The 85 patients with confirmed hyperemesis were on average 25.8 years old (median = 25, IQR: 21–30), and the mean gestational age at the time of diagnosis was 13.2 weeks (median = 11.9; IQR: 9.9–14.7). Most were white (54.1%), in a stable relationship (32.9% cohabitating with their partners and 31.8% married), originated from the state of São Paulo (58.8%), with an educational level of up to complete secondary education (56.5%) and had received prenatal care in the same hospital (50.6%). Regarding occupation, most were housewives (44.7%); the remainder were students (7.1%) or worked in low-paying jobs, such as store clerks (4.7%). Concerning their obstetric history, for a little more than a third of the patients, it was their first pregnancy (36.5%), and almost half were nulliparous (44.7%), while almost one-fifth reported a previous miscarriage (18.8%). There was only one twin pregnancy (1.2%) and two cases of fetal malformation (2.3%). Six patients had data from two successive pregnancies, with recurrence of hyperemesis. A previous history of hyperemesis was observed in four other patients, totaling 10 patients with recurrent hyperemesis (11.8%).

From a nutritional point of view, the average pre-gestational body mass index (BMI) was 24.6 kg/m${}^2$, with an IQR between 21.6 and 26.7, while the average BMI at admission was 21.9 kg/m${}^2$, with an IQR between 18.5 and 23.8.

With regard to the initial clinical picture, two-thirds of the patients (76.5%) presented with clinical signs of dehydration, and approximately one-fourth reported abdominal pain (27.1%) and being in good general condition (25.9%). Considering the difference between the weight at admission and the average weight of the patients before pregnancy, the average weight loss was 7.784 kg (approximately 12% pre-pregnancy weight, on average). Weight loss was $\ge$5% in 94.4% of cases; $\ge$10% in 64% of cases; and $\ge$20% in almost 10% of valid cases.

Compared to initial laboratory tests, approximately two-thirds of the patients had some laboratory alteration, as shown in Table 1.

The 85 pregnant women were hospitalized on average for 15 days, ranging from 1 to 142 days, and in general had more than one hospitalization (range of 1 to 9 hospitalizations, with an average of 1.8). The weight gain between admission and delivery was on average 9.788 kg. Although this may seem appropriate, when considering the total weight gain throughout pregnancy and using the pre-pregnancy weight, it was observed that the weight gain was inadequate, with an average of only 2.594 kg (IQR: –4 to 9.3). Thus, many patients were unable to recover their pre-pregnancy weight.

As shown in Table 2, most of the patients were provided with intravenous hydration and electrolyte or vitamin replacement therapy. Only 6 (7.1%) patients were provided with enteral nutrition, and only one (1.1%) was provided with parenteral nutrition. A large majority used more than one anti-emetic drug (mean of 2.7 drugs per patient). Notably, a large number of patients received anxiolytic or even antidepressant medication.

The rate of preexisting clinical complications was 23.5%, including the presence of the following conditions: urinary tract infection (6); depression (3); cardiomyopathy (2); chronic hypertension (2); syphilis (2); pyelonephritis (2); and other psychiatric disorders (3: schizophrenia, psychosis, and factitious disorder). Considering only complications directly related to HG, the rate of clinical complications was 7% (6/85), including the following conditions: prerenal acute kidney injury because of dehydration (3), and Mallory-Weiss syndrome (3). Overall, clinical complications occurred in 30.5% (26/85) of patients.

We evaluated the birth data of 40 patients. As shown in Table 3, the cesarean delivery rate was 60% and the median gestational age at delivery was 38.2 weeks. There were three stillbirths (7.5%), 12 babies (35.3%) were born prematurely (37 weeks), and the low birth weight rate was 32.5%.

Five of these patients (5/40 = 12.5%) had postpartum complications, including acute anemia with coagulopathy, subaponeurotic hematoma with pulmonary thromboembolism, hematuria, hemorrhage with fever, and puerperal mastitis. One of these women was admitted to an intensive care unit.

Regarding the initial laboratory profile, we found a general condition of greater severity, with worse rates in most parameters in comparison with other authors, as shown in Table 4 (Ref. [11, 13, 18, 19]).

Our rate of electrolyte changes was one of the highest in the literature. The rate of hypokalemia was the highest among all authors surveyed, while the rate of hyponatremia was surpassed only by data reported in a French study [19].

Regarding liver function tests, as shown in Table 4, our findings are similar to those of Chraïbi et al. [19], but are more severe than those of other authors. On the other hand, our rate of hyperbilirubinemia is the highest among all authors consulted [11, 13]. Furthermore, Goodwin [20] established that mild liver enzyme elevations would be present in 20–30% of patients with HG, i.e., at rates lower than that found in the present study, which was 46.4%. Indeed, some authors have established a direct relationship between weight loss and liver changes [11]. The etiology of abnormal liver enzymes in HG is unclear. The liver enzymes return to normal levels promptly when vomiting ceases and adequate nutrition is resumed. It has been suggested that abnormal liver function is a combined effect of hypovolemia, malnutrition, and lactic acidosis that would occur in HG and may thus be related to the severity of HG [21]. This is in agreement with our finding of prerenal renal failure in 3.5% of cases.

The rate of anemia in the present study was similar to that reported in a Malaysian study [13], but more than double the rate reported in the French study [19]. However, our rate of anemia was lower than those described in an online study [11], both in patients with a milder condition (27.3%) and in those with a more severe condition (36.4%). However, these rates may refer to the final condition and not at the initial point.

The rates of transient hyperthyroidism (38.6%) found in the present study are relatively high, but are in agreement with those reported in several other studies in the literature. However, there are several different types of analysis used in the various studies, some distinguishing transient hypothyroidism and others encompassing all thyroid disorders. Some considered only TSH (Thyroid Stimulating Hormone) suppression, while others also considered the levels of free T4 and anti-thyroid antibodies. It should be noted for this comparative analysis that we excluded from our cases series four cases of pre-pregnancy hyperthyroidism or those with a diagnosis made during pregnancy that could explain the vomiting. This same methodological concern was not reported in any of the previously evaluated studies. In two of these studies [13, 19], the excluded cases were described, and none were reported as being excluded due to thyroid disease. In any case, a review [20] indicated that transient hyperthyroidism occurs in 50–70% of women with HG. In a meta-analysis study [21], higher levels of free T4 and lower levels of TSH were described in patients with HG compared to controls.

As shown in Table 2, the treatment performed in the present study consisted mainly of intravenous hydration, with possible electrolyte replacement (in half of cases), in addition to antiemetic and anxiolytic drugs (in 36.5% of the cases). The most commonly used antiemetic was dimenhydrinate (87.1%), followed by metoclopramide (85.9%) and ondansetron (38.8%). This treatment regimen differs marginally from that of other authors. An American study reported that among the patients with milder and shorter-lasting symptoms, the treatments were distributed as follows: 55% seabands (wristbands that exert acupressure on the P6 Nei-Kuan point), 45% ondansetron, 41% promethazine, 38% antacids, 20% metoclopramide, 12% homeopathy, and 9% total parenteral nutrition. Among the most severe and longer-lasting cases, the treatment was slightly different: 72% seabands, 70% ondansetron, 66% promethazine, 56% antacids, 49% metoclopramide, 20% homeopathy, and 10% total parenteral nutrition [22].

The uncommon use of ondansetron in our sample is noteworthy, as it is a medication used in more than half of the patients in some series [11], and is known to be quite efficient in controlling symptoms [5]. It appears that there was some parsimony in its use due to a fear of using a newer medication. Indeed, this is a genuine concern, because there is still some discussion in the medical literature about the safety of this drug. In fact, the safety of ondansetron has yet to be established. Recently, a meta-analysis [23] that studied 12 papers, concluded that there is an increased risk of ventricular septal defects (OR = 1.11) and cleft lip (OR = 1.22) and perhaps cleft palate (OR = 1.48). Although this perception of risk must be balanced against the increase in absolute risk, which still seems to be small [24], we understand that the use of this medication should be done with caution, reserving its prescription for more serious cases.

Furthermore, the rate of use of anxiolytics (36.5%) in the present study was extremely high, compared to Canadian data describing the use of anxiolytics in only 3.4% of pregnant women [25]. The use of antidepressants (6%) was also twice the rate described in an international systematic review study: 3.0% [26].

We used enteral nutrition 7.1% of the time and parenteral nutrition in only one case (1.2%). This is a very small rate given the severity of the cases. As a comparison, the rates described for parenteral nutrition in the United States vary between 15.9% and 35.1% and, for enteral nutrition, between 2.3% and 20.2% of cases of HG [11, 27, 28].

Furthermore, there are few reports in the medical literature on the use of parenteral or enteral feeding in HG. Holmgren et al. [28] described higher rates of the use of enteral nutrition (20.2%) and parenteral nutrition (35.1%) in their 94 pregnant women with HG, reporting fewer complications in enteral nutrition (tube displacement in 10.5%) and major complications in parenteral nutrition, with 66.4% having infection, thromboembolism, or both. Despite this, the enteral route was often rejected by the patients.

An interesting treatment protocol was established by a service in Norway [29], where there was a progressive evolution of conduct every three days if there was no improvement in intake or weight gain. With such procedures, high rates of the use of peripheral parenteral nutrition (31.72%) and enteral nutrition (19.17%) were described. Women with enteral nutrition exhibited greater weight loss at admission (mean of 5 kg) and at the beginning of treatment (mean of 5.5 kg) than the other groups (mean of 4 kg), representing a more severe condition. However, weight gain at the end of pregnancy was similar between the groups, demonstrating the effectiveness of this approach.

Thus, it seems reasonable that nutritional support should be initiated in women with HG who continue to lose weight and who do not respond to pharmacological and non-pharmacological treatments. The decision to initiate enteral or parenteral nutrition should be individualized, taking into account the gestational age, comorbidities, and preferences of the patient, as well as experience and institutional resources. In general, enteral nutrition is preferable, given the increased health risks associated with parenteral nutrition during pregnancy [5].

There appears to be a consensus that enteral feeding and total parenteral nutrition should be considered if intravenous therapy is not successful in reducing symptoms and there is still a caloric deficit.

Given such satisfactory results with enteral nutrition and considering the severity of our cases, with significant weight loss, we understand that we still have little use of this resource in our sample, maybe because we still have little experience. We believe that a similar situation exists for most Brazilian services and perhaps for many other services around the world. Therefore, the data presented here are of importance in the sense of alerting us to a change in our protocols. It would therefore be interesting to make greater use of this treatment option in the future, seeking better results.

As can be seen from the analysis in Table 5 (Ref. [10, 12, 14, 30, 31, 32]), the obstetric and neonatal results of our sample were more severe than those found by most international authors who investigated this topic.

Initially, 60% of the deliveries in our sample were cesarean, a rate higher than that of our tertiary service, which was approximately 45–50%, but also higher than that of all other authors. There are indications that the cesarean rate increases according to the severity of the HG case. Dodds et al. [10], analyzing 1270 women with HG in Nova Scotia, Canada, found higher cesarean rates among patients with a weight gain 7 kg than among other patients: 26.4% vs. 19.9%; OR = 1.4 (95% confidence interval [CI]: 1.0–1.8).

We also observed higher rates of premature birth, SGA, and low birth weight newborns in the sample, in addition to lower Apgar scores. Furthermore, we observed lower mean weight and gestational age at birth. Such rates could be related to the greater weight loss and lower weight gain observed in the sample. The study by Dodds et al. [10] demonstrates this association when comparing patients with HG with a weight gain 7 kg with other patients, noting an increased risk for LBW newborns, (12.5% vs. 3.4%, OR = 2.8), SGA (14.6% vs. 9.2%; OR = 1.5), prematurity (13.9% vs. 4.1%; OR = 3.0), and fifth minute Apgar 7 (5.6% vs. 0.6%; OR = 5.0). This association is reinforced by the data from Stokke et al. [29]. Norwegian patients hospitalized with HG who achieved a weight gain 7 kg had 20% LBW and 9% prematurity. Patients who achieved a weight gain $\ge$7 kg had 7% LBW and 5% prematurity (significant only for LBW). In the multivariate analysis, a weight gain 7 kg was the variable that attributed the highest risk for the LBW event. Authors such as Hastoy et al. [33] also observed an association between weight gain during pregnancy in women with HG and LBW. In a case-control study, they noted that when weight gain during pregnancy was 7 kg, there was a significant risk in HG patients compared to control patients for LBW (OR = 2.0; 1.0–3.1) and for fetal growth restriction (OR = 1.7; 1.1–2.4), but not for prematurity (OR = 1.6; 0.8–2.8).

It is worth noting that 64% of our patients reported weight loss $\ge$10% and that the average weight gain rate during pregnancy was only 2.59 kg, with a median of 2.2 kg and an IQR between –4 kg and +9.2 kg. Such a picture is compatible with a significant risk for inadequate neonatal outcomes. In fact, in our data, the LBW rate doubled (42.9% vs. 21.4%) when weight loss was $\ge$10%, reinforcing the importance of such a large weight loss without satisfactory recovery in neonatal outcomes. Veenendaal et al. [34] established, in a systematic review with meta-analysis, that there is a higher risk in HG for low birth weight newborns, with rates of 6.4% in patients with HG and 5.0% in controls (OR = 1.42; 95% CI: 1.27–1.58). A similar risk was found in relation to preterm births, with rates of 7.4% in pregnancies with HG and 5.8% in normal pregnancies (OR = 1.32; 95% CI: 1.04–1.68).

The predominance of male fetuses (56.4%) observed in our sample was surprising, and is different from that observed in other studies. Schiff et al. [35], in a study with pregnant women hospitalized with HG in the US state of Washington, reported that the chances of these mothers of having a female baby were 50% higher than in healthy controls when the diagnosis of HG was made in the first trimester (OR = 1.5; 95% CI: 1.4–1.7), a risk that disappeared when the diagnosis was made later. As our average gestational age at diagnosis was 13.2 weeks, already later than the first trimester, this could partly explain the male predominance. Askling et al. [36] explained the predominance of female newborns by suggesting the possibility of female fetuses inducing greater amounts of beta-hCG (human chorionic gonadotropin). We speculate, therefore, that the cases in the present study may have had a lower hormone dosage, and that the clinical picture would not only be more severe, but it would also be explained depending on other causal factors. On the other hand, there does not appear to be any plausible evidence for hCG being associated with HG [1, 2, 5]. We look forward to new evidence of a genetic involvement in HG [37]. Perhaps the most severe cases are related to alterations in gene sequences and the milder cases are not. If so, this difference in fetal sex may be of importance in future investigations.

The high rates of inadequate neonatal outcomes, prematurity, SGA, and low birth weight found in the present study are possibly related to higher rates of fetal distress, low Apgar scores, greater use of neonatal ICUs, and higher perinatal mortality. Our neonatal intensive care unit utilization rate was relatively high, as seen in the comparison, for example, with data from Fiaschi et al. [12]: 10% vs. 1.6%.

Our stillbirth rate, 6.9%, was also very high, as can be seen in comparison with other studies. Hastoy et al. [33], for example, reported a 1.5% perinatal mortality, while Roseboom et al. [38] reported a 0.3% rate. Källén [9], in turn, reported a stillbirth rate of 0.47% and a neonatal mortality of 0.37%, that is, a perinatal mortality of 0.84%. At the other extreme, Vikanes et al. [30], in a Norwegian population-based study, described a zero perinatal mortality rate in their sample consisting of 814 pregnant women with HG.

Such results, which show worse rates than those of most other authors, may represent a more severe HG, probably related to the use of more restrictive diagnostic criteria of a well-established protocol, excluding dubious and less severe cases. In addition, we can consider that the treatment given to these women should have been better, more effective, and extensive, to ensure adequate weight gain and less maternal and perinatal repercussions. Thus, more restrictive diagnostic criteria may be useful to ensure a more accurate and early diagnosis of this condition, allowing more successful referral and treatment.

One of the strengths of this study is the fact that it is the first to survey a series of HG cases in Brazil, with a detailed description of laboratory, clinical, obstetric, and neonatal aspects. In addition, it advances the field by describing the findings based on more restrictive criteria, without considering only the presence of vomiting, but rather more compromised conditions, which would require hospitalization and multidisciplinary care. In this sense, the description of these cases leads to a series with one of the most severe HG characteristics in the literature, and serves to guide Brazilian obstetric services to best address these cases.

On the other hand, our study has some limitations and weaknesses.

Because the study was performed in a tertiary and university hospital, the possible presence of clinical and obstetric pathologies in these patients may have partially compromised the impact of obstetric and neonatal outcomes, which may be associated not only with HG, but also with these other pathologies. However, many of these diseases had already been excluded before the final analysis, as they could mimic HG-like conditions. Certainly, the inclusion of only 40 pregnant women with birth data is also a limitation of this study. There was a loss of approximately half of the cases hospitalized for HG, which compromises to some extent the assessment of perinatal outcomes. It possible that the patients who delivered elsewhere had good results and therefore did not return to our hospital, but this is just an assumption. In any case, the data we were able to obtain denote serious repercussions, and these are the only data on this topic currently available in Brazil. Finally, a larger sample would have led to a more robust comparative analysis.