The Edmonton Obesity Staging System and pregnancy outcomes in women with overweight or obesity: A secondary analysis of a randomized controlled trial

Summary The Edmonton Obesity Staging System (EOSS) is a proposed clinical practice tool to determine obesity severity. In a secondary analysis of the Pregnancy Exercise and Research Study (PEARS) (a mobile‐health‐supported lifestyle intervention among pregnant women with body mass index [BMI] ≥25 kg/m2), we apply the EOSS and explore relationships with pregnancy outcomes. In early (14–16 weeks) and late (28 weeks) pregnancy, fasting lipids and glucose were measured, blood pressure was extracted from medical records and maternal well‐being was assessed using the WHO‐5 Well‐being Index. Pearson's correlations, chi‐square statistics and multiple logistic regression were used to identify relationships. One‐way analysis of variance was used to compare groups. Pregnant women (n = 348) were mean (SD) age 32.44 (4.39) years and median (interquartile range) BMI 28.0 (26.57, 29.88) kg/m2. Using metabolic criteria only, 81.9% and 98.9% had raised EOSS scores in early and late pregnancy. From early to late pregnancy, EOSS scores increased by 60.1%. Of these, 10.5% experienced a 2‐point increase, moving from stage 0 to stage 2. There was a potential relationship between early EOSS and large for gestational age (χ 2 = 6.42, df (2), p = .04), although significance was lost when controlled for confounders (p = .223) and multiple testing. Most women with BMI ≥25 kg/m2 had raised EOSS scores, limiting the clinical utility of the tool.


| INTRODUCTION
Obesity is recognized by the World Obesity Federation, the World Health Organisation and others as a chronic, progressive, relapsing disease. 1 Potential obesity can be identified using body mass index (BMI) ≥30 kg/m 2 . The severity of obesity can be described using BMI classes, namely class 1 or low-risk (30-34.9), class 2 or moderate risk (35-39.9) and class 3 or high-risk obesity (≥40 kg/m 2 ). 2 Pre-pregnancy overweight or obesity is common, affecting up to 42% of women in the United States. 3 In pregnancy, a raised BMI is associated with increased risk of adverse maternal and child outcomes. [4][5][6] Women with obesity require appropriate support and management to reduce the risk of complications in pregnancy and beyond. 7 While useful on a population basis, BMI does not provide insight into body composition, an important predictor of health outcomes, or the impact of excess adiposity on markers of health. 8,9 The American Association of Clinical Endocrinologists recommends the use of complication-based schema to inform obesity management. 10,11 The Edmonton Obesity Staging System (EOSS) is intended to provide clinically relevant insight into health-related risk for those with obesity. 12 It involves the classification of those with obesity into distinct groups based on their medical, psychological and functional health status. Stage 0 is given when the individual has no signs of obesity-related risk, stage 1 is given when there are subclinical risk factors and stages 2-4 are given in the presence of established obesity-related comorbidities. The tool was compared by its developers to the 'tumour, node, metastasis' system in oncology medicine. 12,13 Outside of pregnancy, higher EOSS scores have been associated with increased risk of postoperative complications and mortality after bariatric surgery. 14 A recent review of the evidence suggests that EOSS scores may better predict health service usage and treatment outcomes compared to BMI. 15 To date, only one study has been published using the EOSS in a pregnant population. In this study of women attending for induction of labour, the rate of caesarean delivery was higher in women with a BMI ≥25 kg/m 2 and stage 3 EOSS scores, compared to those in stage 2 and below. 16 In our study, we are the first to apply stage 0-2 EOSS to a general pregnancy cohort of women with BMI ≥25 kg/m 2 , recruited as part of a randomized controlled trial. The aim of this is to determine the severity and change in EOSS scores in women with overweight or obesity but otherwise healthy pregnancies. As a secondary aim, we explore potential relationships between EOSS and pregnancy outcomes. This will provide valuable information about the potential clinical utility of the scheme in identifying risk during pregnancy.  17,18 In brief, women were randomized to the intervention using computer-generated allocations in a ratio of 1:1 for usual care versus intervention. The intervention involved a single education session at the start of the study. This included advice on low glycaemic index diets, delivered by a research dietitian or nutritionist. It also included an exercise prescription of 30 min of physical activity for 5 days a week, given by an obstetrician. This information was re-enforced through a specifically designed smartphone application, fortnightly emails and two face-to-face study visits, all underpinned by behaviour change theory. Women in the control group were managed according to local and national guidelines and the advice they received may have varied in relation to nutrition, physical activity and gestational weight gain. 17 Women were screened for eligibility at their first antenatal visit by reviewing their patient charts. Women were eligible if they had a BMI between 25.0 and 39.9 kg/m 2 , singleton pregnancy, absence of previous GDM or any other medical illness requiring treatment. The primary outcome was an oral glucose tolerance test to diagnose GDM, according to the International Association of Diabetes in Pregnancy Study Groups criteria at 28-30 weeks' gestation. 17

| Study sample
This study uses data from the PEARS trial ( Figure 1). 18 A total of 18 (9 in the intervention and 9 in the control) of the 565 women included in the original trial did not attend their first study visit. 18 Of the remaining women, data were available for baseline blood pressure (n = 447), fasting lipids (n = 505) and fasting glucose (n = 485). The sample in this secondary analysis (n = 348) represents those women for whom data were available in early pregnancy on all the cardiometabolic markers needed for EOSS classification (n = 350), excluding one twin pregnancy (n = 2 mother-child pairs) in the intervention group. More information can be found in Figure 1 at the baseline visit, in which there were six options. The options ranged from 'completed no schooling' to 'completed third level'. Economic advantage was assessed using the Pobal Haase-Pratschke (HP Pobal) Deprivation Index, a neighbourhood deprivation score based on Irish census data that considers the relative advantage or disadvantage of the mothers' location of residence. 19,20 Women were classified as 'above' or 'below' average economic advantage using HP Pobal scores greater than or less than one. A composite measure for socioeconomic status was created based on maternal education (completed third-level education yes/no) and estimated economic advantage (economic advantage or disadvantage) using the method described by O'Brien et al. A four-level categorical variable was created ranging from 'economic disadvantage and did not complete third level', 'economic disadvantage and completed third level', 'economic advantage and did not complete third level' and 'economic advantage and completed third level'. 21 This variable was further classified into socioeconomically advantaged (yes/no). Women classified as 'economically disadvantaged and did not complete third level' were assigned a 'no' scoring while all other forms of advantage were classified as 'yes'. Blood pressure was extracted from antenatal medical records. Average systolic and diastolic values over the early (10-16 weeks gestation) and late (28 weeks) study periods were calculated to allow for EOSS classification.

| Maternal well-being
Maternal well-being was assessed in early (14-16 weeks) and late (28 weeks) pregnancy using the World Health Organisation (WHO)-5 Well-being Index. 23 Participants were asked to answer five distinct resulting in a total score that could range from 0 (lowest possible wellbeing) to 25 (highest possible well-being). The raw scores were multiplied by 4 to create a percentage. A score below <13 suggests reduced well-being while a score less than <7 suggests potential depression. [24][25][26]

| Application of the EOSS
Firstly, cardiometabolic markers were used to classify women into no risk (stage 0), some risk (stage 1) and higher risk (stage 2) ( Table 1).
Different cut-offs have been used for the individual cardiometabolic markers in the EOSS across a variety of studies, as detailed in the recent review by Atlantis et al. 15 In the only study to apply the EOSS in pregnancy to date, women were classified as having an EOSS score T A B L E 1 Individual criteria used for application of the Edmonton Obesity Staging System (EOSS) of 1 in the presence of subclinical cardiometabolic risk factors associated with obesity, namely borderline hypertension not requiring medical therapy, impaired glucose tolerance or abnormal gestational diabetes screen; however, no specific cut-offs were provided. 16 As pregnancy or female-specific cut-offs are yet to be established in the context of the EOSS, we selected the biochemical cut-offs used by Canning et al. 15,27 The criteria used to apply the EOSS in early pregnancy included fasting glucose, total cholesterol, LDL cholesterol, HDL cholesterol and triglyceride (Table 1). Higher EOSS scores indicate greater metabolic derangement. An EOSS score of 0 was assigned if all the cardiometabolic markers were within the cut-offs.
We did not apply high risk or end-stage criteria (EOSS stages 3 and 4) to this cohort as the presence of known conditions such as angina pectoris, myocardial infarction, heart failure or type 2 diabetes were exclusion criteria for trial. 27,28 Women with a well-being score <13 were given an EOSS score of 1 and those with <7 were given a score of 2 (Table 1). In late pregnancy, the same criteria were used except for fasting glucose as this was not available at 28 weeks because an oral glucose tolerance test was measured at this time instead. 18 As detailed above, GDM was identified at 28-30 weeks gestation, using the criteria of the International Association of Diabetes in Pregnancy Study after an oral glucose tolerance test. 17 Pregnancy-induced hypertension or pre-eclampsia was identified from medical records. A diagnosis of GDM, pregnancy-induced hypertension or pre-eclampsia resulted in a late pregnancy stage 2 EOSS score, as described by Demsky et al. 16 We did not have data available on obesity-associated functional complications or other conditions such as kidney disease, so these were not included in the scoring.

| Outcomes
The primary outcome in this study is EOSS score in early and late pregnancy. Secondary outcomes included gestational weight gain, mode of delivery, birthweight, birth length, head circumference and a variety of categorical outcomes including pre-term delivery (<37 weeks gestation), small for gestational age (SGA) (birthweight <10th centile), large for gestational age (LGA) (birthweight >90th centile), macrosomia (birthweight >4000 g) and low birthweight (<2500 g). Data on neonatal outcomes were retrieved from medical records.

| Statistical analysis
Categorical variables are presented as number and frequency (%).
Continuous variables were assessed for normality through visual inspection of histograms, the Kolmogorov-Smirnov test for normality, and inspection of descriptive data including the mean, median and skewness. All non-normally distributed data were log 10 transformed for regression analysis. Continuous variables are presented as mean (standard deviation) if they are normally distributed or median and interquartile range (25th, 75th centile) for skewed data. Comparison statistics were generated through independent sample t-tests. 3 | RESULTS women with and without obesity. There were also no differences in early (p = .877) or late (p = .580) pregnancy when well-being was added to the criteria (Figure 2).

| Application of the EOSS in early and late pregnancy
Using the metabolic criteria only, 285 (81.9%) had a raised EOSS score (≥1) in early pregnancy, while 273 (98.9%) had an EOSS score ≥1 in late pregnancy. Of these, 113 (32.5%) had stage 2 EOSS scores in early pregnancy, while 244 (88.4%) had an EOSS score of 2 in late pregnancy (see Table 3). In early and late pregnancy, the most common factor resulting in a stage ≥1 score was a total cholesterol level >5.2 mmol/L. The next most common value resulting in raised EOSS classification was a reduced HDL cholesterol. Of those with available data, the majority had a high well-being score (>13) in early (n = 206/291, 70.8%) and late pregnancy (n = 1984/223, 82.5%).
A well-being score <7, which is suggestive of potential mental health disorder such as depression, was found in 8 (2.7%) women in early and 6 (2.7%) in late pregnancy.

| Pregnancy outcomes according to EOSS score
In un-adjusted, chi-square testing, data were suggestive of a relationship between early pregnancy metabolic EOSS scores ≥2 and incidence of LGA (χ 2 = 6.42, df (2), p = .04, q = 0.01) (  LGA (>90th centile) As a sub-analysis, the incidence of LGA births was compared in women with raised EOSS scores ≥1 (n = 30, 11.3%) versus EOSS scores of zero (n = 5, 8.8%) and no statistically significant difference was seen (χ 2 = 0.32, df (1) On sub-analysis, the potential relationship between EOSS and LGA appeared due to EOSS stage 2 scores; however, this relationship also did not survive adjustment for multiple comparisons. No relationships with pregnancy or birth outcomes were found for late pregnancy, using any criteria.

| Interpretation
Outside of pregnancy, the EOSS has shown some promise in predicting adverse outcomes such as increased mortality and risk of postoperative complications after bariatric surgery. 14 Using data from individuals with a BMI ≥25 kg/m 2 in the National Health and Human Nutrition Examination Surveys up to 2006, those with EOSS stage 2 had higher mortality compared to those with 0 or 1. 13 The impact of higher EOSS scores was also seen in the study by Kuk et al., in which they found an increased risk of mortality in data from 29 000 participants in the Aerobics Center Longitudinal Study, with EOSS stage 2 but not stage 1 obesity. 29 In some studies, the predictive value appears to be greater with increasing EOSS stages (i.e. stages 3 and 4). These are applied when the individual has advanced obesityassociated conditions such as established organ damage, significant psychopathology or functional impairments. 14  In our study, unadjusted analyses were suggestive of a relationship between raised EOSS scores and LGA. Sub-analysis suggested that this was driven by stage 2, but significance was lost when controlled for multiple testing. Categorization of women with raised EOSS scores was mostly driven by cholesterol. A recent observational study of over 500 pregnant women found an association between maternal lipids and LGA, independent of BMI and GDM. 31 Only one other study has explored the role of EOSS in pregnancy. Demsky et al.
applied the EOSS to 345 women attending for induction of labour.
They found the overall rate of caesarean delivery was 35.8%, 29.9%, 43.2%, and 90.5% for women assigned an EOSS category 0, 1, 2, and 3, respectively. 16 We did not find an association with caesarean delivery, but like our study, their data suggest the predictive value of EOSS is most evident at later EOSS stages. In the absence of data on advanced obesity-associated co-morbidities to allow for stage 3 or 4 classification, our study suggests EOSS stage 0-2 may not be useful in predicting outcome due to the large proportions of women classi- in their scoring. 33 We found that the main driver of higher EOSS score was total cholesterol and its' inclusion in the scoring system may have limited its' potential in stratifying risk of adverse outcomes. Potter and Nestel found cholesterol levels increased by 50% throughout gestation. 34 Given the high levels of cholesterol expected during pregnancy, future work could consider the use of alternative scoring tools, such as varying criteria that do not include total cholesterol. This includes the criteria for the metabolic syndrome or the cardiometabolic disease staging system. 10,28,35 This is further supported by evidence suggesting that high triglycerides and low-HDL increase LGA. 36 41 Darmady and Postle found cholesterol levels remained elevated up to 36 weeks post-partum. 39 The degree of change experienced may also influence associations with outcomes. In a prospective study of 575 women, Bever et al. found a 10 mg/dL increase in triglycerides from preconception to 28 weeks was associated with increased odds of LGA and a 10 mg/dL decrease was associated with reduced odds of SGA and LGA. 42 On this basis, more research on the extent of cardiometabolic changes experienced in pregnancy is warranted.
A recent systematic found that a low glycaemic index/load dietary pattern is associated with reduced fasting glucose, LDL cholesterol, Apo B triglycerides and systolic blood pressure, in adults with type two diabetes and a raised BMI. 43 Cha et al. found that adolescents with higher modified-EOSS scores reported lower diet quality that those with lower risk factors. In addition, those with a modified-EOSS score of zero reported greater percentage energy from protein consumption. 44,45 While previous studies found the PEARS intervention group had lower dietary glycaemic load and higher protein intake compared to the control group, we did not find a difference in EOSS scores in early to late pregnancy based on the study group. 18,46 Interventions to promote a more favourable cardiometabolic profile in women during preconception may therefore prove more efficacious in ensuring healthier levels during pregnancy and post-partum. 47-50

| Strengths and limitations
Our study takes a novel and pragmatic approach to exploring the cardiometabolic changes experienced in pregnancy by tracking women against clinically relevant cut-offs for dyslipidaemia, rather than comparing raw values. Pregnancy-specific cut-offs for cardiometabolic parameters have not yet been developed. 15,16 The American Heart Association, the American College of Cardiology and others advocate for the use of sex-specific considerations in cardiovascular risk assessments. 51 A relevant core outcome set is the Core Outcome Set for Studies on Obesity in Pregnant Patients, and this project highlighted a need for greater measurement of outcomes relating to emotional functioning. [52][53][54][55] Use of a validated and internationally relevant questionnaire to assess well-being allows for meaningful comparison with the literature and addresses gaps in EOSS studies and pregnancy obesity that did not consider mental health. 13 Use of the Benjamini-Hochberg adjustment for multiple comparisons adds statistical rigour to the data. 56 This study is not without limitations. Selection of women with sufficient criteria for EOSS classification may introduce selection bias. We did not have sufficient data to allow for application of the functional aspect of EOSS, or application of stage 3 or 4 EOSS.
Other studies in non-pregnant individuals have used functional limitation and activities of daily living in the functional assessment but interpretation of this may be difficult in pregnancy due to the associated functional decline throughout gestation, especially for those with obesity. 13,57,58 This was an exploratory analysis of data collected as part of a randomized controlled trial, and it is possible that our study did not have statistical power to find differences in pregnancy outcomes. Regardless, the spread of EOSS stages in our cohort does not support application in its current format for the purpose of treatment prioritization.

| CONCLUSION
Most women with overweight or obesity have raised EOSS scores in early pregnancy, and EOSS score did not predict pregnancy outcomes.