Introduction

The question whether there is damage to neurodevelopment after receiving a single anesthesia has frequently been queried from the family of children, while surgeons and anesthesiologists also do not answer it properly. Sometimes, the doctor would say it is an uncertain conclusion in children. However, preclinical studies have established that anesthesia is toxic to the brain in neonatal animals [1]. In addition, animals exposed early in development have documented abnormal attention, learning and memory, and behavior changes [2]. Although the answer to the question is clear on animal studies, the evidence lack of human trials, have made conclusions regarding human brain development after exposure highly tentative. Until now, the results of currently available clinical and epidemiological studies remain inconclusive to guide clinical decision-making [3, 4]. To our knowledge, clinical research investigating the neurodevelopmental outcome of children after single general exposure to anesthesia within age of four and three has not been well established. Therefore, we performed a meta-analysis and meta-regression of cohort study with accumulating evidence to discuss the topic of outcome measurement after a single anesthesia administered before 3 and 4 years.

Methods

Ethical approval and patient consent are not required since this is a systematic review and meta-analysis of previously published studies. The systematic review and meta-analysis were conducted and reported in adherence to preferred reporting items for systematic reviews and meta-analyses (PRISMA) [5].

Study identification and data extraction

We used PubMed, EMBASE, and The Cochrane Library database to search relevant studies over the last 10 years by using the following search terms: ([including a keyword search using the words ‘anesthesia’ or a MeSH search using exp ‘General anesthetics’] and [including keyword search using ‘pediatric’ or ‘child*’ or ‘newborns ‘or ‘neonate’ or ‘young’] and [including a MeSH search ‘cohort study’ or a keyword search using ‘cohort analysis’]). Only those studies published in the English language were included; we did not define the minimum number of cases in studies to be included for meta-analysis. The study ladder chart is shown in Fig. 1. By this search strategy, 6052 papers were identified, including 2854 papers in PubMed, 737 papers in Embase, and 2461 papers in the Cochrane Library database.

Fig. 1
figure 1

The ladder chart of the literature search and selection

Full size image

The following inclusive selection criteria were applied: (1) compare the effect of general anesthesia on later neurodevelopment. The evaluation of neurodevelopment includes language and learning disabilities, cognition, behavioral development, academic performance, and autistic disorder as well as attention deficit; (2) independent retrospective or prospective study (3) with available data to estimate hazard ratio (HR) with 95 % confidence interval (CI). Whenever available, we extracted the adjusted HR for other confounding factors, since the adjusted estimates might reflect the true odds of effect, and we select the studies that should have a single anesthesia outcome before 4 years old.

After reviewing all papers, 13 eligible studies were selected for the present meta-analysis [6–18] and one of them originates from conference data [18]. The following variables were extracted from each study if available: first author’s surname, publication year, total number of cases, number of cases in exposure, adjusted HR with 95 % CI of outcomes, timing at exposure, evaluation time, evaluation standard, and evaluation items for neurodevelopment. The information was collected independently by two investigators (HFZ and LLD), and any discrepancy was resolved by discussion. The study quality was assessed using the 9-star Newcastle-Ottawa Scale (NOS). (The NOS for assessing the quality of cohort studies in meta-analyses. http://www.ohri.ca/programs/clinical_epidemiology/oxford.htm).

Meta-analysis and statistical methods

The meta-analysis was planned, conducted, and reported in adherence to the standards of quality for reporting meta-analysis [19]. For each study, an adjusted HR with 95 % CI was retrieved to estimate the association between anesthesia and neurodevelopmental outcomes. The heterogeneity among studies was assessed by Cochran Chi-square Q statistics and I 2 statistics, which determined the appropriate use of either fixed-effects (Mantel–Haenszel method) or random-effects (DerSimonian and Laird method) models. Heterogeneity was considered as either a p value <0.05 or I 2 > 25 % [20, 21]. Whenever significant heterogeneity was present, we searched for potential sources of heterogeneity. Influence analysis (sensitivity analysis) was conducted by omitting each study to find potential outliers. The potential publication bias was assessed graphically in the Begg funnel plot of log HRs against their standard error (SE), and the degree of asymmetry was tested using Egger’s test and Begg’s test.

We also computed the trend from the correlated log HRs estimates across categories of exposure levels by meta-regression as previously described [22, 23]. We made calculations assuming a log-linear relationship between HR and early time exposure before 3 and 4 years old. Results were considered as statistically significant for p < 0.05. All of the statistical analyses were performed using Stata/SE version 12.0 (Stata Corporation, College Station, TX, USA).

Results

Study characteristics

A detailed outcome of the search and selection result is shown in Table 1. A total of 85,014 children were included and 12,220 of them had experienced general anesthesia due to surgery procedures. The minimum age at exposure was at birth to 108 days and the maximum age was at the age of 48 months, with a mean age of 30.5 months, and 84.6 % papers study the relationship between the single general anesthesia and neurodevelopmental outcomes before age of 3 among 13 studies. Year of birth ranged from 1976 to 2010. Children after exposure were followed up from 1 year to 16 years. Of the 13 included studies, six were conducted in the USA [9, 12–14, 16, 18], two in the Netherlands [15, 17], two in Australia [7, 10], one in China [6], one in Denmark [11], and one in Singapore [8]. The number of ICD-9 evaluation standard accounts for 46 % in all 13 studies.

Table 1 Outcome of selected studies included in the meta-analysis
Full size table

Of 13 study findings, only one (Bartels et al. [17]) used the unadjusted HR of exposure to anesthetic agents with neurodevelopmental outcome for the original paper did not report the HR and 95 % CI. However, DiMaggio et al. [24] had calculated them in their previous meta-analysis. The findings from the other 12 papers had adjusted HR for association between exposure to single anesthesia and neurodevelopmental outcomes. Two of the 13 studies reported in the subgroup analyses were in accordance with the exposed age [6, 15]. Eleven of 13 studies were utilized to analyze the association between exposure to anesthesia and neurodevelopmental outcomes before age 3. The quality of studies evaluated by the NOS is shown in Table 1.

Effect measure synthesis: primary outcome

The Q statistic for the 13 estimates for single exposure was 21.09 (p = 0.049) with heterogeneity among the studies (I 2 = 43.1 %). Therefore, the random-effects model was used to analyze the data and an association between anesthesia with neurodevelopmental impairment was observed with a pooled HR of 1.27 (95 % CI 1.10–1.45, p = 0.001). Moreover, either graphical inspection for funnel plots or quantitative evaluation from Egger’s test and Begg’s test indicated a presence of publication bias (p = 0.014 and p = 0.012, respectively).

Sensitivity analysis

Influence analysis was conducted to ascertain the reason of publication bias. The studies conducted by Bong et al. [8] and DiMaggio et al. [16] showed results that were completely out of range of the others and probably contributed to the heterogeneity. When deleted the Bong et al. [8], the Q statistic was 16.19 (p = 0.134) with an I 2 of 32.1 % and quantitative evaluation from Egger’s test and Begg’s test indicated the presence of publication bias (p = 0.051 and p = 0.047, respectively). Similarly, the Q statistic was 16.02 (p = 0.140) with an I 2 of 31.3 % and publication bias is exist from Egger’s test and Begg’s test (p = 0.032 and p = 0.034, respectively) after omitting DiMaggio et al. [16]. After excluding both of these, no evidence of heterogeneity and publication bias was observed among the remaining studies (I 2 = 9.2 %), and the relationship between anesthesia with neurodevelopmental impairment was observed with a pooled HR of 1.17 (95 % CI 1.07–1.28, p = 0.001; Fig. 2). Either graphical inspection for funnel plots or quantitative evaluation from Egger’s test and Begg’s test indicated the absence of publication bias (p = 0.121 and p = 0.161, respectively). No individual study in the 11 studies affected the overall HR dominantly, because omission of any single study made no substantial difference.

Fig. 2
figure 2

Forest plot showing the association of single general anesthesia with the risk of neurodevelopment before age 4 in 11 studies after omitting Bong et al. [8] and DiMaggio et al. [16]

Full size image

Effect measure synthesis: second outcome

On the basis of the above study, we explored further the association between childhood exposure to single general anesthesia and neurodevelopment before 3 years old. Considering heterogeneity, similarly, we used the fixed-effects model to analyze nine studies. Using the homologous method, we could obtain the results exposed to single general anesthesia before age of 3. The Q statistic for the nine estimates for single exposure was 5.92 (p = 0.656) with an I 2 of 0 %, and the relationship between anesthesia with neurodevelopmental impairment was observed with a pooled HR of 1.18 (95 % CI 1.07–1.30, p = 0.001; Fig. 3).

Fig. 3
figure 3

Forest plot showing the association of single general anesthesia with the risk of neurodevelopment before age 3 in 9 studies

Full size image

The relationships between primary and second outcome

Meanwhile, we have compared the difference between before age of 4 and before age of 3 exposed to single general anesthesia in the neurodevelopmental impairment. The outcome shows that there is no significance difference, with the p value of 0.9 (Fig. 4).

Fig. 4
figure 4

The relationships between within age 4 and age 3 exposed to single general anesthesia

Full size image

Meta-regression analysis

We have already demonstrated an association between anesthesia and neurodevelopmental impairment, and a meta-regression was further performed to explore whether the risk of neurodevelopmental impairment could be predicted according to one parameter, which is the time at exposure. The adjusted HR results of individual eligible studies listed in Table 1 are plotted in Fig. 5a, b, including 11 studies in age of four and nine studies except Bong et al. [8] and DiMaggio et al. [16]. For the time at exposure, the slope of the black bold line (by log converted HR) in Fig. 5a, b represented the risk of neurodevelopmental impairment due to per 1-year early exposure to anesthesia. Regression analysis showed that the risk of neurodevelopment disorders for anesthesia decreases as the age of the child increases. The HR of effect of early exposure per year is 0.85, with 95 % CI of 0.49–1.46 and a non-significant p value of 0.518, indicating time at exposure (before age of 4 years in 13 studies) may have no effect on neurodevelopmental impairment. However, when heterogeneity and publication bias is considered in the above studies, the HR of effect of early exposure per year in age of 4 is 1.08, with 95 % CI of 0.89–1.33 and a non-significant p value of 0.374 and the HR of effect of early exposure per year in age of 3 in nine studies is 1.09, with 95 % CI of 0.94–1.27 and a non-significant p value of 0.205; indicating time at exposure (before age of 4 or 3 years) might have a limited effect on neurodevelopmental impairment.

Fig. 5
figure 5

a Meta-regression analysis: effect of time at exposure of single general anesthesia on neurodevelopment in children in 11 studies before age of 4. The black line indicates the effect of time at exposure on neurodevelopmental deficits. b Meta-regression analysis: effect of time at exposure of single general anesthesia on neurodevelopment in children in nine studies before age of 3 after influence analysis

Full size image

Therefore, surgeons should take note that children with early single exposure may have a higher risk of neurodevelopmental impairment compared to those with later exposure and it is better that exposures to single anesthesia before the age of 3 years be avoided in selective operation.

Discussion

It is estimated that more than 6 million children (including 1.5 million infants) will receive general anesthesia for both surgical and nonsurgical procedures in the United States in 2012 [25]. Numerous preclinical and retrospective human studies have reported that the provision of anesthetic to infants and children may be associated with adverse neurodevelopmental outcomes. However, few meta-analyses have been done to enlighten both researchers and clinicians regarding any iatrogenic effects of anesthesia in a deeper lever. DiMaggio et al. [24] have performed a Bayesian meta-analysis to investigate the association between pediatric anesthesia and neurodevelopmental impairment. The synthesized HRs based on seven unadjusted and six adjusted measures for the association of anesthesia with an adverse developmental outcome were 1.9 (95 % CI 1.2–3.0) and 1.4 (95 % CI 0.9–2.2). However, they have a concentration on any exposure on anesthesia and failed to include all eligible studies available at the age of analysis. The results of the present meta-analysis suggest a potential influence of anesthesia on the later neurodevelopmental deficits, especially those before 3 years old exposed to single anesthesia. Considering the existence of heterogeneity, the results of exposed to single general anesthesia before age of 3 and age of 4 have no evident difference after deleting Bong et al. [8] and DiMaggio et al. [16]. Similarly, the results of meta-regression analysis indicate that there is no evident different trend from the viewpoint in the time at exposure. Moreover, it seems that the range between age of 4 and age of 3 is no significant risk factor for neurodevelopmental impairment, that is to say, postponing surgical procedures that require first anesthesia if possible have better performed in children after 3.

Our findings from pooled meta-analysis are consistent with most, although not all, findings only from currently available literature—cohort study. Randomized trials provide strong clinical evidence and attractive purity of its design. However, they are not without limitations and challenges, and the RCT has drawbacks particularly in pediatric anesthesia that include the ethics of randomization, choice of comparison group, choice of relevant dose in the treatment arm, and definition of suitable outcome measure as well as protocol adherence [26, 27]. Indeed, due to numerous biases in these observational cohort studies—bias related to selection of the population and the control subjects and their comparability as well as outcome measure assessment, it has become extremely important to evaluate studies [28]. We, here, select the 9-star NOS to assess all 13 studies and use adjusted HR to do a meta-analysis (except Bartels et al. [17]), which can lead to selective bias. Commonly, the method of HR extrapolation requires comment. If not reported, the HRs was calculated from the data available in the published article. However, we can not obtain the adjusted HR from the origin article. We get the value of HR in the meta-analysis from DiMaggio et al. [24].

Significant heterogeneity was detected among the studies included in the meta-analysis. Heterogeneity was still observed when we limited the scope of analysis to studies investigating Bong et al. [8]. When studies were limited to DiMaggio et al. [16], heterogeneity was still detected. Therefore, we omit both of them, and heterogeneity reduced in a significant range. For the first study from Bong et al., the heterogeneity is from the size of the sample. The sample size is relatively small, with a total of 579 samples. The existing bias in the DiMaggio et al. study comes from unmeasured confounders, measurement error, and misclassifications in their comparison groups, leading to an underestimate of the true association between anesthesia and hernia repair with outcome. Therefore, the primary outcome in this study was not standardized, and as such, it could easily be subjected to variation owing to differences in local practice patterns and misclassification from diagnostic coding.

The other limitations, as Sun et al. [25] described, should be declared. First of all, there should be other prognostic factors not controlled in the meta-analysis. Differences in surgical techniques, varying patient populations, changes in definition of recurrence, and difficulty with long-term follow-up, all hamper firm conclusions. Second, pediatric anesthetic neurotoxicity is a complicated and complex issue. There are many confounding factors at play in addition to the potentially toxic effects of anesthesia, including maternal health, drug exposures during pregnancy and delivery, preexisting medical conditions in the child, environmental characteristics, and as Servick noted [29], a lack of funding as well as ethical issues. Third, the evaluation of neurodevelopment should use standard measurement forms, and different forms would result in different outcomes of neurodevelopmental assessment.

Publication bias is a well-known problem in meta-analysis. The meta-analysis was restricted to papers published in English, and language bias might exist. However, our analysis did not suggest publication bias, so the summary statistics obtained may only approximate the actual average. After all, no statistical approach is a panacea for potentially biased or confounded data. Selective bias exists in meta-analysis for we just select all the studies in last 10 years considering the updated anesthetic agents.

Despite the limitations listed above, our study did find that single general anesthesia was possibly associated with neurodevelopment, especially exposed anesthesia before the age of 3. The outcome of the meta-analysis hints that we had better shift the focus to come the child who need operation under single general anesthesia within 3 years of age, rather than within 4 years old. To become a useful prognostic marker for individual patients and in the context of targeted therapy, these results clearly need to be confirmed by well-designed clinical experiments, for instance prospective studies, with larger, more clearly defined patient populations.