Intrathecal catheterisation after observed accidental dural puncture in labouring women: update of a meta-analysis and a trial-sequential analysis

Highlights

• After observed dural puncture, intrathecal catheter insertion has been advocated.

• Conventional meta-analysis shows post-puncture headache and blood patches are reduced.

• Trial-sequential analysis tests for the presence of firm evidence in meta-analyses.

• Trial-sequential analysis reveals insufficient evidence for drawing firm conclusions.

• We call for an evidence-based clinical practice guideline.

Abstract

Background

Our meta-analysis from 2013 showed that inserting a catheter intrathecally after an observed accidental dural puncture can reduce the need for epidural blood patch in labouring women requesting epidural analgesia. We updated our conventional meta-analysis and added a trial-sequential analysis (TSA).

Methods

A systematic literature search was conducted to identify studies that compared inserting the catheter intrathecally with an epidural catheter re-site or with no intervention. The extracted data were pooled and the risk ratio (RR) and 95% confidence interval (95%CI) for the incidence of post-dural puncture headache (PDPH) was calculated, using the random effects model. A contour-enhanced funnel plot was constructed. A TSA was performed and the cumulative Z score, monitoring and futility boundaries were constructed.

Results

Our search identified 13 studies, reporting on 1653 patients, with a low risk of bias. The RR for the incidence of PDPH was 0.82 (95%CI 0.71 to 0.95) and the RR for the need for epidural blood patch was 0.62 (95%CI 0.49 to 0.79); heterogeneity of both analyses was high. The TSA showed that the monitoring or futility boundaries were not crossed, indicating insufficient data to exclude a type I error of statistical analysis. Contour-enhanced funnel plots were symmetric, suggesting no publication bias.

Conclusions

Conventional meta-analyses showed for the first time that intrathecal catheterisation can reduce the incidence of PDPH. However, TSA did not corroborate this finding. Despite increasing use in clinical practice there is no firm evidence on which to base a definite conclusion.

Introduction

Labour epidural analgesia can be complicated by accidental dural puncture¹ which often leads to post-dural puncture headache (PDPH).2, 3 In the case of an observed dural puncture the catheter can be inserted intrathecally,⁴ an approach that has become increasingly popular over the last two decades and that was favoured by more than two-thirds of lead obstetric anaesthetists in the UK, according to a survey in 2013.⁵ The catheter can plug the dural hole, thereby limiting the efflux of cerebrospinal fluid,⁶ which otherwise results in traction on pain-sensitive structures. Alternatively, epidural puncture and catheter insertion may be repeated at the same or a different interspace which, however, carries the risk of a second dural puncture.

In 2013 we published a meta-analysis and found a significant reduction in the need for epidural blood patch (EBP), but not in the incidence of headache, from intrathecal catheterisation (ITC) after accidental dural puncture.⁷

This study updates the evidence by including relevant literature from the past five years. Moreover, we performed trial-sequential analyses (TSAs) that test for the presence or absence of sufficient evidence for the conclusions obtained with conventional meta-analytical methods. Recent investigations showed that conventional meta-analyses alone were often premature and bear the risk of false-positive results.⁸

Our meta-analysis followed the STROBE statement.⁹ The study was registered with PROSPERO (CRD42018118403). We performed a literature search on December 12th, 2018 in the databases PubMed, Google Scholar, Cochrane Library, Embase, Web of Science with the following search terms: (Spinal or intrathecal or subarachnoid) and (catheter or anaesthesia or analgesia) or continuous spinal and (Spinal or intrathecal or subarachnoid) and (catheter or anaesthesia or analgesia) and (Inadvertent or unintentional or accidental) dural puncture and postdural puncture headache or epidural blood patch. We also screened the bibliography of the retrieved articles for further relevant references. We aimed to be as inclusive as possible and therefore also included abstracts.

Studies were included according to the PICO acronym: P(atients): parturients scheduled for labour analgesia or anaesthesia for caesarean section; I(ntervention): inserting an intrathecal catheter upon observed accidental dural puncture; C(ontrol): epidural catheter re-site at the same or another interspace; O(utcomes): occurrence of PDPH, need for EBP. Additional secondary outcomes were the incidence of caesarean section in the labour analgesia studies, and side-effects.

The methodological quality of observational studies was assessed with the ROBINS-I tool¹⁰ and the quality of prospective randomised trials with the Cochrane risk of bias tool.¹¹ Two researchers (MH, MK) independently screened the articles retrieved from the literature search for eligibility, and two researchers (MH, NH) performed the quality assessment and extracted the data (MH, CvdM).

Conventional meta-analysis: For the dichotomous outcomes (such as incidence of PDPH, need for EBP and incidence of caesarean section) risk ratios (RRs) and 95% confidence intervals (95%CI) were calculated. A condition for calculating pooled effect estimates was the presence of at least 100 patients per treatment arm and at least three studies. If there were more than two ITC groups, we combined the data of the groups by simple addition. We applied a random effects model. I² statistics were computed to assess heterogeneity.

Trial-sequential analysis: Trial-sequential analysis is a statistical approach that allows the power of meta-analytical calculations to be assessed and it therefore controls for type I and II errors of statistical analysis. Details have been described previously.¹² Briefly, a cumulative Z-curve (the Z-test value at each meta-analysis update), the conventional level of significance (e.g. Z-score=1.96 for a P-value threshold of 5%, independent of the quantity of evidence that has accumulated), the number of patients in the meta-analysis, the estimated required information size, and the trial sequential significance and futility boundaries are constructed. The TSA significance boundaries adjust the thresholds for significance and the risk of type 1 error is less than 5%. If the Z curve crosses the monitoring and futility boundaries, then there is sufficient evidence in the individual analyses.

Contour-enhanced funnel plots were performed if there were more than 10 studies. Plot asymmetry is an indicator of the existence of publication bias. We used the statistics programs RevMan for conventional meta-analysis, TSA software (version 0.9.5.10 Beta. Copenhagen Trial Unit, Copenhagen, Denmark) for TSA and R Studio version 1.0.136 for contour-enhanced funnel plots.