Hyperemesis gravidarum has for practical purposes been defined as intractable vomiting of pregnancy severe enough to require hospital admission1 affecting up to 2.3% of pregnancies.2 Hyperemesis gravidarum is thus differentiated from the less severe nausea and vomiting of pregnancy, which affects up to 85% of pregnancies.3 Hyperemesis gravidarum is the second most common indication for hospitalization in women with successful pregnancies.4 Patients affected by hyperemesis gravidarum are dehydrated and starved with associated metabolic, electrolyte, and endocrine disturbances: hyponatremia is present in 43–49% and hypochloremia in 33–40% on hospital admission.5,6 Ketonemia and resultant ketonuria is the consequence of the switch to an alternative energy source when dietary glucose is insufficient for metabolic needs.7 The brain in the satiated state uses glucose as the exclusive energy substrate.8 The average requirement by the adult brain for glucose is 100 g per day and the recommended daily amount for glucose in pregnancy is 175 g.9 Adults should get 45–65% of their calories from carbohydrates and young women in the first trimester of pregnancy should consume 2,400 calories per day.9 Approximately 25% of patients with hyperemesis gravidarum treated with metoclopramide and standard saline rehydration were still ketonuric at 24 hours.10
Our objective was to estimate whether the addition of 5% dextrose to 0.9% saline rehydration solution in the first 24 hours after hospitalization, thereby providing 150 g of glucose over 24 hours intravenously (equivalent to 600 calories) when nutritional intake is likely still limited, would result in faster resolution of ketonuria and a more rapid general recovery, culminating in increased perception of well-being.
PATIENTS AND METHODS
The trial was conducted in a university hospital in Kuala Lumpur, Malaysia. Ethical oversight was provided by the University of Malaya Medical Center Medical Ethics Committee (approval date September 22, 2010, reference number 811.8). This trial was performed in compliance with the Declaration of Helsinki. The trial is registered with a public trial registry with the identifier ISRCTN65014409. All participants provided informed written consent.
Women at their first hospitalization for hyperemesis gravidarum (intractable nausea and vomiting of pregnancy with dehydration and starvation clinically judged to require hospitalization for intravenous rehydration and antiemetic drug administration) were enrolled within 2 hours of ward admission by their health care providers and randomly assigned to receive either 5% dextrose–0.9% saline or 0.9% saline by intravenous infusion at a rate 125 mL/h over 24 hours in a double-blind trial. Women already under intravenous rehydration therapy were not recruited. We did not provide outpatient intravenous rehydration and antiemetic therapy for hyperemesis gravidarum; women who needed these therapies were all admitted. Other inclusion criteria were age 18 years or older, ketonuria by urine dipstick of at least 1+ on admission, gestation 16 weeks or less, plasma glucose 110 mg/dL or less, and sodium 125 mmol/L or greater on admission.
We excluded women in the case group with multiple gestation, established nonviable pregnancy, pre-existing medical conditions that can cause nausea and vomiting (eg, culture-proven symptomatic urinary tract infection, dengue fever), gastrointestinal causes of vomiting (eg, gastroenteritis, gastritis, peptic ulcer), medical causes of vomiting (eg, diabetic ketoacidosis), and women with underlying medical problems (eg, established gestational hypertension, diabetes, heart disease, renal disease, and thyroid disorder).
In our center, as standard initial treatment, patients with hyperemesis gravidarum received intravenous rehydration with 0.9% saline solution (potassium chloride was added as required if hypokalemic), 10 mg oral thiamine daily, and an intravenous antiemetic (typically 10 mg metoclopramide 8 hourly).11 Oral intake was allowed as tolerated at a pace decided by the affected women. We did not ask participants to record their oral intake.
Participants were recruited by health care providers as they were admitted to the gynecology ward. Blood was taken for renal function, plasma glucose, and full blood count. They also were given the visual numerical rating scale for nausea to be filled out at recruitment and 8, 16, and 24 hours after admission.
Randomization was affected by the sequential opening of numbered, sealed, opaque envelopes stating “Protocol A” or “Protocol B.” Randomization sequence on a one-to-one ratio was computer-generated (using http://www.random.org) for 223 participants by a coauthor (P.C.T.) who played no role in enrollment.
Intravenous solutions were prepared by a coauthor (M.J.N.) using standard hospital issue 5% dextrose–0.9% saline or 0.9% saline solution in 500-mL containers with the manufacturer's label stripped off and relabeled as solutions A or B. The solutions and containers were identical in appearance except for the A or B labels. Randomization was to either A or B (either 6×500 mL .9% saline or 6×500 mL of 5% dextrose–0.9% saline) to be given intravenously over 24 hours (ie, 500 mL of solution every 4 hours). One gram (9.5 mmol) of potassium chloride can be added to each 500-mL solution as required to correct hypokalemia. Participants and health care providers were thus both blinded to the allocated solution.
A multivitamin preparation, P-Trovite, which contains 250 mg thiamine, was given intravenously to all participants before starting the trial rehydration solution to prevent Wernicke's encephalopathy that may arise as a consequence of dextrose infusion in a severely thiamine-deficient state. Intravenous antiemetic was also prescribed according to health care providers to all participants for 24 hours or until significant symptom relief was established.
During the 24-hour study period, urine was checked by dipstick 8 hourly for ketonuria and glycosuria. If there was significant glycosuria 2+ or greater present, the patient's capillary glucose level was checked with a glucometer and if reading 8 mmol/L or greater, infusion fluid was to be changed to open-label standard 0.9% saline. Participants were instructed to mark the nausea visual numerical rating scale (10 points, high score denoting more severe nausea) before initial administration of the intravenous fluid regime and then at 8, 16, and 24 hours. At 24 hours, participants were also asked to mark their perceived well-being over the study period with a 10-point visual numerical rating scale (higher score, greater well-being). At the conclusion of the 24-hour main study period, open-label intravenous fluid was started if still required.
Primary outcomes were resolution of ketonuria and well-being (by 10-point visual numerical rating scale) at 24 hours. Secondary outcomes included frequency of vomiting in the 24-hour study period; nausea visual numerical rating scale at 8, 16, and 24 hours, hyponatremia (135 mmol/L or less); hypokalemia (3.5 mmol/L or less); hypochloremia (99 mmol/L or less); hyperglycemia (8 mmol/L or greater) at the end of the 24-hour main study period; duration of intravenous antiemetic and intravenous rehydration during hospitalization; and the hospital admission to discharge interval. Data were extracted onto the case report form after hospital discharge from case notes, laboratory reports, and direct contact of participants if needed.
There is no trial comparing intravenous fluid regimes for hyperemesis gravidarum to guide sample size calculation. A previous trial of promethazine compared with metoclopramide (using 0.9% saline as the standard intravenous fluid) in our center has shown that 25% of patients with hyperemesis gravidarum had not cleared their ketonuria and well-being by visual numerical rating scale is 7.6±2.2 in the metoclopramide arm at 24 hours.10 Sample size calculations were made with the PS program. Assuming a 10% compared with 25% (relative risk 0.4) rates for persistence of ketonuria at 24 hours for 5% dextrose–0.9% saline and 0.9% saline arms respectively, α=.05, power 80%, one-to-one recruitment ratio, and applying the χ2 test, 100 women are required in each arm. Factoring in a 10% dropout rate, a total of (200/0.9) 223 women are needed for a suitably powered study. Similarly, assuming that well-being visual numerical rating scale score will improve by 1 with 5% dextrose–0.9% saline, applying the Student’s t test, 77 participants will be required in each arm. If the Mann-Whitney U test were to be applied instead of the Student’s t test in the event of nonnormally distributed visual numerical rating scale scores, a 10% increase in numbers is appropriate and factoring in a 10% dropout rate, a total of only (154×1.1/0.9) 189 women are required for a powered study on this outcome. We planned to recruit 223 women.
Data were entered into SPSS 17. Analysis was by intention to treat after exclusions for criteria infringements. Normality of data distribution was checked with the Kolmogorov-Smirnov test. Normally distributed continuous data were analyzed with the Student's t test. Two-by-two categorical data sets were analyzed with the Fisher’s exact test and larger categorical data sets with the χ2 test; ordinal data and nonnormally distributed continuous data were analyzed with the Mann-Whitney U test. A repeated-measures analysis of variance was applied to the nausea visual numerical rating scale scores and to ketonuria status. All tests were two-sided and P<.05 was considered significant.
The trial was conducted from November 9, 2010, to February 6, 2012. Recruitment was stopped on reaching the targeted number of participants. The recruitment flowchart of participants through the trial is shown in Figure 1. Exclusions were mostly the result of study criteria infringements (shown in Fig. 1) that were ascertained only after randomization as a result of our pragmatic trial process where allocated treatment was started before full test results were available. All participants received P-Trovite intravenously and had their rehydration regimen administered as allocated.
The characteristics of the participants in the two trial arms are shown in Table 1. Participants in the two arms have similar characteristics except for mean serum potassium in which there was a mean difference of 0.1 mmol/L between the trial arms. All other parameters from the standard renal function test, full blood count, and random glucose as well as biometric variables of blood pressure and pulse at recruitment were similar (result not shown).
Table 2 displays the primary outcomes of ketonuria and well-being visual numerical rating scale score according to treatment allocation. Ketonuria was still present at 24 hours in 10 of 101 (9.9%) compared with 11 of 101 (10.9%) (P<.99) (relative risk 0.9, 95% confidence interval [CI 0.4–2.2) and median (interquartile range) well-being score was 9 (8–10) compared with 9 (8–9.5) (P=.73) for 5% dextrose–0.9% saline and 0.9% saline arms, respectively, at the end of the 24-hour main study period.
Of the secondary outcomes (Table 3), the nausea visual numerical rating scale scores at 8 and 16 hours showed a significant reduction (P<.01 and P=.03, respectively) in favor of the 5% dextrose–0.9% saline arm but the difference had dissipated by 24 hours. Repeated-measures analysis of variance of between-subject nausea visual numerical rating scale scores showed a significant result (P=.046), but interaction of treatment and time was also significant (P=.004). Hospital stays (defined as interval in hours from randomization to documented medical decision to discharge) were 43±21 compared with 48±21 (P=.14) for 5% dextrose–0.9% saline and 0.9% saline arms, respectively.
There was a statistically significant difference in the serum potassium level after completion of the 24-hour study period (mean±standard deviation): 3.9±.4 compared with 3.8±.3 mmol/L (5% dextrose–0.9% saline compared with 0.9% saline arms; P=.04); the mean .1-mmol/L difference in serum potassium level is identical to the pretreatment mean difference between the two trial arms and is likely to have arisen from the differing baseline. The proportion classified as hypokalemic (3.5 mmol/L or less) was not different (19.1% compared with 28.7%; P=.17, relative risk .6, 95% CI 0.3–1.2 between the trial arms at 24 hours.
Post hoc, using the paired t test, for the entire trial cohort, mean glucose, serum sodium, and chloride levels increased in the first 24 hours after hospitalization (87–100 mg/dL, 134–138, 101–106 mmol/L respectively, all P<.001), whereas that for serum potassium decreased marginally from 3.92 to 3.84 mmol/L (P=.014). Adjusting for antiemetic regimen did not alter the result for persistence of ketonuria at 24 hours (adjusted odds ratio 0.8, 95% CI 0.3–2.0, P=.60; 5% dextrose–0.9% saline compared with 0.9% saline). Restricting analysis to only the metoclopramide-exposed cases also did not alter the nonsignificant results on persistence of ketonuria (P=.81) and well-being score (P=.43); the significant result on the nausea visual numerical rating scale by repeated-measures analysis of variance (P=.049) in favor of the 5% dextrose–0.9% saline arm remained.
If participants excluded postrandomization as a result of enrollment criteria infringements were included in the analysis, outcomes between the trial arms were still similar. Similarly, adjusting for the potassium level also did not materially alter the results. There was no protocol curtailment as a result of hyperglycemia during the 24-hour study period.
To our knowledge, intravenous rehydration regimes in the management of hyperemesis gravidarum have not been previously studied. We performed an unrestricted online PubMed search (http://www.ncbi.nlm.nih.gov/sites/entrez) on May 13, 2012, using the search terms “hyperemesis gravidarum hydration trial” and identified only a review article. There seems to be little change since the 2004 Practice Bulletin from the American College of Obstetricians and Gynecologists, which stated that no study has compared different fluid replacements for nausea and vomiting of pregnancy.3
Intravenous rehydration is a mainstay of management in hyperemesis gravidarum, but reflecting the dearth of evidence, the fluid regime is typically not defined in guidance and reviews.3,12,13 A recent review cautions against the use of dextrose in the rehydration fluid in the rare event of Wernicke's encephalopathy being precipitated as a result of thiamine deficiency and the overrapid correction of hyponatremia possibly leading to central pontine myelinolysis,14 whereas another recommends total food avoidance and parenteral administration of carbohydrate, vitamins, and amino acids in the process of correcting volume and electrolyte imbalances in hyperemesis gravidarum.15 Our data suggested that intravenous multivitamins, antiemetic, and a rehydration regime comprising 3 L of 0.9% saline over a 24-hour period in the initial management of hyperemesis gravidarum is seemingly effective: at 24 hours, in 89%. ketonuria had resolved, hyponatremia reduced from 76% to 18% (with only a single case of a 11-mmol/L increase in serum sodium and another case in which serum sodium level peaked outside our upper normal limit of 144 at 145 mmol/L), nausea visual numerical rating scale reduced from a median score of 9 to 2, and 54% (54 of 101) had stopped vomiting altogether.
In the context of pediatric acute gastroenteritis, intravenous dextrose in the rehydration fluid on an outpatient basis reduced the risk of a return visit requiring admission.16 It has been postulated that reduced carbohydrate intake leads to free fatty acid breakdown, excess ketones, and an increased likelihood for continued nausea and vomiting and that the addition of dextrose to intravenous rehydration therapy reduced free fatty acid breakdown and reduced the incidence of outpatient treatment failure as a result of intractable nausea and vomiting.17 Our data showed a temporary reduction in nausea score in the 5% dextrose–0.9% saline arm at 8 and 16 hours, but the reduction did not seem to be mediated through a reduction in ketonuria nor was there any reduction in vomiting.
Our trial protocol did not routinely include potassium supplementation in the rehydration solution in either trial arm, although potassium could be added to correct hypokalemia at health care providers' instruction. This resulted in a marginal drop of 0.1 mmol/L in mean serum potassium level and an increase in the proportion classified as hypokalemic (3.5 mmol/L or less) from 17% to 24% (the lowest recorded serum potassium of 3.0 mmol/L in one woman) at 24 hours. In 51% of participants, serum potassium level fell in the first 24 hours. There seems to be a reasonable case for routine supplementary potassium in the initial management of a typical case of hyperemesis gravidarum.
There was a small 0.1-mmol/L difference in the mean serum potassium level at recruitment and similarly at end of the 24-hour study period between the trial arms. The small difference at recruitment was probably a chance occurrence, which was carried through to the end of the study period. A 0.1-mmol/L difference in mean serum potassium is unlikely to be of clinical significance. Adjustment for potassium level at recruitment did not affect primary outcome of persistence of ketonuria between trial arms.
Despite 150 g of dextrose by intravenous administration (equivalent to 50% of the daily normal carbohydrate intake in an average diet and 600 calories) to women in the 5% dextrose–0.9% saline arm, the rise in plasma glucose was similar in magnitude at 24 hours when compared with the 0.9% saline arm. This finding may reflect oral food intake near the 24-hour mark in the 0.9% saline arm because nausea and vomiting improved or increased glucose metabolism in the 5% dextrose–0.9% saline arm driven by availability. We did not ask the participants to document oral intake in this trial.
This trial has limitations and strengths. In the 0.9% saline arm, the persistent ketonuria rate was only 11% compared with an anticipated 25% as assumed in the sample size calculation using earlier data.10 This would have reduced the trial's power to demonstrate a difference. The improved ketonuria resolution rate in the 0.9% saline arm of our current trial may be the result of the routine use of P-Trovite multivitamin supplement or a more rigorous and consistent per-protocol administration of the prescribed intravenous rehydration solutions. On the other hand, because the standard deviation in the well-being visual numerical rating scale was 1.5 (compared with a standard deviation of 2 used in sample size calculation), this would have increased the power of the trial to detect a 1-point increase in well-being visual numerical rating scale score to 99.7%. We believe our finding is generalizable for similarly managed inpatient hyperemesis gravidarum populations.
Dextrose saline solution was not superior to normal saline solution in the initial intravenous rehydration of women hospitalized for hyperemesis gravidarum on a range of outcomes. However, because of the theoretical concern of Wernicke's encephalopathy with dextrose infusion when in a thiamine-deficient state, normal saline may be a better choice.
1. Verberg MF, Gillott DJ, Al-Fardan N, Grudzinskas JG. Hyperemesis gravidarum, a literature review. Hum Reprod Update 2005;11:527–39.
2. Tan PC, Jacob R, Quek KF, Omar SZ. The fetal sex ratio and metabolic, biochemical, haematological and clinical indicators of severity of hyperemesis gravidarum. BJOG 2006;113:733–7.
3. Nausea and vomiting of pregnancy. ACOG Practice Bulletin No. 52. American College of Obstetrics and Gynecology. Obstet Gynecol 2004;103:803–16.
4. Gazmararian JA, Petersen R, Jamieson DJ, Schild L, Adams MM, Deshpande AD, et al.. Hospitalizations during pregnancy among managed care enrollees. Obstet Gynecol 2002;100:94–100.
5. Tan PC, Tan NC, Omar SZ. Effect of high levels of human chorionic gonadotropin and estradiol on the severity of hyperemesis gravidarum. Clin Chem Lab Med 2009;47:165–71.
6. Tan PC, Jacob R, Quek KF, Omar SZ. Indicators of prolonged hospital stay in hyperemesis gravidarum. Int J Gynaecol Obstet 2006;93:246–7.
7. Comstock JP, Garber AJ. Ketonuria. In: Walker HK, Hall WD, Hurst JW, editors. Clinical methods: the history, physical, and laboratory examinations. 3rd ed. Boston (MA): Butterworths; 1990.
8. Hasselbalch SG, Knudsen GM, Jakobsen J, Hageman LP, Holm S, Paulson OB. Brain metabolism during short-term starvation in humans. J Cereb Blood Flow Metab 1994;14:125–31.
10. Tan PC, Khine PP, Vallikkannu N, Omar SZ. Promethazine compared with metoclopramide for hyperemesis gravidarum: a randomized controlled trial. Obstet Gynecol 2010;115:975–81.
11. Tan PC, Yow CM, Omar SZ. A placebo-controlled trial of oral pyridoxine in hyperemesis gravidarum. Gynecol Obstet Invest 2009;67:151–7.
12. Tan PC, Omar SZ. Contemporary approaches to hyperemesis during pregnancy. Curr Opin Obstet Gynecol 2011;23:87–93.
13. Niebyl JR. Clinical practice. Nausea and vomiting in pregnancy. N Engl J Med 2010;363:1544–50.
14. Jarvis S, Nelson-Piercy C. Management of nausea and vomiting in pregnancy. BMJ 2011;342:d3606.
15. Jueckstock JK, Kaestner R, Mylonas I. Managing hyperemesis gravidarum: a multimodal challenge. BMC Med 2010;8:46.
16. Levy JA, Bachur RG. Intravenous dextrose during outpatient rehydration in pediatric gastroenteritis. Acad Emerg Med 2007;14:324–30.
17. Reid SR, Losek JD. Rehydration: role for early use of intravenous dextrose. Pediatr Emerg Care 2009;25:49–52.