Nausea and vomiting affect most women in early pregnancy, and for a small group the symptoms become relentless. Hyperemesis gravidarum (HG) sits at this severe end of the spectrum. It brings dehydration, weight loss, electrolyte disturbance and nutritional deficiency, and it remains one of the commonest reasons for hospital admission in the first trimester (1). For many years the condition was loosely defined, which made studies hard to compare. The Windsor consensus has since described HG as symptom onset before 16 weeks, severe nausea or vomiting, an inability to eat or drink normally and marked limitation of daily activities (2).

Grading severity is not simple either. The revised Royal College guideline recommends validated tools such as the PUQE and HELP scores, and it advises against using ketonuria to judge dehydration or severity (3). Recent clinical reviews carry the same message and give more weight to hydration, nutritional status and functional impact (4,5). This matters for research, because studies that grade severity in different ways may reach different conclusions about the same risk factor.

Understanding of why HG develops has moved forward quickly. Fejzo and colleagues showed that fetal production of the hormone GDF15 and the mother's sensitivity to it both contribute substantially to the risk of HG (6), and this work now points towards targeted treatment (7). Older explanations based on human chorionic gonadotrophin and thyroid stimulation still carry some weight; a meta-analysis found higher free thyroxine and lower TSH in affected women (8). None of these pathways, however, explains why some women deteriorate far more than others once vomiting starts.

Nutritional reserve is one plausible reason. Women with HG often eat less than half of their recommended nutrient intake, and intake falls further as symptoms worsen (9,10). Those who enter pregnancy already thin have little to spare. The effects are not limited to the mother, since a large meta-analysis linked HG with preterm birth and low birth weight (11). In a registry of more than 124,000 births, maternal underweight was among the characteristics associated with HG (12). Whether low pre-pregnancy body mass index (BMI) also predicts how severe the illness becomes, rather than simply whether it occurs, is much less clear.

The question is relevant in Khyber Pakhtunkhwa, where undernutrition among women of reproductive age remains a public health concern. Local data on BMI and HG severity are scarce. We therefore aimed to determine the association between low pre-pregnancy BMI and severe HG among women with HG admitted to Moulvi Ameer Shah Memorial Hospital, Peshawar.

This hospital-based cross-sectional study was conducted in the Department of Obstetrics and Gynecology, Moulvi Ameer Shah Memorial Hospital, Peshawar, Khyber Pakhtunkhwa, Pakistan for the duration from January 2026 to June 2026. The records of 146 pregnant women admitted with a clinical diagnosis of HG were included. Women were eligible when HG was the main reason for admission and pre-pregnancy weight and height were documented. The study was approved by the institutional ethics committee. For each woman, pre-pregnancy weight and height, HG severity, urinary ketones, serum electrolytes and length of hospital stay were recorded on a structured proforma. BMI was calculated as weight in kilograms divided by height in metres squared and grouped by World Health Organization adult cut-offs: underweight below 18.5 kg/m2, normal 18.5–24.9 kg/m2, overweight 25.0–29.9 kg/m2 and obese 30.0 kg/m2 or above. The treating team graded HG as mild, moderate or severe on clinical assessment. For analysis, mild and moderate cases were combined as non-severe HG. The primary exposure was underweight versus non-underweight BMI and the primary outcome was severe HG. Two women recorded as underweight had a BMI of exactly 18.5 kg/m2, which falls in the normal range under WHO criteria; they were kept in their recorded group for the main analysis and moved to the normal group in a sensitivity analysis. Data were analyzed in Python 3.13 with SciPy 1.18. Categorical variables were expressed as frequencies and percentages with Wilson 95% confidence intervals. The association between underweight and severe HG was tested with Fisher's exact test because one cell contained no observations, and the Yates-corrected chi-square value was reported alongside it. Effect size was expressed as a prevalence ratio (PR) with a log-method 95% confidence interval and as a prevalence difference with a Newcombe hybrid score interval. As no underweight woman had non-severe HG, the odds ratio was not estimable and logistic regression could not be fitted because of complete separation. For the four-by-three table of BMI category against severity, the Pearson chi-square statistic was calculated; because four of twelve cells had expected counts below five, significance was confirmed with a Monte Carlo exact (Fisher–Freeman–Halton) test based on 200,000 random tables. Ordered trends were examined with the linear-by-linear association test, Kendall's tau-b, Spearman's rho and Goodman–Kruskal gamma. The relation between severity and electrolyte imbalance was assessed with Fisher's exact test, and ketonuria grade and length of stay were described by severity group. A two-sided p-value below 0.05 was taken as significant.

A total of 146 women with HG were analyzed. Underweight was the largest BMI group (66 women, 45.2%), followed by normal weight (54, 37.0%), overweight (16, 11.0%) and obesity (10, 6.8%). Severe HG was recorded in 86 women (58.9%), moderate HG in 38 (26.0%) and mild HG in 22 (15.1%) (Table 1).

Table 1. Distribution of pre-pregnancy BMI categories and HG severity (N=146)

BMI, body mass index; HG, hyperemesis gravidarum; CI, confidence interval.

Severity fell away clearly as BMI rose (Table 2). All 66 underweight women had severe HG. In the normal BMI group, 20 women (37.0%) had severe and 34 (63.0%) moderate disease. No overweight or obese woman had severe HG; most overweight women (75.0%) and all obese women had mild disease. The overall association was highly significant (Pearson χ²=190.80, df=6, p<0.001).

Table 2. HG severity according to pre-pregnancy BMI category

^{Row percentages shown. Pearson χ2=190.80, df=6, p<0.001; four of twelve cells had expected counts below five (minimum 1.51), so the result was confirmed by}

^{the Monte Carlo exact test (Table 4).}

In the primary comparison, severe HG occurred in 66 of 66 underweight women (100.0%) and in 20 of 80 non-underweight women (25.0%). Underweight women were four times as likely to have severe disease (PR 4.00, 95% CI 2.74–5.85), an absolute difference of 75.0 percentage points (95% CI 63.2–83.2; Fisher's exact p=6.64×10⁻²⁴). When the two women with a BMI of exactly 18.5 kg/m² were moved to the normal group, the association was almost unchanged (PR 3.73, 95% CI 2.61–5.33; Fisher's exact p=8.85×10⁻²³) (Table 3).

Table 3. Association between underweight and severe HG, with sensitivity analysis

PR, prevalence ratio (log-method CI); difference, prevalence difference (Newcombe hybrid score CI). The odds ratio was not estimable because no underweight woman had non-severe HG.

Tests that respect the ordering of both variables confirmed a consistent downward gradient in severity as BMI increased (Table 4). The linear-by-linear association was strong, and Kendall's tau-b and Spearman's rho both indicated a large negative correlation. Goodman–Kruskal gamma reached −1.00, meaning that no pair of women showed higher BMI together with more severe disease. Severity was also closely tied to the clinical indicators. Every woman with severe HG had an electrolyte imbalance, while none of the 60 non-severe cases did. Women with severe HG stayed in hospital for 4–6 days and had ketonuria of 2+ to 4+, whereas non-severe cases stayed 1–3 days with negative or 1+ ketones.

Table 4. Trend analysis and supplementary tests of association

Severity coded as mild < moderate < severe; BMI coded as underweight < normal < overweight < obese.

In this study, low pre-pregnancy BMI tracked closely with severe HG. Every underweight woman had severe disease, compared with one in four of the remaining women, giving a prevalence ratio of four. Severity eased step by step from the underweight group to the obese group, and the finding held when borderline BMI values were reclassified. To our knowledge, this is one of very few reports from Khyber Pakhtunkhwa to examine BMI and HG severity together. Our findings agree with a growing body of work that links poor nutritional reserve with more serious illness. In 685 Chinese women with HG, Hong et al. found BMI to be one of four independent predictors of admission, with higher BMI linked to lower risk (13). Ghanem et al. in Saudi Arabia reported lower body weight and BMI among women with HG than among controls, along with hyponatraemia in 29% and hypokalaemia in 21% of cases (14). Soysal et al. likewise observed lower BMI and albumin in affected women and an inverse relation between the HALP score and PUQE-measured severity (15). In 246 women, Balkaş et al. showed that albumin-based nutritional indices worsened as severity increased and correlated with ketonuria grade and length of stay (16). Taken together, these studies suggest that markers of depleted reserve, whether anthropometric or biochemical, gather at the severe end of the disease. Other studies have emphasised different factors. In a case-control study from southern Ethiopia, the independent determinants of HG were antenatal follow-up, multiple pregnancy and previous or family history of HG (17). In a Finnish study adjusted for BMI, Lindström et al. found that women with long and repeated admissions had been admitted earlier in pregnancy, while a personal or family history of nausea did not strongly predict overall severity (18). These differences largely reflect design. The Ethiopian study compared women with and without HG, and the Finnish study defined severity by admission pattern, whereas we looked at clinically graded severity among women who already had the condition. They also remind us that BMI is only one of several influences on how the illness unfolds. The strength of the association in our study was greater than in earlier reports. No underweight woman had mild or moderate HG, and severity was closely aligned with electrolyte imbalance, ketonuria grade and length of stay. Other cohorts have shown more overlap; in the Irish series of Doherty et al., ketonuria was present in only 42.8% of women and PUQE scores were split almost evenly between moderate and severe (19). Part of this difference may reflect local grading practice, since clinical assessment of severity often takes weight, ketones and electrolytes into account, whereas current guidance recommends validated symptom scores (3,5). The magnitude of the effect should therefore be interpreted with some caution. The clinical message is still worth stating. Nutritional depletion in HG can lead to Wernicke encephalopathy from thiamine deficiency (20), and such cases have been described in Pakistan (21). Underweight women admitted with HG should receive thiamine early, close electrolyte monitoring and dietetic input, in line with current guidance (3,22). Because the evidence for many antiemetic and supportive treatments is still of low certainty (23), careful attention to fluids and nutrition remains central to care. The local setting adds weight to these points. In the Women First trial, which recruited mothers from four countries including Pakistan, 17% were underweight before pregnancy, and maternal underweight was linked to poor linear growth in their children (24). The milder disease seen in overweight and obese women in our study should not be read as a benefit of excess weight, which carries its own obstetric risks. HG has also been associated with preterm birth and low birth weight (11), which makes early recognition of the most vulnerable mothers more important. Finally, a bidirectional link between HG and eating disorders has been reported (25), so very low BMI in a woman with HG may sometimes deserve a gentle enquiry about eating history. This study has limitations. It was cross-sectional, single-centre and included only women with HG, so it cannot show whether low BMI raises the risk of developing HG. Severity was graded by the treating team rather than against a validated score. Complete separation prevented adjusted analysis, and data on previous HG, multiple pregnancy, comorbid illness and socioeconomic status were not available. Pre-pregnancy weight may partly rest on recall, and a hospital sample may not represent women treated as outpatients. Its strengths include the use of exact and trend tests suited to sparse tables and a sensitivity analysis around the BMI threshold. Future work should be prospective and multicentre, use the Windsor definition with a validated score such as PUQE-24, measure weight and height at booking and collect enough confounders to allow adjusted models.

Low pre-pregnancy BMI was strongly associated with severe HG among women with HG admitted to Moulvi Ameer Shah Memorial Hospital, Peshawar. Every underweight woman had severe disease, and severity eased steadily as BMI increased. Recording weight and height at the first antenatal contact is simple and may help identify women who need early nutritional and electrolyte support. Prospective studies with validated severity scores are needed to confirm the size of this effect.

Declaration

Funding: This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.

Acknowledgements: The authors thank the medical officers, nursing staff and record-keeping personnel of the Department of Obstetrics and Gynecology, Moulvi Ameer Shah Memorial Hospital, Peshawar, for their help with data collection.

Conflict of Interest: The authors declare no conflict of interest.

Author Contributions

All authors contributed to the conception and design of the study, data collection, analysis and interpretation, and drafting or critical revision of the manuscript. All authors approved the final version and agree to be accountable for the work.

Data Availability

The data supporting this study are available from the corresponding author on reasonable request.