The Effects of Pre-Pregnancy Body Mass Index and Gestational Weight Gain on the Risk of Preeclampsia at a Tertiary Referral Hospital, Northern Tanzania


 Background: Pre/eclampsia and other hypertensive disorders of pregnancy contributed to 18% of the maternal mortality reported in Northern Tanzanian. There is increasing prevalence of obesity in Tanzania which is related to excessive weight gain in pregnancy. Both high BMI and excessive gestation weight gain are identified to increase risk of PE and subtypes, however this is still inconclusive and little is known about the joint effect of pre-pregnancy BMI and GWG on risk of PE and its subtypes in Africa. We evaluated the independent and joint effects of pre-pregnancy BMI and GWG on the risk of pre-eclampsia and its subtypes among women who delivered at Kilimanjaro Christian Medical Center (KCMC) from October 2018 to May 2019, Northern Tanzania.Methods: We performed a retrospective birth cohort study from October 2018 to May 2019 at KCMC, Tanzania. Pre-pregnancy BMI was categorized using WHO categories into Underweight (˂ 18.5kg/m²), Normal weight (18.5-24.9kg/m²), Overweight (25-29.9kg/m²) and Obese (≥ 30kg/m²). Gestational Weight Gain (GWG) was categorized using the 2009 Institute of Medicine (IOM) guidelines into Inadequate, Adequate and Excessive weight gain in pregnancy. Multinomial logistic regression analysis was used to adjust for confounders using relative risk, 95% confidence interval for the risk ratios (RR) that did not cross 1 and p<0.05 were regarded statistically significant.Results: Among the 1309 women analysed, 5.3% were Underweight, 51.1% Normal weight, 26.9% Overweight and 16.7% were Obese. About 43.5% had excessive GWG. Women with PE were 9.5%. Both obesity and excessive GWG independently increased risk of PE with adjusted RR=2.42, 95%CI: 1.48-3.96 and RR=1.77, 95%CI: 1.16-2.69 when compared to normal BMI and adequate GWG respectively. Jointly, Obesity and Excessive GWG had the highest risk of PE (ARR=4.95, 95%CI: 2.21-11.10). The increased risk was similar for Mild PE (MPE), Severe PE or eclampsia (SP/E) and Late Onset PE (LOPE). No association was found for Early Onset PE (EOPE).Conclusion: Pre-pregnancy Obesity and Excessive GWG independently and jointly increases risk for PE and the risk varies by PE subtype.


Background
Preeclampsia (PE) is a serious and life-threatening complication of pregnancy which is characterized by a new onset hypertension and either proteinuria or end organ damage occurring after 20 weeks of gestation. Eclampsia (E) is the development of convulsions in a woman with PE. This pregnancy syndrome is associated with a large burden of maternal and foetal morbidity and mortality with substantial contributions to prematurity of the foetus and long term maternal cardiovascular and renal disease (1). Preeclampsia (PE) is estimated to affect 2-10% of pregnancies worldwide (2). PE and eclampsia are estimated to cause over 50,000 annual maternal deaths globally (3). At a tertiary hospital in Northern Tanzania, PE and other hypertensive disorders of pregnancy contributed to 18% of maternal deaths. (4). After obstetric haemorrhage, PE is the second leading cause of pregnancy related intensive care unit admissions (5). In many African countries, it continues to be a major public health challenge being a leading cause of prematurity and foetal growth restriction. It has a perinatal mortality rate ve times higher than that among babies born to healthy mothers (6).
There is wide geographic and regional variation in the incidence of PE with the USA reporting an estimated incidence of 5.9% (1). The World Health Organization (WHO) estimates the incidence of PE to be seven times higher in developing countries (2.8% of live births) than in developed countries (0.4% of live births), with Africa reporting ranges from 1.8%-16.7%. Nigeria reported the highest incidence at 16.7% (2). The incidence rates from other African countries such as South Africa, Egypt, Tanzania and Ethiopia vary from 1.8% -7.1% (2). The incidence for Tanzania was reported at 1.7% (7).
Women with PE in one pregnancy have a substantially increased risk of recurrence in subsequent pregnancies with an absolute recurrence risk of 25% reported in Northern Tanzania (8).The underlying pathogenesis of PE remains unclear. However it is characterized by defective placentation, placental ischaemia, abnormal spiral artery remodeling and oxidative stress at the maternal foetal interface. Other factors include angiogenic imbalance in the maternal circulation with resultant endothelial and end -organ damage (9). The disease can be understood in terms of both placental and maternal dysfunctions. Several studies have suggested PE as a heterogeneous disease with two stages, namely, Early Onset or placental disease (EOPE) and Late Onset or maternal disease (LOPE) with different epidemiology, clinical presentation and associated morbidity (10).
Early Onset Preeclampsia (EOPE) occurs before 34 weeks of gestation and it involves abnormal placentation with shallow trophoblastic invasion, insu cient spiral artery remodeling leading to reduced placental perfusion. It is associated with increased risk for intrauterine growth restriction and a 3-25 fold increased risk of severe maternal complications. These include abruptio placenta, disseminated intravascular coagulation, pulmonary oedema and aspiration pneumonia (1,6,11) and a 20 fold higher risk of maternal mortality (6).
Late Onset PE or maternal disease occurs at or after 34 weeks of gestation and it involves maternal systemic in ammation and oxidative stress which results in vascular endothelial dysfunction leading to multi organ failure.
It is frequently associated with maternal obesity, larger placental volume, normal foetal growth, normal birth weight and more favorable maternal and neonatal outcomes.
Preeclampsia (PE) is also subcategorized into severe (SP/E) and mild (MPE) disease. Severe P/E has a blood pressure ≥ 160/90 mmHg with or without proteinuria, and or evidence of end organ damage, such as thrombocytopenia, impaired liver function with persistent severe right upper quadrant pain or epigastric pain.
Others are new onset renal failure, pulmonary oedema or new onset cerebral or visual disturbances. Mild PE has a blood pressure 140/90 ≥ BP 160/110 mmHg, and no features of severe disease (12).
A wide range of factors are associated with PE such as parity, placental factors, multifoetal gestation and excessive weight gain during pregnancy. Others are some pre-pregnancy maternal factors like age, race, prepregnancy overweight and obesity, diabetes mellitus and chronic hypertension (6).
Pre-pregnancy Body Mass Index (BMI) and gestational weight gain (GWG) are two modi able risk factors implicated in the development of PE. Obese women who become pregnant and their foetuses are predisposed to various adverse pregnancy related complications. Infants and later adults of obese mothers have correspondingly increased rates of morbidity, mortality and obesity (13). In Tanzania, the prevalence of obesity among women of reproductive age has increased progressively from 6% in 2010 to 10% in 2015 (14) with Kilimanjaro region reporting an increase in obesity from 10.8% in 2010 to 19.7% in 2015 (14).
The association between GWG and the risk of PE is still inconclusive with some studies reporting a positive association (6) while others found no association (15).
Studies which evaluated the association between BMI, GWG and the risk of PE by different subtypes in low resource settings like ours are limited. More so, there are very few studies investigating the joint effect of prepregnancy BMI and GWG on risk of PE and its subtypes.
The current study therefore, aimed to evaluate the independent association of pre-pregnancy BMI, GWG and their combined effect on the risk of developing PE and its subtypes in a low resource setting.

Study design, setting and period
We conducted a hospital based retrospective cohort study from October 2018 to May 2019 at Kilimanjaro Christian Medical Centre (KCMC), a zonal referral hospital in Northern Tanzania located in Moshi urban district.
The hospital caters for about 15 million people and it has over 3,000 deliveries annually. It also receives patients from the surrounding districts in the northern zone including Kilimanjaro, Tanga, Arusha, Manyara and neighboring districts of Kenya. Obstetric assessments, vital signs check, weight measurements at booking and every Antenatal clinic (ANC) visit thereafter, height measurement at booking, screening for pregnancy complications, health education on nutrition and pregnancy danger signs, counseling and testing for HIV are done regularly.

Study population and eligibility
Eligible study participants were all women who delivered at KCMC between October 2018 and May 2019 at gestational age ≥ 28 weeks. All women with history of chronic hypertension, multiple pregnancies, rst ANC visit gestational age ≥ 20 weeks, missing information on the following, blood pressure measurements,, proteinuria, admission weight, rst ANC visit weight (kg) and height (m) measurements were excluded from the study.

Data collection procedure
The hospital registration numbers from the birth register book were used to trace patient admission les and antenatal clinic records from medical records or post-natal ward. Patient les were then reviewed and all important variables were obtained. These variables were socio-demographic data, anthropometric measurements of the mother, conditions, diseases and complications during the current pregnancy, maternal obstetric history, and medical history of the mother. WHO BMI guidelines were used to categorize participants' BMI into Underweight ( 18.5), Normal weight (18.5-24.9), Overweight (25.0-29.9) and Obese (≥ 30) categories (16). Prepregnancy BMI for this study was calculated using the rst ANC visit weight (kg) divided by the square of height (m²) when the gestational age recorded in the ANC record at rst visit was ≤ 20 weeks (The last normal menstrual period was used to calculate the gestational age at rst ANC visit). The pre-pregnancy weight (kg) was the recorded ANC weight at rst visit when the gestational age was ≤ 20 weeks.
The 2009 Institute of Medicine (IOM) recommendations for gestational weight gain were used to categorize GWG into Adequate, Inadequate and Excessive categories based on the respective BMI categories (17). Gestational Weight Gain was calculated by subtracting the pre-pregnancy weight (kg) from the weight (kg) at delivery. The ranges for underweight, normal weight, Overweight and obese were 12.5-18.0 kg, 11.5-16.0 kg, 7.0-11.5 kg and 5.0-9.1 kg respectively. BMI and GWG with their respective categories were the main independent variables in this study. PE and its subtypes were the outcome of interest in this study.
A total of 2,305 deliveries were recorded in the delivery book between October 2018 and May 2019. There were 34 missing patient les that could not be traced. The remaining 2,271 patient les were therefore reviewed. Of these, 962 patient les were excluded due to exclusion criteria (619 lacking records for BMI, blood pressure and proteinuria; 58 history of chronic hypertension; 65 multiple pregnancies and 220 les due to having the gestational age at rst booking recorded in the antenatal record more than 20 weeks of gestation). We ended up with 1,309 singleton deliveries for nal analysis. A total of 125 PE/E participants with their respective BMIs and GWG and 1,184 with no PE/E participants with their respective BMIs and GWG were analyzed (Fig. 1). The 125 PE participants were sub categorized into MPE (45), SPE (80), EOPE (34) and LOPE (91) subtypes (Fig. 1).
Pre-eclampsia was identi ed as BP ≥ 140/90 mmHg measured on two separate occasions at least four hours apart with ≥ 1 + proteinuria on dipstix occurring after 20 weeks of gestation. Mild preeclampsia was identi ed when BP ≥ 140/90 mmHg and < 160/110 mmHg, proteinuria (≥ 1 + and < 2 + on dipstix) without symptoms of severity. Severe pre/eclampsia was identi ed as BP ≥ 160/110 mmHg and proteinuria (≥ 2 + on dipstix) with additional symptoms of severity i.e. headache, blurred vision, epigastric pain, decreased urine output and convulsions. Early onset preeclampsia (EOPE) was identi ed with onset of PE before 34 weeks of gestation and late onset preeclampsia (LOPE) was identi ed with onset of PE at 34 weeks or more of gestation. PE and subtypes where the outcome variables in the study.

Data analysis
Data was analyzed using Stata Version 13.0 after adequate data cleaning process. Numerical data was summarized using mean and standard deviations while categorical variables were summarized using frequency and proportions. Chi-square and Fisher's tests were used to compare the maternal characteristics by BMI categories and PE with their corresponding p-values. P value < 0.05 was regarded as statistically signi cant.
Multinomial logistic regression analysis was employed to compute risk ratios (RR) for PE and its subtypes by BMI and GWG. Both crude and adjusted RR was employed to identify signi cant association between exposures and the outcome of interest (PE and its subtypes). All noted risk factors for PE in this study were adjusted to control possible confounders and effect modi ers. 95% con dence interval was used to identify the signi cant risk of PE and its subtypes. Further analysis was done for the joint effects of BMI and GWG on the risk of PE and subtype.
Adjusted RR for PE and its subtypes and the probability for interactions between BMI and GWG was estimated using test of homogeneity. The ndings of the study were presented using tables and gure.

Results
Out of the 1,309 singleton birth deliveries analyzed, the majority of participants were in the age group 20-34 years (80.2%), with a mean age of 28.3 years and standard deviation of 5.6 years. The ages ranged from 15 to 47 years old. The median gestation age at rst ANC visit was 15 weeks with Inter-quartile range (IQR) of 12 to 17 weeks.
The mean pre-pregnancy BMI was 25.5 kg/m² with a standard deviation of 4.6 kg/m 2 . Underweight women were 5.3%, normal weight women 51.1%, overweight women 26.9% and obese women 16.7%.
Majority of the study participants were married (85.4%), multiparous (65.2%), unemployed (64.9%), non-smokers and non-alcohol users, 98.4% and 96.6% respectively; with good antenatal visit attendance (67.8%) and had at Page 6/23 least a secondary school education (71.9%). A high proportion of women (52.1%) were from an urban residence and were self-referred from home for delivery. Participants with Gestational Diabetes Mellitus (GDM) were only 0.5% (Table 1). The highest proportion of obesity was noted in women with age 35 years and above, with secondary school and above education, employed, married, urban dwelling, non-alcohol drinking, non-smoking, with 4 or more ANC visits, multiparous, self-referred from home and in those with GDM (Table 1).   (Table 3). and (RR = 0.75, 95%CI: 0.23-2.51) respectively. These relationships were not statistically signi cant (Table 4). slightly increased risk of PE when compared to the normal weight mother with adequate GWG, (RR = 1.03, 95%CI: 0.23-1.10). The association was not statistically signi cant.
Normal weight with inadequate GWG was found to be protective against risk of PE development when compared to the normal weight mother with adequate GWG (RR = 0.25, 95%CI 0.08-0.83). This association was statistically signi cant. Overweight pre-pregnancy BMI with excessive GWG had a two folds increased signi cant risk of PE when compared to the pre-pregnancy overweight mother with adequate GWG, (RR = 2.58, 95%CI: 1.03-6.43).
Conversely the pre-pregnancy overweight mother with inadequate GWG had a reduced risk of PE development when compared to the pre-pregnancy overweight mother with adequate GWG (RR = 0.9, 95%CI: 0.22-3.36), but this was not statistically signi cant. The pre-pregnancy obese mother with excessive GWG had nine fold increased risk of PE development when compared to the obese pre-pregnancy mother with adequate GWG and this was statistically signi cant, (RR = 9.40, 95%CI, 1.91-46.23) ( Table 5).  (Table 6).

Discussion
This hospital based retrospective birth cohort study revealed a prevalence of PE among our study population to be 9.5%. This is higher than the previously reported prevalence of 3.3% from our center (18 (19)(20)(21).
The increased prevalence of PE as observed in this study highlights the increased contribution of hypertensive disorders of pregnancy including PE to the alarmingly high maternal mortality ratios, of 12% reported by Bergsjø et al., (2010) and 18% reported by Maro et al., (2016) for the same institution (4,22). This trend is attributable to the increasing rates of urbanization with associated sedentary lifestyles, increased consumption of high fat, high sugar, highly re ned and highly processed foods. Sadly, there is a concomitant decrease in the consumption of fruits, vegetables, nuts and legumes which are believed to prevent obesity and thus PE. This is consistent with the nding by Endeshaw et al which reported an association between increased risk for PE in obese young rural Ethiopian obstetric patients and reduced dietary intake of folate, fruit and vegetables (21).
Excessive GWG was also associated with increased risk of PE which is consistent with ndings as reported by some authors (19). They however did not nd any association between overweight, obesity and risk of PE. This was likely due to the small sample size (N = 462) of their retrospective cohort among Cameroonian women. Our ndings were also comparable to the ndings of a meta-analysis performed by Zabih and colleagues (23) which explored the association between high BMI and risk of PE and subtypes (24). The authors further reported that BMI was appreciated to rise within each BMI category and noted that women with BMI 35 had a triple risk of PE compared to women with BMI of 30 or 31.
Our study further showed that Obesity was associated with signi cantly increased risk of PE and its subtypes, MPE, SPE and LOPE. This was in keeping with ndings by other researchers (6,25,26). The variations in observed risk by PE subtypes can be explained by the fact that PE subtypes may have different clinical and biochemical features and different haemodynamic states. Early onset PE for instance, is typically linked to abnormal placentation with resultant placental insu ciency and is postulated to have a genetic component. This would then cause foetal growth restriction and other adverse maternal and neonatal outcomes. Maternal mortality is reported to be twenty fold higher in women who develop PE women at 32 weeks of gestation (6). Late Onset PE on the other hand is more likely to be related to maternal factors and typically involves normal foetal growth, larger placental volume, normal birth weight and more favorable maternal and neonatal outcomes. Maternal obesity has been identi ed as a crucial risk factor in both scenarios (27). Other similar studies have also observed increased risk of PE with maternal overweight and obesity. (6,18,(28)(29)(30).
Furthermore, we found that overall excessive GWG was associated with a twofold increased risk of PE.
Speci cally, excessive GWG in the pre-pregnancy overweight and obese mothers was signi cantly associated with increased risk of PE. This is in keeping with the ndings reported by previous works (31,32). This nding was also appreciated for MPE and LOPE (6,33). Inadequate GWG on the other hand was noted to be protective for SPE across the different BMI categories. Interestingly, Saftlas and co-workers (15) reported that higher than expected gestational weight gain did not increase the risk of PE and therefore was not associated with PE. The difference between their study and this present one is most likely explained by variations in study populations (different ethnic/racial distribution) and sources of GWG data (self-reported vs. medical records).
In addition, we also found that inadequate GWG in the normal weight mothers was associated with a signi cantly reduced risk of PE, which is consistent with ndings reported by Fouelifack et al. (19). Most other studies consistently reported a reduced risk of PE and subtypes in the normal weight and underweight sub categories with inadequate GWG.
The joint effects of BMI and GWG on risk of PE and subtypes from this index study, indicates that pre-pregnancy obesity and excessive GWG were jointly associated with a signi cantly increased risk of PE, MPE, SPE and LOPE. This was contrary to the study done by Shao et al., (2017) which found no such signi cant interactions .These differences in ndings could be explained by the fact that the Chinese study had a smaller proportion of obese women in their cohort when compared to our study. However, the mechanism through which obesity and excessive GWG affect hypertensive disorders is still unclear. It may be related to excessive adiposity acquisition and endothelial dysfunction resulting from in ammation associated with obesity (34).
We also reported on the heterogeneity of PE subtypes and this may be the explanation for the variations in the effect of pre-pregnancy BMI and GWG on risk of PE subtypes. Our ndings are consistent with those reported from Western and Asian populations (6,24,34) although very few assessed this joint effect by PE subtypes (6). To the best of our knowledge, there are no studies which assessed these joint effects of pre-pregnancy BMI and GWG on risk of PE and subtypes in the African population, particularly in low resource settings, thus this study.
A main strength of our study was that although it was conducted at a single tertiary level hospital, the large sample size which included participants from several districts and the neighboring districts of Kenya served to increase the statistical power of the study. This may allow for generalization to the reproductive aged women in Kilimanjaro region, Tanzania.
Furthermore, potential disease misclassi cation was minimized by the use of hospital records which were readily available to us. Detailed information on demographic factors, medical histories and lifestyle factors as well as the exclusion of participants with potential confounders was another strength of our study. There was however no adjustment for unmeasured confounding variables such as diet, physical activity before pregnancy as well as maternal and paternal genetic factors which may have biased the results.
However, potential BMI exposure misclassi cation is a limitation of our study because of weight gained in early pregnancy. The average gestational age at rst booking was 14.5 weeks of gestation. Analysis was restricted to rst ANC visit gestational age ≤ 20 weeks for study participants. Another limitation of the study was that the weight of patients at onset of PE was weight on admission for delivery; it should have been weight of onset of PE.
The small distributions in the PE subtype categories made it challenging to compute effect estimates for the GWG subcategories and risk of PE subtypes. Total GWG was used instead to compute the statistical estimates.

Conclusions
Our ndings showed high proportions of overweight and obesity among obstetric women who delivered at KCMC.
Pre-pregnancy overweight and obesity are critical modi able factors associated with increased risk of PE. Notably, excessive GWG may magnify this effect. The independent and joint effects of High BMI and excessive GWG on risk of PE were supported by our ndings. This makes overweight and obesity in reproductive age women a major health concern and especially a priority for maternal and child health.
Community sensitization and mobilization to raise awareness on the risks of overweight and obesity on development of PE should be vigorously pursued. Also, lifestyle modi cation campaigns which will target women of reproductive age and promoting nutritional education, healthy eating habits, preconception normalization of weight and exercise are necessary.
Future multi Centre studies with larger sample sizes that can study the heterogeneous nature of PE subtypes in relation to pre-pregnancy BMI and GWG and hopefully con rm our ndings and improve the certainty of our statistical estimates are recommended. Committee with certi cate number 2348. Permission was also sought from the Head of department of the department of Obstetrics and Gynaecology. Con dentiality was observed with the use of participant identi cation numbers. All data was stored under lock and key, unlinked to patient identi ers.

Consent for publication
Not applicable Availability of data and material The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare no competing interests Funding