Our systematic review was carried out in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (Supplementary Material 1) [15]. A comprehensive search strategy was developed by two authors (LK, KAE). After discussion with specialist subject librarians, the agreed search terms were:

An extensive literature search was carried out by LK in March 2020 (and updated in February 2021 and February 2023) using Ovid MEDLINE, EMBASE and CINAHL Plus searching databases from inception. Searches were restricted to full-text, human studies reported in the English language.

Inclusion and exclusion criteria were developed to target population level studies that used administrative and routinely collected healthcare data. Studies were eligible for inclusion if (i) they presented obesity prevalence data for pregnant women; (ii) BMI was recorded pre-pregnancy (for cases of singleton or multiple births) or recorded in early pregnancy (for singleton pregnancies) [16] and (iii) the study used the WHO classification boundaries for overweight and obesity, or alternative boundaries for classification specific to race/ethnicity of the cohort studied. Studies were excluded if they had exclusion criteria relating to (i) age; (ii) race; (iii) maternal health status; (iv) birth outcome; (v) gestation at birth; or if they had (vi) reported fewer than one hundred cases; (vii) reported only livebirths; or (viii) required voluntary participation. Exclusion criteria (i) to (v) and (vii) were chosen as these characteristics are associated with BMI and have the potential to bias estimates of prevalence. The exclusion criterion of fewer than 100 cases was used as smaller studies may not reflect populations being studied. Studies that required voluntary participation were also excluded as women who are self-conscious of their BMI status may not volunteer or consent to studies which require this measurement. After removing duplicates, screening of all titles and abstracts was performed by two authors (LK & KAE March 2020 and February 2021; MMcG & LK February 2023), and full texts obtained for studies identified by all authors at this stage. Lastly, hand searching of full text papers was performed (LK) to identify any additional publications suitable for inclusion. Full text copies of the remaining articles were assessed for inclusion by three authors (LK, KAE, MMcG) with each receiving an allocation and all assessments being reviewed by the full team. Consensus decisions were made as a team on inclusion or exclusion where necessary. Additional studies were excluded if they had the same study population from the same curated data set as another included study.

A detailed data extraction form was developed (LK, KAE, MMcG) (Supplementary Material 2). In addition to inclusion and exclusion criteria, these forms required details of the country and region, participant characteristics, the BMI classification used, and the number and percentage of pregnant women recorded within each BMI category. Each author extracted data for a proportion of the included studies, with 10% of papers randomly selected for independent verification by one other author. Study authors were contacted via email for clarification of inconsistencies or missing data.

The Hoy Risk of Bias tool was used to assess the quality and validity of measured data, and the risk of bias of included studies (LK, KAE, MMcG) (Supplementary Material 3) [17]. This validated tool for assessing the risk of bias in prevalence studies is comprised of ten questions, assessing four domains of bias. Questions 1 to 4 focus on the external validity of studies (assessing selection bias and nonresponse bias); and questions 5 to 10 consider the internal validity of studies (evaluating measurement bias and analysis-related bias). If criteria were fulfilled and the answer to the question was ‘Yes’ (low risk), a score of 0 was assigned. If the criteria were not met (answering ‘No’ or ‘Unclear’ to the question, meaning high risk), a score of 1 was assigned. To ensure harmonisation between authors when assessing each criterion, additional guidance was developed to reflect the context of this systematic review of population-based studies reporting prevalence of maternal overweight and obesity (Supplementary Material 3). If the overall score was 0–3, 4–6 or 7–10, the risk of bias was classified respectively as ‘Low’, ‘Moderate’ or ‘High’, applying the same approach as previous studies [18, 19].

A comprehensive study database was designed (LK, MMcG) to capture study characteristics and reported prevalence of each BMI category. The data was entered by MMcG, with a random 10% sample checked for quality assurance by a second author (LK). If separate study populations were examined in one paper, each were added into the database separately, and noted as Study ID A, B, etc. to ensure there was no overlap between these study populations.

All data preparation and analysis was performed using R [20]. For each included study, the midpoint date was calculated. Prevalence was calculated as the percentage of total pregnancies classed as obese (BMI ≥30 kg/m²), and overweight/obese (BMI ≥25 kg/m²).

Random effects meta-analysis was conducted on untransformed prevalence data using the DerSimonian and Laird [21] method to summarise estimates and 95% confidence intervals for the pooled prevalence of obesity and overweight/obesity. The percentage of total variability due to between-study heterogeneity was estimated using the I² statistic, however this was expected to be large due to anticipated temporal and regional differences in maternal obesity from study cohorts spanning different time periods. In the first instance this was performed for all study cohorts reflecting global prevalence over the full time-period covered by the included studies. This process was then repeated for each region of the world and each decade to describe the prevalence in different geographical regions over time. The pooled prevalence for obesity and overweight/obesity were then visualised as bar charts with 95% confidence intervals.

The relationship between prevalence and time was visualised for (i) obesity, and (ii) overweight/obesity using scatterplots of individual study cohorts. For overall global trends, a linear fitted line with 95% confidence intervals was created from all study cohorts. To visually compare trends between different geographical regions, linear fitted lines were created for maternal obesity prevalence estimates grouped by area of the world.

Linear regression was then used to model the trends in global prevalence over time, first using the full dataset containing all study cohorts. The midpoint year of the study period for each included study was used as the independent variable, and prevalence of i) obesity and ii) overweight and obesity was used as the dependent variable. As a sensitivity analysis, further linear regression models were fitted using data from i) 1980 onwards, ii) 1990 onwards, iii) 2000 onwards and iv) 2010 onwards. To compare prevalence between areas of the world, linear models were then fitted using midpoint year of study and area of the world as independent variables, and Europe, as the category with the most studies, selected as the reference. The estimated differences in prevalence and 95% confidence interval were calculated for each model and p values considered statistically significant at ≤0.05.

The study protocol was registered with the Research Registry (UIN: reviewregistry998) on 23^rd September 2020.

After screening 11,635 papers, 94 studies met the inclusion criteria (PRISMA Flow Diagram, Figure 1). Nine of the included articles reported data from multiple cohorts [22-30] resulting in 121 study cohorts.

Overall, data from 49,009,168 pregnancies were included. Characteristics of included studies are presented in Table 1 (n = 94). Of the 121 study cohorts present in the included studies, the majority were European populations (n = 71), with 23 cohorts in North America, 12 cohorts in South America, ten from Australia and Oceania, and five from Asia. No studies were found representing populations from Africa. Included study cohorts represent the populations of 25 different countries, with all but two studies having taken place in high income countries [31, 32], according to the World Bank definition [33]. Fifty-one of the study cohorts were countrywide studies, using data from national databases, whilst 70 were regional studies which used data from databases that covered a smaller geographical region within a country, or in some cases a single hospital or group of hospitals. The years covered by the included studies span from 1959 to 2022. Midpoint dates of included studies ranged from 1962 to 2021. Seventy-four study cohorts included only women who delivered singleton births and 36 included both singleton and multiple births, whilst eleven were unclear as to the order of the included pregnancies. Maternal BMI was measured pre-pregnancy for women in 57 study cohorts, in early pregnancy for 38 cohorts, and a combination of both pre-pregnancy and early pregnancy measurement in five study populations. Timing of BMI measurement was unclear in 21 study cohorts. BMI was categorised according to the WHO classification in the majority of study cohorts (n=95); three studies adjusted BMI classification according to ethnicity. In the remaining study cohorts that did not use the WHO BMI classification or ethnicity-specific classification, only the boundary between underweight and healthy differed, and therefore did not affect the current analysis of prevalence of overweight and obesity.

Quality assessment of the 94 included studies is reported (Supplementary Material 4). Eighty studies (85.1%) were evaluated as having a low risk of bias and 14 studies (14.9%) were judged as moderate. The more common reasons for receiving a higher score were greater than 2% missing values reported or that weight or BMI was ascertained through self-report. Given that the assessed risk of bias of included studies was at highest moderate, this was not thought likely to affect the interpretation of our results.

The prevalence for each BMI category is presented separately for all included study cohorts in Table 2 and visualised via box plots in Figures 2, 3. Pooled prevalence, synthesised by area of the world and decade, is presented in Table 3 and visualised in Supplementary Materials 5, 6. Globally, prevalence of obesity (BMI ≥30 kg/m²) in pregnancy has more than tripled from 4.7% (95% CI 2.6–8.6) pre-1990 to 16.3% (95% CI: 15.1–17.5), or one in approximately six pregnancies, in the decade 2010 to 2019 (Figures 2 and Supplementary Material 5). Only two studies reported prevalence of obesity since 2020, both from North America, but these appear to show a continuation of the upward trend, with a pooled prevalence of obesity estimated as 24.5% (95% CI: 21.4–27.8), or one in every four pregnancies. Prevalence of combined overweight & obesity (BMI ≥25 kg/m²) in pregnancy has also increased globally, more than doubling from 20.6% (95% CI: 16.4-25.6) pre-1990 to 43.8% (95% CI: 42.2-45.4) in the decade 2010 to 2019 (Figures 3 and Supplementary Material 6). Asia did not have sufficient study cohorts to allow for a pooled estimate of prevalence to be estimated for each decade. Across the remaining areas, only North America and Europe had sufficient results to stratify by at least three decades (1990s, 2000s, 2010s). Pooled prevalence was possible for 2000s and 2010s for Australia & Oceania and South America. Across each area of the world, an upward trend similar to the global trend was observed, with North America demonstrating the highest prevalence in obesity and overweight/obesity in each decade. Percentage of total variability due to between-study heterogeneity, as measured by the I² statistic, was large in all meta-analyses, ranging from 98.0 to 100%.

Figure 4 further demonstrates the upward trend of prevalence of both obesity, and overweight/obesity in pregnancy. This trend is apparent globally (Figure 4a,b) and for each area of the world (Figure 4c,d). North America demonstrates some of the highest prevalence rates for obesity and overweight/obesity, with Australia and Oceania appearing to increase at a greater rate and reaching similar rates as North America by 2010.

Linear regression performed on all studies reporting (i) obesity prevalence (n = 120 study cohorts) and (ii) combined overweight and obesity prevalence (n = 110 study cohorts), both demonstrated a statistically significant increase over time (p < 0.001), with prevalence increasing by (i) 0.34% (95% CI: 0.23–0.45) and (ii) 0.64% (95% CI: 0.44–0.84), respectively, with each year.

Using Europe as a reference and adjusting for midpoint year of study, obesity prevalence was compared across areas of the world. A significantly higher prevalence was observed for study cohorts from North America (6.90%, p < 0.001) and Australia & Oceania (5.59%, p = 0.002). Analysis of combined overweight and obesity rates demonstrated a significantly greater prevalence in North America (9.83%, p < 0.001).

The scale of this systematic review, along with the strict inclusion and exclusion criteria applied, provides a comprehensive insight into the current state of global and continental maternal overweight and obesity prevalence. Our findings demonstrate a significant increase in maternal obesity and overweight/obesity prevalence over time, with our linear regression analysis estimating a 0.34% and 0.64% increase in obesity prevalence and combined overweight and obesity prevalence, respectively, with every passing year. Pooled analysis from the most recent studies (post 2010) indicates that globally, one in six pregnancies are now complicated by obesity. In North America, since 2020, one in four pregnancies are complicated by obesity.

In comparison to a previous review of pre-pregnancy BMI from Martínez-Hortelano et al, global obesity prevalence found in our study was slightly lower for the full time period (13.6% v. 16.3%) [9]. This may be due to inclusion of earlier studies, with lower prevalence, in our study. However, when results were stratified by continent, our pooled prevalence of maternal obesity was higher, and particularly striking when the analysis was limited to the decade 2010–2019: North America (23.1% vs. 17.6%), Europe (14.6% vs. 9.1%), Australia and Oceania (20.6% vs. 11.3%). Our findings likely reflect the more recent studies included in our review, and our strict inclusion/exclusion criteria which ensured that results were reflective of population-based prevalence and not biased by including studies that may have unintentionally excluded pregnant women with higher BMI’s.

The estimated global trend of increasing overweight and obesity prevalence in pregnant women mirrors that seen in the general female population [114]. It is estimated that mean BMI has increased by 0.59 kg/m² per decade from 1975 to 2014 [114], and 0.5 kg/m² per decade from 1980 to 2008 [115]. In 2014, obesity prevalence in the general female population was estimated to be 14.9% [114], reflected by WHO estimates in 2016 (15%) [2]. However, a much higher prevalence has been estimated for the general female population in high income countries; 25.3% in 2014, and 26.2% in 2016 [2]. Our linear regression analysis suggests that global obesity prevalence in the pregnant population was 17.8% in 2014, and 18.4% in 2016. Using our model, the current global prevalence of maternal obesity is 20.9% (95% CI 18.6 to 23.1%), and is projected to reach 23.3% (95% CI 20.3 to 26.2%) by 2030.

The highest prevalence of maternal obesity is found in North America and Australia/Oceania, who also demonstrate the highest prevalence of combined overweight/obesity. These findings are also recognised in the general female population; with national bodies including, the National Academy of Medicine (USA) [116], Health Canada [117], the National Health and Medical Research Council (Australia) [118] and the National Institute for Health and Care Excellence (UK) [6], recognising the urgent need for strategies to support weight management, including pre-conception care [119]. A steeper increase in obesity and overweight/obesity prevalence was seen in studies carried out in Australia/Oceania. Our findings are similar to those reported by Finucane et al., who noted greatest increase in BMI in the general population of females in Australia/Oceania (1980–2008) [115].

Comparing prevalence of maternal overweight and obesity between areas of the world is challenging. Studies included in this review ranged significantly in the type of data used (e.g. routinely collected healthcare data versus vital statistics) and scale of the population covered (e.g. hospital database versus region or countrywide datasets), making it difficult to assess comparable studies. In Nordic countries, reporting of all births to national organisations is required [120]. As a result, the number of population wide studies included in our review from Sweden (n = 12) was higher than other countries, which may have skewed the results for Europe. Many studies carried out in USA were excluded as much of the published data currently relies on information obtained from birth certificates, meaning stillbirths are often excluded. As a result, North America contributed only three countrywide studies [99, 107, 111]. A further challenge was the ability to reliably compare trends on a global scale, as the number of studies per geographical region varied greatly. As previously stated, no studies undertaken in Africa met the inclusion criteria, mainly due to the lack of routinely collected data from retrospective studies. The lack of data available for analysis may reflect local constraints in collecting healthcare information, including lack of digitised healthcare records, late or non-presentation of women for antenatal care, and lower levels of tertiary education resulting in poor self-reporting of pre-pregnancy weight [121]. Of the five Asian studies that were included, only two used the WHO recommended BMI classification specific to the Asian population [34, 36] (BMI of 23–24.9 kg/m² considered overweight and ≥25 kg/m² obese) [122]. This may have resulted in possible misclassification of at-risk overweight and obese Asian women. Furthermore, pre-pregnancy BMI may be influenced by country income [114]; only two studies included in this review took place outside a high-income country [31, 32], due to limited publication of routinely collected data from low- and middle-income countries. Again, this publication bias may reflect the paucity of standardised national databases and digitised clinical records [123]. Maternal BMI is also linked to ethnicity [124, 125] and parity [126] however included studies had insufficient data to enable us to explore this association further.

There is also a paucity of guidance in the field of evidence synthesis of prevalence, which has been previously recognised [127]. This has led to variability in approaches to the methods, risk of bias assessment, data synthesis and reporting of such reviews. We echo the Prevalence Estimates Review-Systematic Review Methodology Group (PERSyst) call to action to develop guidance and reporting standards for these types of review [127].

The use of strict inclusion and exclusion criteria ensured high quality studies of population level data representative of all births were included. In addition, use of an appropriate risk of bias tool for studies investigating prevalence [17] supports the interpretation and generalisation of results in this review. No studies were judged to have a high risk of bias, ensuring that our results were not unduly influenced by selection bias, sampling bias, non-response bias or data-collection bias, and that they were suitably representative of population-level data.

A key limitation is that our review mainly included studies from countries categorised as high income at the time of review, this may have been in part due to the exclusion of reports not written in English. Translation of non-English language reports was not within the scope of this review due to funding constraints. A further limitation is that the income status of countries where specific studies took place was not assessed at the time of their study period but instead when the review was undertaken, as a result some income status’ may have differed. Additionally, several study populations were included that had overlapping data, for example, studies using national data from the Swedish Medical Birth Registry. Instead of withdrawing studies and risking exclusion of valid data for certain years, we attempted to ameliorate these effects by calculating and reporting midpoint year of studies. Further MeSH terms, such as “antenatal” or “perinatal$” may have yielded additional studies. The review did not evaluate underweight or undernutrition, which we acknowledge can also be linked to adverse pregnancy outcomes [128, 129].

Our findings provide key epidemiological and clinical insights. This review demonstrates an increasing trend in overweight and obesity prevalence in pregnant women across the world, with particularly high prevalence in high income countries, notably in North America, Australia/Oceania and Europe.

To improve accurate surveillance of maternal BMI, key public health stakeholders must focus their efforts on provision of better data informatics to assist in cohesive collection of healthcare data, particularly in low- and middle- income countries. A robust evidence base is required to allow comparable analysis, with this review pointing to the need for further research to be focused in Asia, South America, Africa and in low- and middle-income countries worldwide, using ethnically appropriate BMI classification boundaries. While BMI is a useful tool for population level surveillance, it has limitations at an individual level [130, 131], and there is a need for development of useful tools to aid clinical assessment of personal risk [132]. One example is the Edmonton Obesity Staging System [130, 131]. Novel tools such as this must also be evaluated for use at both an individual and population level in women of child-bearing age.

The increasing trend towards maternal obesity in pregnancy demonstrated by this review is set to provide further challenges to provision of obstetric care. Interpregnancy weight gain, independent of parity and baseline BMI, increases adverse obstetric and neonatal outcomes, including gestational diabetes, hypertensive disorders of pregnancy, increased need for operative delivery, fetal macrosomia and stillbirth [43, 133, 134]. Women with increased BMI in pregnancy are also more likely to have pre-existing multimorbidity (two or more long-term physical or mental health conditions) which is likely to further complicate their pregnancy and postpartum course [135].

Effective strategies to support optimisation of weight before conception are essential to reduce maternal, fetal and neonatal risks and to help women establish a healthy weight trajectory beyond pregnancy.