Lessons learned from big data (APRICOT, NECTARINE, PeDI)





Abstract


Big data in paediatric anaesthesia allows the evaluation of morbidity and mortality of anaesthesia in a large population, but also the identification of rare critical events and of their causes. This is a major step to focus education and design clinical guidelines. Moreover, they can help trying to determine normative data in a population with a wide range of ages and body weights. The example of blood pressure under anaesthesia will be detailed. Big data studies should encourage every department of anaesthesia to collect its own data and to benchmark its performance by comparison with published data. The data collection processes are also an opportunity to build collaborative research networks and help researchers to complete multicentric studies. Up to recently, big data studies were only performed in well developed countries. Fortunately, big data collections have started in some low and middle income countries and truly international studies are ongoing.



Introduction


Most of the paediatric anaesthesia literature from the last decade has been concentrated on the potential neurotoxicity of anaesthetic drugs on the developing brain. Based on evidence from pre-clinical studies, the anaesthesia community was really concerned about the risk of delayed or impaired long term neurodevelopment when a neonate, infant or young child is exposed to general anaesthesia. Several retrospective and prospective clinical studies were performed and some others are ongoing. Contrary to studies performed in animals, most clinical studies failed to confirm that, in healthy children, a single short exposure to anaesthesia is responsible by itself of any neurological damage resulting in delayed achievement of neurocognitive functions, nor were they able to identify a specific phenotype to look for, except perhaps for some impact on visual memory as shown in monkeys more than 2 years after exposure to anaesthesia [ ], and more recently in children.


At the same time, because the anaesthesiologist is also responsible for protection against pain, surgical stress, inflammation etc, efforts of the scientific community highlighted the need for good quality anaesthesia to guarantee a favourable outcome beyond the potential risk of neurotoxicity intrinsic of the anaesthetic medications. This is the goal of the SAFETOTS international initiative ( https://www.safetots.org ): it defines the competences (5 W: who, where, what, when and how) necessary to provide safe anaesthesia in children and insists on the need to maintain physiological and psychological homeostasis in the perioperative period (summarized as the 10 Ns). Moreover, anaesthesia is never administered for trivial reasons, but for diagnostic or surgical procedures. In addition to emergencies, some of these procedures need to be performed electively at a young age to avoid future damage: for example, cleft lip and palate repair for language learning, cardiac surgery to prevent failure to thrive or pulmonary hypertension etc. For less critical procedures, it is often recommended to postpone anaesthesia until the patient’s age is out of a window of vulnerability based on evidence coming from animal studies. However, the human neurodevelopment extends over a much longer period of time (up to adolescence) than in animal models, and the timing and duration of these windows of vulnerability vary according to which function is considered. Moreover, neuroplasticity can compensate for some mild functional disturbances. On the other hand, even if rare, severe critical events occurring during anaesthesia may have long lasting effects. For example, a drop in oxygenation, blood pressure, glycaemia, etc, can result in a poor neurological outcome especially if the derangement is prolonged in time. In addition, the preoperative patient conditions (i.e. co-morbidities and/or pre-existing medical conditions) may predispose patients to unfavourable outcome, if these conditions are not identified and adequately addressed. The present chapter is dedicated to the results of studies on anaesthesia-related morbidity, mortality and severe critical events in children.



Big data to study anaesthesia-related morbidity, mortality and critical events: APRICOT and NECTARINE


Large data collections represent an unique opportunity to evaluate anaesthesia-related morbidity and mortality, and identify the incidence of rare critical events as well as their relative risk factors. Due to the rarity of these events in paediatric anaesthesia and the variability in practice that might influence their occurrence, the only possibility to study critical events is collecting data on a large multi-institutional level. One can use either an ongoing registry format (e.g., WakeUpSafe, Paediatric Regional Anaesthesia Network or PRAN, Paediatric Difficult Intubation registry or PeDI) [ ] or a time-limited data collection format (e.g., APRICOT, NECTARINE) [ , ]. Interestingly, the former is more frequently used in North America and the latter in Europe. This is probably due to organizational and funding issues.


The Clinical Trial Networks (CTNs) have been launched by the European Society of Anaesthesiology and Intensive Care (ESAIC) many years ago with the aim to connect hospitals from different countries, collaborate and collect data on rare events. Many CTNs have been successfully concluded under the umbrella of the ESAIC. APRICOT and NECTARINE are two of these. The first was a prospective cohort study exploring incidence and risk factors of predefined severe critical events in children undergoing anaesthesia. Similarly, NECTARINE explored predefined critical events, morbidity and mortality in neonates and infants undergoing anaesthesia. Both CTNs were successful and concluded with unprecedent number of patients entered into the database. The APRICOT database had more than 31,000 patients included, while NECTARINE more than 5500. The APRICOT main publication went out in the Lancet Respiratory Medicine Journal [ ], while the main outcome of NECTARINE was divided into two main publications: the first reporting incidence of critical events [ ] and the second addressing the incidence of difficult airway management [ ].


One of the main challenges in designing APRICOT was the definition of these severe critical events since variability in defining them properly and homogeneously influences the outcomes and jeopardizes the comparison between various outcome trials performed in various settings and geographical locations [ ]. Hence, one of the first lessons learned from APRICOT was the use of the same nomenclature, which has a great impact on the establishment of good clinical practice guidelines and can be used as a basis for completing an incident-reporting system for departmental quality improvement initiatives, competency-based curriculum during training as well as for continuous professional development [ ].


The higher than expected incidence of severe critical events demonstrated in APRICOT pointed out the lack of standardization of care and variability of practice among European countries. Analysis of the identified risk factors allowed to develop two major frameworks for progress in paediatric anaesthesia safety. Firstly, and for the first time, a threshold of 3 years of age was established as the cut-off for an increased incidence in the severe perioperative critical events, mainly respiratory in toddlers and haemodynamic in infants. Secondly, the identification of a population at higher risk for severe critical events, which is represented by children with history of prematurity, handicap (metabolic or genetic disorder, or neurological impairment), snoring, airway hypersensitivity, and those having a medical condition with fever or under medication. These findings constituted prerequisites for the elaboration of guidelines in several European countries recommending that children meeting these criteria should be anaesthetized by an experienced anaesthesiologist with sufficient paediatric training and ongoing paediatric practice.


The impact of the experience of the anaesthesia team on the quality of anaesthesia management in children has been questioned in the literature [ ]. APRICOT involved large and complex datasets, and one of its advantages was the ability to perform sophisticated analyses, such as multivariate analysis, which considers multiple variables simultaneously. Correcting for variables that may affect the outcome is an essential aspect of such analysis. Applying multivariate analysis provided evidence for the role of the anaesthesiologist’s experience with a 1% and 2% decrease in the incidence of severe respiratory and cardiac adverse events, respectively, for each additional year of experience. However, one must not be mistaken, and one should be alerted about the hazards for experience only-based knowledge since some physicians could counter rigorous research evidence to provide weak clinical evidence. While APRICOT underscores the considerable diversity in practices, the study’s robustness stems from the consistent observation of numerous occurrences. This approach enables the detection of infrequent events and the identification of corresponding risk factors.


Moreover, APRICOT provided insight into hidden factors, which served for building hypothesis and emphasising education in areas where the identified severe critical events could be predicted and prevented. For example, APRICOT provided further evidence for the role of the presence of airway sensitivity (wheezing in the past 12 months, recent upper respiratory tract infection, asthma diagnosis and passive smoking) in increasing the risk for respiratory critical events [ ]. It has been well demonstrated that children with airway sensitivity benefit from premedication with a salbutamol aerosol for the prevention of perioperative respiratory adverse events, a finding that should be part of all anaesthesia management plans particularly in pre-school children and in those undergoing ENT surgeries . In addition, one of the main risk factors identified for both cardiac and respiratory critical events was prematurity. The higher incidence for critical events requiring unplanned intervention in premature neonates was also confirmed in the NECTARINE study [ ], which provides further evidence for referring this population to specialized centres where perioperative management could be guaranteed by a multidisciplinary approach and specialized paediatric anaesthesiologists [ ].


Big data may also highlight hidden or so far unknown risk factors. APRICOT revealed that sex significantly impacted the incidence of severe cardiac critical events with a 25% higher relative risk in girls than in boys (2.3% vs 1.7%, respectively with RR 0.74; 95% CI [0.63–0.87]). This association is surprising since various factors may influence the incidence of complications including type of surgery, choice of anaesthesia management and pre-existing medical condition. To our knowledge, such association has never been pointed out in the literature before and a difference between sexes in the incidence of complications has only been reported recently for respiratory adverse events in children undergoing caudal anaesthesia under sedation without airway instrumentation [ ]. Nevertheless, such finding deserves further consideration: knowing that detecting an association does not imply causality but to gain a comprehensive understanding of this potential risk, it is essential for upcoming research to focus on in-depth exploration and analysis of the potential effect of sex on the incidence of severe critical events.


Another point revealed by the APRICOT study was the relatively low incidence of 30-day perioperative mortality, which was not related to anaesthesia: it was 10 in 10,000 (0.1%, 95% CI [0.07–0.14]), which is comparable to the incidence reported by many studies in developed countries [ ]. This finding is in contrast to that observed in the NECTARINE study [ ] where the incidence of mortality at 30 and 90 days was much higher (3.2%; (95% CI [2.7–3.7] even when correcting for cardiac surgery. This again confirms the specificity of the neonatal population, which requires a multidisciplinary approach in specialized centres with high expertise.



Big data as a base for clinical guidelines


While APRICOT and NECTARINE were focused primarily on morbidity and mortality, both studies had also the merit of collecting information about airway management in children. These available large datasets allowed an insight into identifying the incidence for difficult tracheal intubation that was defined in both trials as needing 3 or more attempts to secure the airway and/or Cormack Lehane grading of 3 or 4. APRICOT revealed that the incidence for difficult tracheal intubation was much higher in neonates and infants (1%; 95% CI: [0–2.2] and 1.1%; 95% CI [0.6–1.6], respectively) [ ]. Moreover, there was evidence for an increase in severe respiratory critical events when facing difficult intubation and when not using a neuromuscular blocking agent for tracheal intubation. However, this rate in neonates and infants was probably underestimated since this result was obtained from a subpopulation that represented 1.2% and 9.4% of the entire cohort, respectively. Focusing more precisely on the neonatal population, the incidence for difficult tracheal intubation in NECTARINE was much higher and of concern with a rate of 6.9%; 95% CI: [6.1-7-7], as it included also patients who had their trachea intubated in the paediatric or neonatal intensive care units [ ]. More striking was the incidence of complications that was associated with difficult tracheal intubation in neonates with a 40% incidence of hypoxaemia and 7.7% of bradycardia. The large dataset from the « Paediatric Difficult Intubation Registry » (PeDI) [ ] revealed that hypoxaemia concomitant to difficult tracheal intubation led to cardiac arrest in almost 15% of the cases [ ]. Building upon these significant findings and the compelling evidence regarding the efficacy of videolaryngoscopy-guided intubation [ ] and nasal oxygenation during neonatal intubation [ ], a panel of experts has collaboratively developed and recently published guidelines for intubation in neonates and infants [ ]. These guidelines aim at proposing an universal strategy applicable to airway management education and the development of a comprehensive curriculum. They meticulously address three pivotal stages: i) preparation (including patient assessment, evaluation of the local situation (staff and assistance), and discussion of the management plan; ii) intubation attempts, which are categorized based on whether the patient is clinically stable or unstable and emphasize the importance of anaesthesia depth and systematic nasal oxygenation; iii) formulation of an extubation plan and/or the emergency pathway depending on whether the patient can be awakened or not. The primary objective of these guidelines is to establish standardized practice, strongly advocating for the routine use of videolaryngoscopy, nasal oxygenation, and careful consideration of neuromuscular blocking agents (and reversal). This is a perfect example on how the insights derived from big data analyses led to inform and contribute to the establishment of evidence-based guidelines.



Using big data to determine normative data: the example of blood pressure


The collection of big data can also help determining normative data during anaesthesia. This was recently the case for blood pressure during anaesthesia. Blood pressure (BP) is routinely used as a surrogate of cardiac output to make sure at short time intervals (a few minutes or beat-to-beat depending on whether non-invasive or invasive monitoring is used, respectively) that organ blood flow is adequate [ ]. There are of course other means to achieve this goal but either their response time is too slow (blood lactates, urine output) or they are either unavailable or not used in routine cases (e.g., monitoring of organ oxygenation (NIRS)). An impetus for further research to determine what is a normal or acceptable BP under anaesthesia followed the description of 6 dramatic cases of postoperative ischaemic encephalopathy in infants by McCann et al. [ ]. Their data showed that systemic hypotension had probably played a role along with hyperventilation as evaluated with capnography. But they also showed that the mean arterial pressure ( MAP ) value that could have been used as a threshold to diagnose and treat hypotension varied according to which population data were used, i.e. the Paediatric Advanced Life Support guidelines or the survey of Society for Paediatric Anaesthesia and Association of Paediatric Anaesthetist of Great Britain and Ireland [ ]. In addition, many readers were not sure they would have considered some numbers as hypotension. Last, normotension is one of the 10 physiological and psychological parameters to be controlled to maintain homeostasis during the perioperative period according to the SAFETOTS initiative [ ].


Let us review recently published big data to try to answer to the following clinical questions:



  • 1

    What is the normal blood pressure in children under anaesthesia ?


  • 2

    What is the threshold to diagnose hypotension during anaesthesia ?


  • 3

    What is the reliability of noninvasive blood pressure measurements?




What is the normal blood pressure in children under anaesthesia ?


The first study to present reference ranges for BP in children under anaesthesia was a retrospective observational cohort study of data from the Multicenter Perioperative Outcomes Group data set [ ]. Noninvasive blood pressure (NIBP) data recorded after the induction of anaesthesia but before incision and after a brief stable portion of surgery in 116,362 cases from 11 centres (USA and The Netherlands) were used. The authors created NIBP systolic, mean and diastolic age-, weight- and sex-specific reference curves and tables representing the 50th percentile, +1 SD, −1 SD, and the upper (+2 SD) and lower reference ranges (−2 SD). Moreover, individualized data can be obtained online: http://www.pediatric-anesthesia.eu/bp_calculator . These curves show that BP measured during general anaesthesia is lower than the reference values obtained in awake patients, which is expected, but also a great interindividual variation: for example, the P50 of the MAP in a 10 kg infant after induction of anaesthesia was 51 mmHg but varied from 75 (P95) to 34 (P5) mmHg.


This large variation of BP in anaesthetized children was confirmed in a monocentric retrospective study [ ] and in smaller prospective studies such as GAS [ ], a Danish [ ] and a Dutch series [ ]. The authors of another retrospective study were able to calculate the median, 5th and 95th centiles for BP under anaesthesia according to age and type of anaesthesia [ ]: in addition to a large variation, they found that total intravenous was associated with a higher BP than inhalational anaesthesia.


When interpreting those numbers and curves, we should remember that they represent what was recorded during the study period and considered acceptable by the anaesthetic team in charge. Moreover, except for the GAS study [ ], they were not related to any outcome: they simply represent the variation in the population studied and an unusually low or high BP value should alert the anaesthesiologist to interpret it in the medical and surgical context.



What is the threshold to diagnose hypotension under anaesthesia?


Given that BP decreases during anaesthesia, we need to determine a threshold value below which hypotension should be considered and treated. Knowing that BP decreases by 10–20% during sleep [ ], we can assume that a decrease of less than 20% from the awake value is safe. A survey in 2009 showed that the majority of responders considered a decrease in systolic blood pressure of 20–30% from baseline as significant [ ]. In the NECTARINE study, the mean [SD] decrease in BP reported by the participants before reacting was 29.3 % [19.5] from baseline [ ]. The clinical interest of this 20% decrease threshold was confirmed in infants using transcranial Doppler or near-infrared spectrometry (NIRS) to evaluate cerebral blood flow or oxygenation, respectively [ , ]. These data were however obtained in a small number of healthy infants and using NIBP and capnography to evaluate the haemodynamics and ventilation.


A critical issue is thus determining what is the baseline BP for an individual child. We all know that measuring BP in an awake child is difficult and often results in a distressing experience; moreover, stress and anxiety can result in an artificially high BP. That is why preoperative BP values are often not available: less than 30% of the children in the GAS study [ ] and less than 2% in a retrospective study [ ] had their BP measured before anaesthesia. In a prospective study in infants, in which awake baseline BP could be measured in 85% of the cases [ ], Weber et al. showed that the incidence and thus the diagnosis (and potential useless treatment) of hypotension varied depending on whether a preoperative value or absolute numbers taken out of the literature were taken as the baseline.


The situation is even more critical in premature infants. The following rule of thumb, based on an experts’ consensus statement [ ], was used for years: MAP under anaesthesia should be greater than the infant’s gestational age in weeks during the first two days of life. This number equals approximately the 10th percentile for normal blood pressure. Moreover, many teams use this rule during a longer period of time than the first 48h of life. Recently, percentile curves based on 5885 NIBP measurements performed during the first week of life in 607 non critical neonates born between 24 and 40 weeks gestational age have been published [ ]. They can be used to determine the normal range of awake BP in this fragile population. Individual data can even be calculated using a calculator: http://bloodpressure_neonate.com . Whether the lowest value should be used as a threshold for the diagnosis of hypotension remains to be determined.


In clinical practice, even if preoperative BP data are available, an easy solution could be starting NIBP as soon as the child loses consciousness during induction of anaesthesia. This allows obtaining data close to the BP of the awake relaxed child, at a time where stress is controlled and the haemodynamic effects of the anaesthetic agents are minimal. During anaesthesia, one should preserve access to a peripheral (usually radial) pulse, the strength of which has been evaluated during induction. In case of low BP value, the NIBP measure is repeated twice (based on the assumption that the two most close values are probably reliable) while the pulse strength is checked: if it is easily felt, no problem. If it is weak or difficult to feel, and/or if the associated NIRS monitoring, if available, shows a decrease by more than 15%, the presence of hypotension is confirmed and one should look for possible reasons: hypovolaemia, overdose, surgical maneuver, hyperventilation … and take action accordingly. One should be more aggressive in preventing and treating hypotension and not wait for the appearance of a shock status with loss of the pulse, which may lead to cardio-circulatory arrest.



What is the reliability of NIBP measurements ?


Noninvasive blood pressure is generally measured using an oscillometric device that evaluates the amplitude of the pulsations of an artery (usually the humeral) following deflation of a cuff placed around the arm. The maximal amplitude is determined to be the MAP, and a proprietary algorithm calculates the systolic and diastolic values. It is recommended to use a BP cuff with a cuff width to arm circumference ratio the closest to 0.5 positioned on the right upper arm [ ]. A systematic review of the studies comparing MAP measured invasively with MAP by NIBP using the Bland-Altman methods for comparison showed that 1) the mean BP difference was outside the ±5 mmHg [SD 5 mmHg] acceptable bias in the majority of studies, and 2) that the NIBP value was usually higher, especially in neonates [ ]. This was confirmed in other studies [ , ]. In a series of 1193 paired pressure measurements in anaesthetized neonates, the percentage error between MAP measured noninvasively and invasively was 26.4%, and NIBP was higher than intra-arterial BP in case of hypotension and lower in case of hypertension. Although invasive blood pressure may overestimate the systolic pressure due to amplification during propagation of the pressure wave, this uncertainty about accuracy should be kept in mind when interpreting NIBP measurements. The accuracy of NIBP needs to be improved, especially in neonates and infants. Moreover, other techniques, such as continuous finger cuff blood pressure, should be developed that facilitate BP measurement both in awake and anaesthetized children.




Practice Points regarding blood pressure in children .




  • given the large interindividual variation of BP in anaesthetized children, there are no magic numbers: the measured values should be interpreted individually in the medical (e.g., cardiac or intracranial pathology), anaesthetic and surgical context



  • when using NIBP, the measured MAP is more reliable than the calculated systolic pressure



  • a reliable baseline value for baseline blood pressure can be measured during induction, as soon as the child loses consciousness



  • in case of low blood pressure, the strength and presence of a peripheral pulse should be verified while controlling the measure and looking at the surgical field



  • NIBP tends to overestimate BP in case of hypotension and underestimate it in case of hypertension


Only gold members can continue reading. Log In or Register to continue

Stay updated, free articles. Join our Telegram channel

Mar 30, 2025 | Posted by in ANESTHESIA | Comments Off on Lessons learned from big data (APRICOT, NECTARINE, PeDI)

Full access? Get Clinical Tree

Get Clinical Tree app for offline access