Mittendorf, Robert MD, DrPH; Covert, Robert MD; Elin, Ronald MD, PhD; Pryde, Peter G. MD; Khoshnood, Babak MD, MPH, PhD; Lee, Kwang‐Sun MD
Department of Obstetrics and Gynecology, Loyola University Medical Center, Maywood, Illinois; Section of Neonatology, Department of Pediatrics, University of Chicago, Chicago, Illinois; Division of Maternal‐Fetal Medicine, Department of Obstetrics and Gynecology, University of Wisconsin, Madison, Wisconsin; and Department of Pathology, University of Louisville, Louisville, Kentucky.
Address reprint requests to: Robert Mittendorf, MD, DrPH, Department of Obstetrics and Gynecology, Loyola University Medical Center, 2160 South First Avenue, Maywood, IL 60153; E‐mail: email@example.com
Funding was provided by the United Cerebral Palsy Research and Educational Foundation, Washington, DC.
Received November 14, 2000. Received in revised form March 2, 2001. Accepted March 8, 2001.
Recently, we raised concern about the safety of tocolytic magnesium sulfate. In a randomized controlled trial, we tested the hypothesis that intravenous magnesium sulfate given to nonpreeclamptic women diagnosed with preterm labor (either as a tocolytic or as a preventive against neurologic injury in those not eligible for tocolysis) could avoid the outcome of neonatal intraventricular hemorrhage or subsequent cerebral palsy. Unexpectedly, interim data monitoring revealed a significant increase in the total pediatric (fetal + neonatal + postneonatal) mortality among those randomized to magnesium treatment.1 Since then, using a case‐control study design in a different data set, we found an association between the total dosages of tocolytic magnesium sulfate exceeding 48 g and perinatal mortality among premature infants weighing 700–1249 g.2
Other investigators have also published reports that raise questions about the safety of tocolytic magnesium sulfate. These observations include elevations of cardiac troponin levels (a well‐established marker of cardiac injury in adults) among human neonates exposed to higher doses of magnesium sulfate,3 dose‐related excess mortality among magnesium‐exposed fetal sheep under conditions of physiologic stress,4 and potentially deleterious physiologic responses related to magnesium sulfate administration in a newborn piglet meconium aspiration model.5
The purpose of this investigation was to explore further the hypothesis of an association between exposures to higher dosages of magnesium sulfate, as used in tocolytic protocols, and total pediatric mortality. Our strategy was to examine a variable that, if this association is causal, would be expected to demonstrate a biologic gradient (in this case, a dose‐response effect of magnesium on its suspected toxic manifestations). The focus of the study, therefore, was to evaluate the relationship between umbilical cord serum ionized magnesium levels and the outcome of total pediatric mortality.
MATERIALS AND METHODS
This study reports the secondary analysis of supplemental data obtained during the conduct of the Magnesium and Neurologic Endpoints Trial. The details of the trial and primary outcomes data have been reported elsewhere.1 The methods portion relevant to the current study is provided. All women admitted with a diagnosis of active preterm labor were screened for eligibility. Mothers having either a primary or coexisting diagnosis of preeclampsia were excluded from the trial because magnesium sulfate is the established standard of care in the United States for seizure prophylaxis in preeclampsia. Furthermore, because the fundamental pathophysiology of preterm labor and preeclampsia are very different, these two diseases cannot be easily studied in the same trial. Also excluded were women with higher‐order multiple gestations (more than twins), because these patients are rare in our patient population and because of the considerable increase in perinatal morbidity in this group compared with singletons or twins. Women found to be eligible were provided a detailed explanation of the study goals and informed consent was requested. The research protocol was approved by the Institutional Review Board of the Biological Sciences Division at the University of Chicago. To ensure a balance among the important variables, mothers were randomized into stratified blocks on the basis of race (black and other) and gestational age (less than 28 weeks and 28 weeks and older). In addition, several months after the beginning of the trial, randomization was also done in stratified blocks on the basis of fetal plurality (twins or singleton).
The treatment portion of the Magnesium and Neurologic Endpoints Trial consisted of two mutually exclusive parts. In the tocolytic portion of the trial, mothers in active preterm labor at less than 34 completed weeks of gestation and with cervical dilatation less than 4 cm were randomized to either tocolytic magnesium sulfate (a 4‐g bolus followed by an infusion of 2–3 g/h) or to another tocolytic agent (ritodrine, terbutaline, indomethacin, or nifedipine) as selected by the attending clinician. In the preventive portion of the trial, mothers in active preterm labor at less than 34 completed weeks of gestation, but in whom the cervix was dilated 4 cm or more (and thus were not eligible for tocolysis), were randomized to receive either a single 4‐g bolus of magnesium sulfate or a comparable volume bolus of saline control.
For all the randomized mothers, we collected data on numerous maternal demographic, obstetric, and neonatal variables. We also attempted, in each case, to obtain several biologic specimens at the time of delivery. These included umbilical cord venous blood that was collected by umbilical venipuncture under direct visualization immediately after cord clamping at delivery. The blood specimens were immediately centrifuged at 1000g for 10 minutes. Aliquots of pipetted sera were then frozen at −70C and stored pending later determination of ionized magnesium levels at the National Institutes of Health (Bethesda, Maryland) using the AVL 988–4 analyzer (Graz, Austria). Technicians and researchers processing the samples were kept unaware of all prior or subsequent health outcomes.
The sample size determinations for the Magnesium and Neurologic Endpoints Trial were based on the anticipated reduction in the occurrence of neonatal intra‐ventricular hemorrhage after the use of antenatal intravenous magnesium sulfate. If the prevalence of neonatal intraventricular hemorrhage was reduced from 18.9% to 4.4% when antenatal intravenous magnesium sulfate was used, as suggested by the observations of Kuban et al,6 then for α = .05, 1 − β (power) = 80%, two tailed, the total number of infants needed would be 140.
Statistical analyses were performed using the unpaired Student t test and Mann‐Whitney U test (Minitab for Windows, Release 11, 1996; Minitab, State College, PA), the Fisher exact test (Stata statistical software, version 5.0, 1997; Stata, College Station, TX), as well as multivariable logistic regression analysis (LogXact 4 For Windows, 1999; CYTEL Software, Cambridge, MA), as appropriate. All tests of statistical significance were two sided, with significance defined as α = .05.
We screened 194 pregnant women who presented in labor with a gestational age of less than 34 completed weeks for enrollment in the trial. Of these, we found 157 were eligible; 149 mothers gave written, informed consent and were randomized into one of the four arms in the trial. The mother, not the child, was the unit of randomization. In the tocolytic arms of the study, 46 mothers (37 singletons + 9 pairs of twins = 55 fetuses) were randomized to magnesium sulfate and 46 (41 singletons + 5 pairs of twins = 51 fetuses) were randomized to other tocolytic agents. In the preventive arms, 29 mothers (28 singletons + 1 pair of twins = 30 fetuses) were randomized to a 4‐g bolus of magnesium sulfate and 28 (27 singletons + 1 pair of twins = 29 fetuses) were randomized to a bolus of saline control.
Among the 159 children whose mothers remained in the trial through delivery, and for whom we know the mortality outcome status, 82 individuals had umbilical cord blood specimens adequate for analysis. The remaining cases included 55 neonates whose cord blood specimens were not collected at the time of delivery and 22 whose collected specimens were not adequate for analysis because of an insufficient quantity (n = 18) or hemolysis (n = 4). No significant difference (Student t test, P = .57) was found in the mean birth weight between the group of children with assessable ionized magnesium levels (1781 g) and the group for whom the ionized magnesium level could not be determined (1848 g). This similarity suggests that the absence of assessable specimens appears to be a random event or one in which selection bias related to birth weight or its correlates appears not to be a factor.
Among the 82 children for whom we had ionized magnesium levels (the difficulty in obtaining blood from the umbilical cords of the smallest newborns limited the number of assessable specimens), seven died (one immediately before delivery, three in the neonatal period, and three in the postneonatal period). Six of these deaths occurred in the tocolytic arms. The median ionized magnesium level from the group having the seven deaths was 0.76 mmol/L; in the group of 75 survivors (47 exposures to magnesium sulfate and 28 unexposed to magnesium sulfate), the median ionized magnesium level was only 0.55 mmol/L. This difference (0.76 mmol/L compared with 0.55 mmol/L) was statistically significant (Mann‐Whitney U test, P = .03). Information on the maternal demographic variables, including age, race, parity, marital status, payor status, cigarette smoking, and use of cocaine during pregnancy is shown in Table 1. No significant differences were found between the children who died and those who lived. Among the obstetric variables (gestational age, preterm premature rupture of membranes, use of antenatal steroids, presence of meconium at delivery, need for cesarean section, birth weight, Apgar scores, fetal plurality [twins], and level of ionized magnesium in the umbilical cord venous serum at delivery) (Table 2), the ionized magnesium level was the only variable to reach statistical significance, although fetal plurality (twins) reached near significance.
Because the birth weight is such an important predictor of infant mortality, we further evaluated the relationship between the ionized magnesium level and mortality while controlling for birth weight in a multivariable logistic regression model. To this end, birth weight categories were formed as follows: less than 1500 g, 1500–2499 g, and 2500 g and greater (Table 3). We chose 0.70 mmol/L as the fulcrum of the dichotomy in the ionized magnesium levels because 0.70 mmol/L was the midpoint between the group of children who died after exposure to magnesium (n = 6, 0.76 mmol/L) and those who survived the exposure to magnesium (n = 47, 0.64 mmol/L). Among the neonates whose umbilical cord blood ionized magnesium level was 0.70 mmol/L or less, 3.3% (2 of 61) died; among those whose ionized magnesium level was greater than 0.70 mmol/L, 23.8% (5 of 21) died (Fisher exact test, P = .01). This association remained significant (adjusted odds ratio [OR] 7.0, 95% confidence interval [CI] 1.02, 79.6, P < .05) when controlling for birth weight, as noted above (LogXact 4 For Windows, 1999; CYTEL Software, Cambridge, MA). Furthermore, to rule out the possibility of confounding by fetal plurality (twins), we did a multivariable logistic regression of twins, dichotomized cord ionized magnesium levels, and dichotomized birth weight (less than 1500 g versus 1500 g or more) (Table 4). In this regression, the only statistically significant, independent predictor of death was the higher levels of umbilical cord ionized magnesium at delivery (adjusted OR = 7.7, 95% CI 1.2, 47.6, P = .03). Thus, children whose ionized magnesium levels are greater than 0.70 mmol/L at delivery are seven times more likely to die than children whose magnesium levels are lower.
As previously reported, using an intent‐to‐treat analysis,1 the total pediatric mortality difference between those fetuses exposed to magnesium sulfate and those not exposed in the Magnesium and Neurologic Endpoints Trial was statistically significant (risk difference 10.7%, 95% CI 2.9%, 18.5%; two‐sided Fisher exact test, P = .02). In the high‐dose tocolytic arms, the risk difference was 15.2% (95% CI 4.4%, 26.0%; two‐sided Fisher exact test, P = .01). In this secondary analysis of supplementary data collected in that study, we found a biologic gradient in the apparent relationship between the magnesium level and pediatric deaths. This apparent dose‐response effect of magnesium sulfate requires careful consideration in the ongoing controversy regarding its safety in the high dosages used for the purposes of tocolysis.
Because of the well‐established efficacy of magnesium sulfate for seizure prophylaxis in preeclamptic women, as well as its possible role as a neuroprotective agent at the lower dosages currently under investigation for that purpose in large randomized clinical trials, it would be inappropriate for us to suggest that the obstetric indications for magnesium sulfate be abandoned. However, it is important to emphasize the failure of existing trials and meta‐analyses to demonstrate any measure of tocolytic efficacy using magnesium sulfate.7,8 In this light, the popularity and widespread use of magnesium sulfate for the treatment of premature labor must be questioned.
Although our data do not definitively establish causality, the evidence supporting a possible deleterious role of magnesium sulfate in the extraordinarily high dosages used for tocolysis is growing. Given our previous experience,1,2 and the additional research findings reported here, the conduct of a randomized controlled trial testing the hypothesis of an association between tocolytic magnesium sulfate and total pediatric mortality clearly cannot be justified at our own institutions. However, we strongly agree with Hannah's recent editorial observation that current data regarding tocolytic agents are inadequate and her appeal for well‐designed randomized clinical trials focusing not only on outcomes‐based indicators of efficacy, but on maternal and fetal safety issues as well (Hannah ME. Search for best tocolytic for pre‐term labour. Lancet 2000;356:699–700). In this regard, we suggest that centers remaining committed to magnesium sulfate as a tocolytic agent conduct a randomized trial evaluating not only its heretofore unproved tocolytic efficacy but, more importantly, focusing on its impact on pediatric short‐ and long‐term outcomes. Unarguably, it is important to prove that a drug has a significant beneficial effect before any potential for toxicity can be considered acceptable.
1. Mittendorf R, Covert R, Boman J, Khoshnood B, Lee KS, Siegler M. Is tocolytic magnesium sulphate associated with increased total paediatric mortality? Lancet 1997;350:1517–8.
2. Scudiero R, Khoshnood B, Pryde PG, Lee KS, Mittendorf R. Perinatal death and tocolytic magnesium sulfate. Obstet Gynecol 2000;96:178–82.
3. Shelton SD, Fouse BL, Holleman CM, Sedor FA, Herbert WNP. Cardiac troponin T levels in umbilical cord blood. Am J Obstet Gynecol 1999;181:1259–62.
4. Reynolds JD, Chestnut DH, Dexter F, McGrath J, Penning DH. Magnesium sulfate adversely affects fetal lamb survival and blocks fetal cerebral blood flow response during maternal hemorrhage. Anesth Analg 1996;83:493–9.
5. Barrington KJ, Ryan CA, Finer NN. Effects of magnesium sulfate in a newborn piglet meconium aspiration model. J Perinatol 2000;20:373–8.
6. Kuban KCK, Leviton A, Pagano M, Fenton T, Strassfeld A, Wolff M. Maternal toxemia is associated with reduced incidence of germinal matrix hemorrhage in premature babies. J Child Neurol 1992;7:70–6.
7. Cox SM, Sherman ML, Leveno KJ. Randomized investigation of magnesium sulfate for prevention of preterm birth. Am J Obstet Gynecol 1990;163:767–72.
8. Gyetvai K, Hannah ME, Hodnett ED, Ohlsson A. Tocolytics for preterm labor: A systematic review. Obstet Gynecol 1999;94:869–77.