjects recruited were the next available singleton infant born at term of the same sex who was not small for gestational age; investigators blinded to birth data. | N/A | 5* | 54.8 | 176 | 22.49 | 23.37 (3.307) | 60 | 31.67 (19) | - | ||||
22.61 (2.597) | 12 | 8.33 (1) | - | ||||||||||
22.55 (3.723) | 96 | 19.79 (19) | - | ||||||||||
25.7 (1.902) | 8 | 50 (4) | - | ||||||||||
22.87 (3.58) | 156 | 24.36 (38) | - | ||||||||||
23.85 (2.768) | 20 | 25 (5) | - | ||||||||||
23.18 (2.319) | 11 | 27.27 (3) | - | ||||||||||
24.655 (3.181) | 9 | 22.22 (2) | - | ||||||||||
PC; Finland; Persons born between 1924 and 1933 at the Helsinki University Central Hospital who resided in Finland in 1971, and remained in Finland thereafter. Investigators were blinded to birth data. | N/A | 4* | 7.1 | 500 | 69.6 | 27.61 (4.358) | 495 | 71.72 (355) | 24.44 (121) | ||||
26.44 (2.906) | 5 | 60 (3) | 20 (1) | ||||||||||
PC; China; Adult offspring of women who had sequential, live, singleton births at Peking Union Medical College between July 1948 to December 1954, still residing in Beijing. Height and weight data were measured by one of two observers. Investigators were blinded to birth data. | N/A | 4* | 86.6 | 628 | 45 | 24.74 (3.07) | 256 | 38.67 (99) | 5.86 (15) | ||||
24.25 (3.23) | 53 | 45.3 (24) | 3.77 (2) | ||||||||||
23.31 (3.07) | 266 | 24.81 (66) | 2.26 (6) | ||||||||||
23.58 (3.25) | 53 | 26.42 (14) | 3.77 (2) | ||||||||||
24.01 (3.149) | 522 | 31.61 (165) | 4.02 (21) | ||||||||||
23.915 (3.242) | 106 | 35.85 (38) | 3.77 (4) | ||||||||||
PC; India; All singletons born alive at Holdsworth Memorial Hospital, Mysore between 1934 to 1954, living within 8 square miles of the hospital, identified between 1993 to 2004. Investigators were blinded to birth data. | N/A | 4* | 4.3 | 1064 | 46.8 | 23.09 (4.201) | 536 | 33.02 (177) | 5.04 (27) | ||||
24.2 (4.3) | 14 | 35.71 (5) | 7.14 (1) | ||||||||||
25.546 (5.267) | 507 | 53.85 (273) | 16.96 (86) | ||||||||||
25.6 (4.4) | 7 | 57.14 (4) | 14.29 (1) | ||||||||||
24.263 (4.932) | 1043 | 43.15 (450) | 10.83 (113) | ||||||||||
24.6 (4.3) | 21 | 42.857 (9) | 9.52 (2) | ||||||||||
PC; Netherlands; All newborn babies in the Department of Obstetrics of the University Hospital in Groningen between 1975 to 1978, were sent a postal questionnaire in 1998. All responders to the questionnaires were invited for medical characterisation where height and weight were measured. | N/A | 4* | 38 | 590 | 20.49 | 22.5 (2.7) | 248 | 14.9 (37) | 1.2 (3) | ||||
22.9 (3.4) | 17 | 23.5 (4) | 5.9 (1) | ||||||||||
23.4 (4.0) | 300 | 26.7 (80) | 7.7 (23) | ||||||||||
23.1 (2.9) | 25 | 28 (7) | 0 (0) | ||||||||||
23.0 (3.5) | 548 | 21.4 (117) | 4.7 (26) | ||||||||||
23.0 (3.1) | 42 | 26.2 (11) | 2.4 (1) | ||||||||||
23.2 (3.2) | 25 | 32 (8) | 4 (1) | ||||||||||
22.7 (3.0) | 17 | 17.6 (3) | 0 (0) | ||||||||||
RC; Sweden; study subjects were randomly identified through national population registers. Inclusion criteria were male subjects aged 18–20, willing to participate in the study. Detailed birth records were identified for these subjects. Height and weight readings were taken by trained personnel. Investigators were blinded to birth data. | N/A | 3* | 94.3 | 1008 | 18.9±0.6 | 22.351 (3.278) | 867 | 16.15 (140) | 3.91 (34) | ||||
22.624 (3.146) | 141 | 14.89 (21) | 2.13 (3) | ||||||||||
22.505 (3.444) | 70 | 15.7 (11) | 1.43 (1) | ||||||||||
22.741 (2.842) | 71 | 14.1 (10) | 2.82 (2) | ||||||||||
Record Linkage Study; Sweden; All women who were singleton-born infants (between 1973 and 1992) AND mothers of first-born singleton infants (between 1990 and 2006). Preterms and subjects with missing data were excluded. Linked with Swedish Birth Registry to find birth characteristics. Investigators were blinded to birth data. | N/A | 4* | 64 | 103941 | 30.3 | 24.5 (4.6) | 87285 | 35.85 (31293) | 11.56 (10087) | ||||
25.7 (5.202) | 16656 | 45.63 (7601) | 17.84 (2972) | ||||||||||
25.9 (5.3) | 8148 | 46.76 (3810) | 18.65 (1520) | ||||||||||
25.5 (5.1) | 8508 | 44.56 (3791) | 17.07 (1452) | ||||||||||
PC; USA; A convenience sample of women with uncomplicated pregnancies were recruited in Gundersen Clinic, Ltd. and Lutheran Hospital in La Crosse, Wisconsin, from April 1, 1989 to March 30. Exclusion criteria included gestational diabetes, pre-eclampsia, cervical incompetence, or other conditions that would possibly lead to pre-term deliveries. Children from these mothers were followed up to adulthood. | N/A | 4* | 54 | 422 | 18–20 | 24.64 (5.64) | 174 | 30.5 (53) | 15.5 (27) | ||||
27.93 (9.4) | 11 | 36.4 (4) | 36.4 (4) | ||||||||||
23.94 (4.8) | 216 | 29.2 (63) | 11.1 (24) | ||||||||||
26.93 (8.67) | 21 | 42.9 (9) | 28.6 (6) | ||||||||||
24.25 (5.2) | 390 | 29.7 (116) | 13.1 (51) | ||||||||||
27.27 (8.79) | 32 | 40.6 (13) | 31.3 (10) | ||||||||||
Record Linkage Study; Denmark; men born as singletons between 1977 and 1983 in Northern Denmark, who presented for compulsory medical evaluation for conscription aged 18–20. Conscription records were linked to the Danish Medical Birth Registry. Investigators were blinded to birth data. | N/A | 5* | 100 | 21,051 | 18–20 | 23.6 (3.92) | 18,913 | 18.04 (3412) | 6.37 (1205) | ||||
24.1 (4.22) | 2138 | 18.52 (396) | 8.98 (192) | ||||||||||
23.9 (4.47) | 344 | 16.57 (57) | 8.72 (30) | ||||||||||
23.8 (4.42) | 189 | 18.51 (35) | 8.99 (17) |
Data from 12 studies
Forest Plot showing the pooled gender, unadjusted mean BMI difference in adult offspring by mode of delivery.
Forest Plot showing the pooled gender, unadjusted OR for incidence of overweight in adult offspring, by mode of delivery.
Forest Plot showing the pooled gender, unadjusted OR for incidence of obesity in adult offspring by mode of delivery.
Meta-regression showed the observed gender difference for BMI was not statistically significant (
Subgroup analyses of CS against VD for all studies reporting type of CS showed similar effect-sizes to the overall analysis (mean BMI difference: 0·43 kg·m−2 [0·02, 0·85] (
Data from Cnattingius
However, data from Svensson
Twenty-five percent of participants in Mi
Mean offspring age at BMI measurement ranged from 18 to 69·6 years in the primary pooled gender analysis. Meta-regression showed a borderline significant decrease in mean difference in BMI with increasing mean offspring age, 0·023 kg·m−2 ([0·0008, 0·046],
Visually the funnel plots (
A summary of all meta-analysis results is presented in
Number of subjects | Random effects | Heterogeneity | Fixed effects | ||||||
Comparison | CS | VD | Pooled result [95% CI] | p-value | I2 | p-value | 95% Prediction interval | Pooled result (95% CI) | p-value |
2995 | 34803 | 0·44 [0·17, 0·72] | 0·002 | 39% | 0.08 | [−0·27, 1·15] | 0·48 [0·30, 0·66] | 0·00001 | |
2995 | 34803 | 1·26 [1·16, 1·38] | 0·00001 | 0% | 0·92 | [1.14, 1.40] | NA | NA | |
2975 | 34647 | 1·22[1·05, 1·42] | 0.01 | 22% | 0.24 | [0.92, 1.92] | 1.21 [1.07, 1.36] | 0·002 | |
3791 | 36438 | 0·42 [0·18, 0·67] | 0·0008 | 40% | 0.07 | [−0.19, 1.03] | 0.49 [0.34, 0.63] | 0.00001 | |
3791 | 36438 | 1·21 [1·02, 1·43] | 0·03 | 56% | 0.0007 | [0.75, 1.95] | NA | NA | |
3776 | 36378 | 1·33 [1·16, 1·53] | 0·0001 | 8% | 0.36 | [1.01, 1.75] | 1.34 [1.19, 1.50] | 0.00001 | |
18134 | 104935 | 0·72 [0·27, 1·18] | 0·002 | 74% | 0·00001 | [−0.66, 2.10] | NA | NA | |
18134 | 104935 | 1·28 [1·12, 1·47] | 0·0004 | 44% | 0.05 | [0.90, 1.82] | NA | NA | |
18126 | 104839 | 1·30 [1·05, 1·62] | 0·02 | 66% | 0·01 | [0.74, 1.19] | NA | NA | |
281 | 17662 | 0·48 [−0·08, 1·04] | 0·09 | 7% | 0·36 | [−0.89, 1.85] | 0.49 [−0.04, 1.03] | 0.07 | |
281 | 17662 | 1·21[0·95, 1·53] | 0·12 | 0% | 0·72 | [0.72,2.03] | NA | NA | |
270 | 17506 | 1·26 [0·78, 2·05] | 0·35 | 39% | 0·20 | [0.01, 127] | 1.29 [0.93, 1.78] | 0.12 | |
261 | 17662 | 0·32 [−0·21, 0·85] | 0·24 | 0% | 0·47 | [−0.84, 1.48] | NA | NA | |
261 | 17662 | 1·20 [0·93, 1·55] | 0·15 | 0% | 0·85 | [0.68, 2.10] | NA | NA | |
252 | 17506 | 1·13 [0·80, 1·59] | 0·50 | 0% | 0·82 | [0.12, 10.3] | NA | NA | |
542 | 17662 | 0·43 [0·02, 0·85] | 0·04 | 10% | 0.34 | [−0.67,1.53] | 0.44 [0.05, 0.82] | 0.03 | |
542 | 17662 | 1·20 [1·01, 1.43] | 0·04 | 0% | 0.98 | [0.82,1.76] | NA | NA | |
522 | 17506 | 1·21 [0·96; 1·54] | 0·11 | 0% | 0.38 | [0.25,5.78] | NA | NA | |
2195 | 16817 | 0·69 [0·36, 1·02] | 0.0001 | 36% | 0.18 | [−0.20,1.58] | 0.64 [0.42, 0.86] | 0.00001 | |
2195 | 16817 | 1·33 [1·20, 1·47] | 0·00001 | 0% | 0.91 | [1.13,1.56] | NA | NA | |
2175 | 16661 | 1.27 [1.03, 1.57] | 0.02 | 39% | 0.18 | [0.79,2.02] | 1.27 [1.09, 1.48] | 0.002 |
In this large systematic review and meta-analysis investigating the association between mode of delivery and adult BMI, overweight and obesity, we found an average increase in BMI of almost 0·5 kg·m−2 in subjects delivered by CS compared to VD, and an increased odds of overweight and obesity >20%; these findings are consistent across sexes. We found some indication that studies of younger populations showed greater effect-size.
Given that rate of CS has changed over time; driven partially by changing clinical practice and by the rise of the maternal choice CS, it is possible that the impact of CS on outcomes may be different dependent on the birth year of the cohort. Studies with a higher background CS rate, and studies where participants were born after 1975, have higher effect sizes and narrower confidence intervals (
There are strengths and limitations to our systematic review and meta-analysis. A key strength is the large population of 142,702 subjects, from ten countries, spanning four continents; a large part of which have not been previously published in this form. Pre-registration of an
Despite these strengths, there were limitations. Funnel plot asymmetry suggested the possibility of reporting bias, but Egger's Test was statistically non-significant and as the outlying studies are in the opposite direction to the pooled results, any bias present probably attenuates rather than accentuates the findings. There was no evidence of a statistically significant gender difference in effect. Furthermore, gender-specific analyses showed higher levels of heterogeneity than the main results, possibly because the population size for each study is halved so the study estimates are less precise.
Collection of mode of delivery data in cohort studies was poor; these data were recorded in only 16 studies out of 33 identified (<15% of the total population of potential studies identified). Also, considerable aggregate data were unavailable for inclusion: 16 studies (a combined population >260,000 subjects) were identified in which relevant data had been collected, yet these were only available from 13, (142,280 subjects). A lack of patient level data prevented adjustment for confounders.
Findings from our meta-analysis are consistent with findings from other investigators. Shortly after completing our review, Li
Goldani
A key consideration is whether the associations we have identified between mode of delivery and offspring outcome are causal or reflect confounding. Plausible causal factors that might lead from CS to greater risk of obesity include differences in offspring microbiome as this differs between CS and VD neonates
To our best knowledge there have been no previous attempts to evaluate the effect of exposure to labour on later-life BMI, by examining outcomes in in-labour-CS and pre-labour-CS. When data were stratified by type of CS, we found no significant differences in adult BMI, overweight and obesity. However, these data were provided in only 5 studies, and the number of CS-delivered participants included in these subgroup analyses was <10% of the total number of CS deliveries in the main meta-analysis. Given the lack of power we do not feel that reliable conclusions can be drawn regarding the possibility of an impact of exposure to labour on study outcomes.
Our meta-analysis shows reduced association between CS and later-life BMI in older subjects. Exposure to post-natal obesogenic environmental factors increases with age, and may mask the association between CS and increased BMI in later life. However, as overweight and obesity track across time and are amplified in an obesogenic environment it seems unlikely that any impact of CS is truly attenuated with age. Hence we consider this is most likely due to the rising prevalence of obesity driving a larger effect-size in younger populations, although, longitudinal follow-up of these cohorts will be necessary to establish this with certainty.
In 2010 the WHO identified increased short-term adverse effects to mother and baby
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(TIF)
(DOC)
(PDF)
We thank all of the authors listed in table 1 who kindly supplied additional unpublished data for inclusion in this meta-analysis. We also thank the UK Data Archive and the Economic and Social Data Service for supplying data from the 1970 British Cohort Study and the 1958 National Child Development Study. These surveys are conducted by the Centre for Longitudinal Studies at the Institute for Education.