Description of studies

This 2015 review update identified three new randomised controlled trials by Kunst 1999, El‐Gohary 2004 and Das 2005, assessing the impact of regional versus general anaesthesia in preterm infants undergoing inguinal herniorrhaphy in addition to the original four randomised controlled trials (Welborn 1990; Krane 1995; Somri 1998; Williams 2001). Krane 1995 is included in the review although some of the data were incomplete and we were not able to obtain this information via personal communication. Two of the newly identified studies are awaiting further classification: Das 2005 did not specify the infant population as preterm and reported a minimum age at operation of less than two months; and Kunst 1999 was excluded in the original version of this review as the published results were of insufficient detail. The newly identified El‐Gohary 2004 trial qualified for inclusion in this review: similar to the other studies in this review, this was published as a complete article.

Details of each study are given in the table 'Characteristics of included studies'. Overall, the studies attempted to include ex‐preterm infants. The inclusion criteria, the intervention (type of anaesthetic agent) and outcomes varied between studies.

All studies compared single‐shot spinal versus general anaesthesia in premature infants (less than 37 weeks) undergoing inguinal herniorrhaphy.

Krane 1995 enrolled 18 infants born at less than 36 weeks' gestation who were post‐term but less than 60 weeks' corrected gestational age at time of surgery, with an American Society of Anesthesiologists (ASA) physical class of I or II, undergoing inguinal herniorrhaphy. The authors excluded significant chronic lung disease, known symptomatic congenital heart disease or symptomatic central nervous system disease. Preoperative respiratory function of the included infants was assessed by a 12 hour study of respiratory rate, haemoglobin oxygen saturation and ECG in the infant's home. On the day of surgery 10 infants received a single shot of spinal anaesthetic with 0.6 mg/kg tetracaine hydrochloride made hyperbaric with an equal volume of 10% dextrose. Three infants in the spinal group who cried during surgical manipulation received brief mask blow‐over of 50% nitrous oxide (N₂O). In the general anaesthesia group, eight infants received balanced anaesthesia, with halothane and muscle relaxation. At the end of the surgery the ilioinguinal nerve was infiltrated with 0.25% bupivacaine for pain relief. When necessary, at the conclusion of surgery, neuromuscular blockade was antagonised with neostigmine and atropine. One infant in the spinal anaesthetic group had inadequate spinal anaesthetic blockade and was then given a general anaesthetic in addition to a local nerve block. The outcome measures of postoperative apnoea (cessation of respiratory movement for more than 10 seconds), bradycardia and periodic breathing were reported to be equal in each group, but individual data were not supplied. Postoperative haemoglobin oxygen desaturation (a decrease in peripheral oxygen saturation (SpO₂ ) to less than 90% for 10 seconds or more), anaesthetic placement failure and anaesthetic agent failure were measured and reported for the two groups. Postoperative pain relief in both groups was managed with paracetamol and no opioids were administered.

Somri 1998 enrolled 40 infants born at less than 37 weeks' gestation who were post‐term but less than 60 weeks' corrected gestational age at time of surgery and with one of the following: a history of neonatal respiratory distress syndrome, bronchopulmonary dysplasia documented by radiography or a history of preoperative apnoea. They did not exclude any infants on entry criteria but excluded from analysis those infants in whom failure to achieve spinal needle placement occurred. On the day of surgery, 20 infants received a spinal anaesthetic with isobaric bupivacaine 0.6 to 0.8 mg/kg. A pacifier was used to soothe restless infants. In the general anaesthesia group 20 infants received thiopentone intravenous induction, 4 to 5 mg/kg, followed by balanced anaesthesia of halothane 0.5% to 1.5% and muscle relaxation with atracurium, 0.5 mg/kg. Neuromuscular blockade was reversed at the end of surgery with neostigmine and atropine. The outcome measures were postoperative apnoea (defined as a respiratory pause of 15 seconds or longer; or less than 15 seconds with bradycardia (defined as a heart rate of less than 100 beats/minute)); postoperative analgesia; postoperative respiratory support; average length of hospital stay; anaesthetic placement failure; and multiple spinal placement attempts. Six infants in the spinal anaesthesia group became restless despite successful spinal blockade and supplementary anaesthesia was administered to obtain sufficient motionlessness for completion of the operative procedure: four infants received inhaled halothane and two received intravenous propofol. Paracetamol was the only analgesic agent used in the postoperative period.

Welborn 1990 (the third study) enrolled 36 infants born at less than 37 weeks' gestation who were post‐term but less than 51 weeks' corrected gestational age at time of surgery and who were healthy, with an ASA physical status I or II. The authors excluded infants who had cardiac, neurological or metabolic disease and those who were receiving methylxanthines or caffeine. In the spinal anaesthesia group, during the early part of the study, nine infants received intramuscular ketamine for sedation prior to performance of lumbar puncture. These results were presented separately. As the authors gained experience the ketamine sedation was stopped due to the high incidence of side effects observed. The remaining 11 infants in the spinal group were comforted with sugar solution. In all infants the spinal anaesthesia was achieved by 1% tetracaine 0.4 to 0.6 mg/kg with an equal volume of 10% dextrose. The 16 infants in the general anaesthesia group received a balanced anaesthetic with halothane and neuromuscular blockade. Neuromuscular blockade was reversed at the end of surgery with neostigmine and atropine. The outcomes measured were episodes of postoperative apnoea (defined as a respiratory pause of less than 15 seconds not associated with bradycardia); and prolonged episodes of apnoea (defined as a respiratory pause of 15 seconds or longer; or less than 15 seconds accompanied by bradycardia (defined as a heart rate less than 100 beats per minute for at least 5 seconds). Other outcomes measured included postoperative respiratory support, length of surgery and temperature on admission to the recovery room.

Williams 2001 (the fourth study) enrolled 28 infants born at less than 36 weeks' gestation, who were post‐term but less than 46 weeks' postconceptional age at time of surgery. The authors excluded infants with pre‐existing cardiac, neuromuscular or metabolic diseases. Pre‐existing abnormal respiratory function with or without the need for supplemental oxygen therapy was noted. On the day of surgery 14 infants received a spinal anaesthetic consisting of 0.5% bupivacaine 1 mg/kg. The 14 infants in the general anaesthesia group received a 2 minimum alveolar concentration (MAC) equivalent value for age of sevoflurane in 100% oxygen and muscle relaxation with atracurium. At the end of surgery, neuromuscular blockade was antagonised with neostigmine and glycopyrrolate. Both groups received a single injection of bupivacaine 0.25% 2 mg/kg into the caudal epidural space for analgesic effects extending into the postoperative period. The outcome measure was postoperative apnoea (a sustained respiratory pause of 15 seconds or longer, or less than 15 seconds if accompanied by an SpO₂ less than 90% or bradycardia). Bradycardia was defined as a heart rate of less than 100 beats/minute for at least five seconds. Anaesthetic placement failure was also measured.

El‐Gohary 2004 (the fifth study) enrolled 30 infants born at less than 37' weeks gestation who were post‐term but less than 46 weeks' postconceptional age at time of surgery. The authors excluded infants with pre‐existing cardiac, neuromuscular or metabolic diseases; coagulopathy; systemic infection; meningitis; infection at the site of puncture; and cases of inadequate block. On the day of the surgery, 15 infants received a single dose of EMLA cream 0.5 g applied to the skin over the sacrococcygeal membrane 90 minutes before receiving a spinal anaesthetic consisting of 1 mL/kg of 0.375% ropivacaine. The 15 infants in the general anaesthesia group received 2 MAC sevoflurane in 100% oxygen. Restlessness or discomfort in the presence of adequate block was alleviated with mask inhalation of nitrous oxide. The outcome measures were episodes of pre and post‐operative apnoea (a sustained respiratory pause of 15 seconds or longer; or less than 15 seconds if accompanied by an SaO₂ less than 90%; or bradycardia, where bradycardia was defined as a heart rate of less than 100 beats/minute for at least 5 seconds; and desaturation was defined as an SaO₂ of less than 90% for more than 10 seconds).

Ongoing studies: the GAS study is a multi‐centre trial that has recently completed recruitment of 722 infants undergoing inguinal herniorrhaphy repair (NCT0075660). Infants were randomised to regional anaesthesia versus combined general/regional anaesthesia. In addition to neurosensory outcomes at two years' corrected age, investigators intend to follow longer term intelligence quotients of study infants at five years' corrected age.

Risk of bias in included studies

See: Characteristics of included studies table.

Overall, the studies included in this review were of inadequate quality. Due to the nature of the intervention contrast, it was deemed impractical to blind the intervention in any of the studies. Three of the five studies attempted to account for this by using blinded assessment. Specific methodological issues of individual trials are discussed below:

Krane 1995
Randomisation in this study was not described adequately. There was no mention of allocation concealment and there was no blinding of intervention but there was adequate blinding of the outcome, with the results being interpreted by a trained technician unaware of the treatment group assignment. There was complete follow‐up of all enrolled infants. One infant with a failed spinal anaesthetic who subsequently received a general anaesthetic was counted as a member of the general anaesthetic group. We have re‐analysed this according to intention‐to‐treat analysis by including the infant in the spinal group. Unfortunately, despite contacting the author, further clarification of this infant was not available.

Somri 1998
Randomisation in this study was inadequately described. Allocation concealment was not mentioned and there was no blinding of either intervention or outcome. There was complete follow‐up of all enrolled infants. Bias was introduced by omitting three infants in which lumbar puncture was unsuccessful and one infant was excluded because of hypothermia.

Welborn 1990
Randomisation in this study was inadequately described. Allocation concealment was not mentioned and there was no blinding of intervention, but a pulmonologist unaware of treatment group assessed outcome. There was complete follow‐up. Subgroup analyses were performed, with and without pre‐anaesthetic ketamine sedation, in the spinal group, following initial concerns by study investigators that pre‐anaesthetic sedation contributed to greater incidence of postoperative apnoea. Results were reported by the investigators in three groups: general anaesthesia, spinal plus ketamine, and spinal without ketamine.

Williams 2001
Randomisation in this study was adequately described with the use of random number tables. There was no mention of allocation concealment and there was no blinding of intervention but the results were analysed by a blinded observer. There was complete follow‐up. Four infants were excluded following randomisation due to anaesthetic failure in the spinal anaesthesia group. These four infants were given a general anaesthetic but were excluded from further analysis by the authors.

El‐Gohary 2004
Randomisation in this study was inadequately described. There was no mention of allocation concealment, blinding of intervention or outcome assessment. Although the number analysed equalled those initially randomised, follow‐up is assessed as incomplete due to the exclusion of an unspecified number of cases with inadequate block. Whether cases of inadequate block were excluded pre‐ or postrandomisation was not mentioned. Two cases in the spinal group requiring nitrous oxide via mask inhalation were analysed as intention to treat.

Effects of interventions

See: Summary of findings for the main comparison Regional (spinal, epidural, caudal) versus general anaesthesia for preterm infants having inguinal herniorrhaphy in early infancy

All five studies in this review reported on the impact of spinal versus general anaesthesia in preterm infants undergoing inguinal herniorrhaphy. The five studies reported different postoperative outcomes. Where different studies reported similar outcomes there was no significant heterogeneity of treatment effect between the studies. However, due to the small number of studies and participants in a proportion of the meta‐analyses, a complete lack of heterogeneity cannot be assumed.

Spinal anaesthesia versus general anaesthesia (Comparison 1)

Primary outcome measures

Apnoea/bradycardia (Analysis 1.1; Figure 1): four studies adequately reported on apnoea/bradycardia. In all four studies the definition of apnoea did not fulfil the NIH 1987 definition of a pause in breathing of greater than 20 seconds or one less than 20 seconds and associated with cyanosis, marked pallor, hypotonia or bradycardia. El‐Gohary 2004, Somri 1998, Williams 2001, and Welborn 1990 defined apnoea as a pause in breathing of 15 seconds, or less if associated with bradycardia. Krane 1995 defined apnoea as a pause greater than 10 seconds, or less if associated with bradycardia, but the results were not adequately reported. Only Somri 1998 demonstrated a statistically significant reduction in the risk of apnoea/bradycardia among infants having received a spinal anaesthetic (RR 0.12, 95% CI 0.02 to 0.89). The remaining three studies showed no significant difference between intervention groups. Meta‐analysis showed no statistically significant difference between the intervention groups (typical RR 0.72, 95% CI 0.48 to 1.06; 4 studies, 138 infants).

Figure 1

---

---

Forest plot of comparison: 1 Spinal anaesthesia versus general anaesthesia, outcome: 1.1 Apnoea/bradycardia occurring 12 to 24 hours following completion of operation.

Any oxygen desaturation (Analysis 1.2; Figure 2): two studies reported on oxygen desaturation (Krane 1995; El‐Gohary 2004). Neither study, alone or in combination, showed a statistically significant difference in the rate of oxygen desaturation between infants who received a spinal anaesthetic and those who received a general anaesthetic (typical RR 0.82, 95% CI 0.61 to 1.11; 2 studies, 48 infants).

Figure 2

---

---

Forest plot of comparison: 1 Spinal anaesthesia versus general anaesthesia, outcome: 1.2 Any oxygen desaturation occurring 12‐24 hours following completion of the operation.

Postoperative respiratory support (Analysis 1.3): of three studies (Krane 1995; Somri 1998; Welborn 1990) reporting this outcome; only Somri 1998, reported the need for respiratory support for longer than one hour postoperatively in four infants having received a general anaesthetic compared to no infants having received spinal anaesthesia. In the meta‐analysis there was a reduced need for respiratory support in infants receiving spinal anaesthesia but this was not statistically significant (typical RR 0.09, 95% CI 0.01 to 1.64; 3 studies, 98 infants).

Postoperative apnoea with preoperative sedatives excluded (Analysis 1.4; Figure 3): four studies reported postoperative apnoea in infants who had not received preoperative sedatives (Welborn 1990; Somri 1998; Williams 2001; El‐Gohary 2004). Only Somri 1998 showed a significant reduction in risk of apnoea/bradycardia among infants who received spinal anaesthetic (RR 0.12, 95% CI 0.02 to 0.89). Welborn 1990 ceased the usage of preoperative ketamine because of apnoea: once ketamine was ceased the incidence of apnoea was dramatically reduced. We have performed subgroup analyses with and without the ketamine group. The effect on postoperative apnoea was statistically significant in the subgroup of infants that had not received preoperative ketamine (typical RR 0.53, 95% CI 0.34 to 0.82; 4 studies, 129 infants). The use of spinal rather than general anaesthesia was associated with a 47% reduction in the risk of postoperative apnoea for those infants not receiving any sedation. The number needed to treat to benefit (NNTB) from receiving spinal rather then general anaesthesia in preventing an episode of postoperative apnoea was approximately four infants (4.40).

Figure 3

---

---

Forest plot of comparison: 1 Spinal anaesthesia versus general anaesthesia, outcome: 1.4 Postoperative apnoea with preoperative sedatives excluded.

Postoperative apnoea with no episodes of preoperative apnoea (Analysis 1.5; Figure 4): this subgroup was determined post hoc. Four studies reported postoperative apnoea in infants with no history of preoperative apnoea (Welborn 1990; Somri 1998; Williams 2001; El‐Gohary 2004). This outcome did not reach statistical significance in any of the studies taken separately. In the updated review and meta‐analysis, there was a statistically significant reduction in the risk of postoperative apnoea in preterm infants without a history of preoperative apnoea and having received a spinal anaesthetic (typical RR 0.34, 95% CI 0.14 to 0.81; 4 studies, 134 infants). The use of spinal rather than general anaesthesia was associated with a 66% reduction in the risk of postoperative apnoea for those infants without preoperative apnoea.The number needed to treat to benefit (NNTB) from receiving spinal rather then general anaesthesia in preventing an episode of postoperative apnoea in infants with no history of preoperative apnoea was approximately six infants (5.72).

Figure 4

---

---

Forest plot of comparison: 1 Spinal anaesthesia versus general anaesthesia, outcome: 1.5 Postoperative apnoea with no preoperative apnoea.

Neurodevelopmental state at two year follow‐up as reflected in the Bayley scales of infant development (MDI/PDI): none of the trials in this review reported on neurodevelopmental state at two year follow‐up or any other neurological outcomes.

Secondary outcome measures

Anaesthetic agent failure (Analysis 1.6; Figure 5): three studies assessed anaesthetic agent failure (El‐Gohary 2004; Krane 1995; Somri 1998). Although each study indicated a higher incidence of agent failure in the spinal group, none of these studies when considered separately reached statistical significance. In the meta‐analysis, infants who received a spinal anaesthetic had a statistically significantly greater risk of failure of anaesthesia arising from the anaesthetic agent, compared to those receiving general anaesthesia (typical RR 7.83, 95% CI 1.51 to 40.58; 3 studies, 92 infants). One spinal anaesthetic agent failure occurred for every four successfully placed spinal anaesthetics administered (typical RD 0.24, 95% CI 0.11 to 0.38), requiring the use of sedatives or inhalational anaesthesia or both.

Figure 5

---

---

Forest plot of comparison: 1 Spinal anaesthesia versus general anaesthesia, outcome: 1.6 Anaesthetic agent failure.

Repeated attempts to achieve successful anaesthetic placement (Analysis 1.7): in the study by Somri 1998, infants undergoing a spinal anaesthetic were more likely to undergo multiple attempts to achieve accurate placement of the spinal needle. This effect was of borderline statistical significance (RR 9.24, 95% CI 0.54 to 157.57; RD 0.21, 95% CI 0.03 to 0.38).

Anaesthetic placement failure (Analysis 1.8; Figure 6): three studies reported on anaesthetic placement failure (Krane 1995; Somri 1998; Williams 2001). None of the included studies alone reached statistically significant difference in the proportion of infants with failure of spinal placement between regional anaesthesia and general anaesthesia groups. In the study by Krane 1995, no infant in either group had anaesthetic placement failure. Overall, there was an increased risk of anaesthetic placement failure in the spinal group of borderline statistical significance (typical RR 7.38, 95% CI 0.98 to 55.52; typical RD 0.15, 95% CI 0.03 to 0.27; 3 studies, 90 infants). There was one spinal needle placement failure for every seven infants in whom spinal anaesthesia was intended.

Figure 6

---

---

Forest plot of comparison: 1 Spinal anaesthesia versus general anaesthesia, outcome: 1.8 Anaesthetic placement failure.

Use of postoperative analgesic agents (Analysis 1.9): the only study to report on this outcome, Somri 1998, showed no significant difference in the use of postoperative analgesic agents between infants receiving spinal anaesthesia compared to those receiving general anaesthesia (RR 0.42, 95% CI 0.15 to 1.18; 1 study, 44 infants). None of the studies used opioids or other respiratory depressants as analgesic agents in the postoperative period. All studies used paracetamol for pain relief.

Duration of surgery (Analysis 1.10): two studies reported this outcome. Welborn 1990 reported no statistically significant difference between the treatment groups; however, insufficient data were supplied to allow calculation in this systematic review. The study by El‐Gohary 2004 showed no statistically significant difference between spinal and general anaesthesia groups (RD ‐2.00, 95% CI ‐5.24 to 1.24).

Average length of hospital stay: one study reported on the length of hospital stay with no significant difference between the treatment groups (Somri 1998). Reporting of data were insufficient to include in the meta‐analysis.

Temperature on admission to recovery: one study reported on temperature on admission to recovery (Welborn 1990). The mean temperature on admission to recovery, 36.5 °C, was the same between the two intervention groups.

With the exception of the subgroups receiving preoperative sedatives and those infants without episodes of preoperative apnoea, other prespecified subgroup analyses were not possible due to inadequate reporting of outcome data and the small numbers of studies and participants.