Is STAN monitoring associated with a significant decrease in metabolic acidosis at birth compared with cardiotocography alone? Review of the three meta‐analyses that included the recent US trial
C. Vayssière,V. Ehlinger,L. Paret,C. Arnaud
DOI: https://doi.org/10.1111/aogs.12923
2016-10-01
Acta Obstetricia Et Gynecologica Scandinavica
Abstract:Sir, We read with interest the recent meta-analysis (MA) by Blix et al. (1) reporting that intrapartum surveillance with STAN monitoring resulted in a significant decrease in metabolic acidosis at birth compared with cardiotocography (CTG) alone (OR 0.64, 95% CI 0.47–0.88). This finding conflicts with the Cochrane MA by Neilson et al. (2) and MA by Saccone et al. (3) which found no decrease in metabolic acidosis with STAN monitoring (RR 0.72, 95% CI 0.43–1.20 and RR 0.81, 95% CI 0.44–1.46, respectively). The conflicting results in these three recent MAs based on the same six trials including the last US trial (4), require explanation. Two main reasons could be suggested. First, revised data for two of the six trials (the Swedish and Dutch trials) were published years later (5,6). Only Blix et al. used the revised data appropriately for a more precise MA of STAN monitoring (1). Moreover, the Finnish trial, unlike the others, presented corrected data for calculating metabolic acidosis using base deficit in the extracellular-fluid, required for comparability. Secondly, the MAs used different statistical methods: Blix et al. (1) performed a trial sequential analysis (TSA), whereas the others (2,3) did traditional aggregate MAs. We would like to clarify whether these contradictory findings are due to the use of different data, the type of statistical analysis or both. Do the trials published so far show that STAN monitoring decreases the risk of metabolic acidosis at birth significantly? The differences between the statistical methods employed lie in the parameters estimated (OR vs. RR), the weights given to each study’s data (Mantel–Haenszel vs. inverse variance method), and the recognition of the heterogeneity between studies (fixed-effects vs. random-effects model) while performing MA. As Blix et al. stated, each method employed has its pros and cons, especially in an MA with moderate heterogeneity between the studies. We conducted an MA of the revised data using five different MA methods available in STATA 14 software (metan function): fixed-effects model with Mantel–Haenszel method, fixed-effects model with inverse variance, random effects model with DerSimonian & Laird method with heterogeneity estimated by the Mantel–Haenszel model, random-effects model with DerSimonian & Laird method and heterogeneity estimated by the fixed-effects inverse variance model, and Peto’s method for pooling ORs. Table 1 provides the results for metabolic acidosis by each method. Although the results of the random-effects model were at the borderline of significance, with an upper confidence interval limit of 0.998, the results were consistent and significant with p-values <0.05 with all five methods. In conclusion, applying the revised and comparable data produces a significant 35% decrease in metabolic acidosis at birth compared with traditional CTG monitoring, as Blix et al. reported. Unlike them, we think that metabolic acidosis, a combination of a low pH at birth and base deficit >12 mmol/L, remains one of the best indicators available at birth for the immediate assessment of fetal well-being and the risk of hypoxic-ischemic encephalopathy (7). By making it possible to reduce the incidence of metabolic acidosis by one-third, as well as by better standardizing decisions for immediate operative delivery, the use of STAN for intrapartum surveillance of women at risk appears to be a valuable alternative to CTG alone. Table 1. Results for metabolic acidosis in all 6 trials published comparing STAN to CTG alone, applying the revised data published*.