Adverse heart rate responses during beach-chair position for shoulder surgeries - A systematic review and meta-analysis of their incidence, interpretations and associations.

BACKGROUND AND AIMS: Evaluations of adverse heart rate (HR)-responses and HR-variations during anaesthesia in beach-chair-position (BCP) for shoulder surgeries have not been done earlier. We analysed the incidence, associations, and interpretations of adverse HR-responses in this clinical setting.
METHODS: We performed a meta-analysis of trials that reported HR-related data in anaesthetised subjects undergoing elective shoulder surgeries in BCP. Studies included prospective, randomised, quasi-randomised and non-randomised, controlled clinical trials as well as observational cohorts. Literature search was conducted in MEDLINE, EMBASE, CINHAL and the Cochrane Central Register of Controlled Trials of the 21st century. In the first analysis, we studied the incidence and associations of bradycardia/hypotension-bradycardia episodes (HBE) with respect to the type of anaesthesia and different pharmacological agents. In the second, we evaluated anaesthetic influences, associations and inter-relationships between monitored parameters with respect to HR-behaviours.
RESULTS: Among the trials designed with bradycardia/HBE as a primary end point, the observed incidence of bradycardia was 9.1% and that of HBE, 14.9% and 22.7% [(for Interscalene block (ISB) ± sedation) subjects and general anaesthesia (GA) + ISB, respectively]. There was evidence of higher observed risk of developing adverse HR-responses for GA subjects over ISB (Risk Difference, P < 0.05). Concomitant use of β-agonists did not increase risk of HBEs (P = 0.29, I 2= 11.4%) or with fentanyl (P = 0.45, I 2= 0%) for ISB subjects (subgroup analysis). Fentanyl significantly influenced the HR-drop over time [meta-regression, estimates (standard error), 14.9 (5.4), 9.8 (4.3) and 17 (2.6); P = 0.007, 0.024 and <0.001; for early, mid and delayed periods, respectively] in GA subjects. With respect to number of subjects experiencing cerebral desaturation events (CDEs), total intravenous anaesthesia (TIVA)- propofol had higher risk over inhalational anaesthesia (P = 0.006, I 2 = 86.7%). Meta-correlation analysis showed relationships between the HR and rSO2(regional cerebral oxygen saturation) or SjvO2(jugular venous oxygen saturation) values (r = 0.608, 95%CI, 0.439 to 0.735, P < 0.001, I 2= 77.4% and r = 0.397, 95%CI, 0.151 to 0.597, P < 0.001, I 2= 64.3%, respectively).
CONCLUSIONS: There is not enough evidence to claim the associations of adverse HR-responses with any specific factor. HR-fall is maximal with fentanyl and its variability is associated with changes in rSO2. Fall in rSO2 could be the common link triggering adverse HR-responses in BCP. Copyright:

INTRODUCTION

Among the undesirable haemodynamic consequences of beach-chair position (BCP) for shoulder (arthroscopic) surgeries, bradycardia, by virtue of its unpredictable occurrence and occasionally adverse anaesthetic consequences, is a cause for concern.[1234] A specifically named haemodynamic event, the 'Hypotension-Bradycardia Episode' (HBE) has been reported in 6-27% of BCP subjects.[345] These studies however lack specificity in documenting isolated significant bradycardia (necessitating the use of atropine). The true incidence of bradycardia remains indeterminate due to several factors such as the frequent use of the terms 'bradycardia' and 'HBEs' as synonyms,[46] use of different definitions of 'bradycardia' by various authors, inclusion of additional causes of 'hypotension' episodes (anaesthetic and pharmacological) and subjective variations in the anaesthesiologist's decision to use atropine, justifiably attributable to a 'play it safe' attitude.

The correlation of incidence of bradycardia/HBE with the type of anaesthesia[5] or the anaesthetic agent deployed has not been conclusively established.[7] While activation of the Bezold-Jarish Reflex (BJR) linked to interscalene block of the brachial plexus (ISB) could be the primary reason for such adverse events,[8] the demonstration of a 'non-empty' heart ventricle during such events suggests otherwise.[91011] Similarly, the association of use of β-adrenergic agonists and adverse heart rate (HR)-responses/HBE[24] is uncertain since these episodes were also reported in patients without their use.[1213] Likewise, while ISB has been linked to such events,[8] the same has not been confirmed with general anaesthesia (GA). There is a paucity of comparative literature on the association of HR-responses in BCP with other parameters like use of maintenance anaesthetic agents or opioids. Several studies indicate a strong association of hypotensive response with regional cerebral oxygen saturation (rSO2)[141516] and jugular venous oxygen saturation (SjvO2) for BCP surgeries done under anaesthesia.[16] But it is unclear whether cerebral desaturation events (CDEs) correlate with (adverse) HR-responses.

The aim of this study was to systematically review all available evidence from trials reporting bradycardia/HBEs for its: 1) incidence, 2) anaesthetic/pharmacological associations, and 3) association of BCP-HR-behaviours with monitored parameters, and to conduct a meta-analysis on the results. Establishing the association of adverse haemodynamic responses with specific anaesthesia-related variables or changes in monitored parameters would be helpful in improving predictability of such events, taking precautionary measures to prevent them and providing an insight into their possible underlying pathophysiological mechanisms.

METHODS

Registration and protocol

This meta-analysis was conducted in accordance with Preferred Reporting Items for Systematic reviews and Meta-analyses.[17] The protocol was registered with PROSPERO (CRD42019119454, crd.york.ac.uk; date of registration, 14/01/2019, and updated on 31/07/2019).

Eligibility criteria

We included prospective, randomised, quasi-randomised and non-randomised, controlled clinical trials as well as observational cohorts with adult subjects (>18 yrs) undergoing elective shoulder surgeries in BCP. Reporting of HR-related data or HR-responses were mandatory to inclusion. Publications in all languages were considered. Subjects received one of the following anaesthetic modalities; (1) Planned GA; (2) Regional anaesthesia (RA): ISB or similar and (3) RA in combination with GA. The use of supplementary sedation was not a barrier to inclusion. We excluded studies wherein subjects underwent surgeries in <45° BCP as well as American Society of Anaesthesiologists (ASA) >3 physical status.

Information sources

An electronic literature search was conducted in MEDLINE, EMBASE, Cochrane Central Register of Controlled Trials and CINHAL. The selection of literature specifically restricted to studies in BCP. We also searched the bibliography of retrieved manuscripts for additional studies pertaining to data encompassing our primary outcome of interest. These comprised studies reporting incidents of isolated bradycardia or HBEs, documenting maximum and minimum average HRs, or measuring serial HR over time periods; with a caveat that both pre-induction and post-induction HR data be available. Twenty- first-century literature were scanned since anaesthesia protocols have remained uniform during this period. Retrospective studies, reviews with inadequate information on primary outcome interests, abstracts and letters to the editor were not included. The detailed search strategy is shown in Supplementary Digital Content File 1.

Study selection and data collection

The manuscripts meeting the inclusion criteria were assessed and data were extracted following a standardised format. Extracted items comprised of study characteristics, risk of bias (RoB) domains,[18] participant disposition, and study outcomes. Patients were categorised according to type of the surgery or anaesthesia, number of subjects and position adopted for surgery (≥45° of BCP, i.e., 45 to 90°). Interventions referred to BCP after induction and achievement of hemodynamic stability. Comparison of variables was pre-BCP versus post-BCP. Outcomes were classified as 'primary' and 'secondary'. The former included HR data before and after BCP at various intervals of time, the incidence of bradycardia/HBE in BCP, influence of anaesthetics over HR-responses and HR-rSO2/SjvO2 associations. The latter included incidence and magnitude of hypotension and associations of mean blood pressure (MBP) with anaesthetic factors, vasoactive drugs and rSO2/SjvO2 in BCP.

Data synthesis and analysis of outcomes

For evaluation of the outcome of interests of this interventional (investigating an effect of BCP on HR) meta-analysis, data were extracted separately from study groups (SGs) of each trial to negate the effect of intergroup variables affecting their outcomes. We categorised the SGs further into study control groups, randomised SGs, non-randomised SGs, physiological control groups. Study control groups received standard anaesthesia care without additional investigating pharmacological agents or technical measures. Physiological control groups were those placed in BCP but not anaesthetised.

The HR data collected included values documented at a single point of time or continuous data at various intervals for a SG. Incidences of bradycardia and/or HBE and rest of the HR data were considered for meta-analysis. Data were collected as a single or combined value in the form of mean and standard deviations (SD) or median and inter-quartile range (IQR), respectively. If multiple data were provided, then they were converted into pooled statistical averages. The data were tabulated under pre-induction [baseline (BL)] and post-induction groups. The latter included data relating to pre-BCP and post-BCP categories after the stabilisation of vitals. These post-BCP HR data were pooled for the time periods mentioned in the respective publication. If recorded data timings were non-specific timings, they were approximated to a specific time by mutual discussion with the two authors. Publications with unreported or inconclusive data that could not be obtained after attempts to contact the authors were excluded from this review.

The data presented in tables, text or images were used as the primary source for extraction. A graph digitizing software (Enguage Digitizer version 10.10, @ Mark Mitchell) was used for efficiently extracting and estimation of numerical raw data whenever text numerical data were unavailable. We substituted the missing SDs with pooled SDs of other studies with the same comparison by ®[(∑ N*SD2)/∑ N] where N = sample size. When range and IQR were available, SD was estimated using the formula SD = range/4 and SD = IQR/1.35, respectively, as described by Cochrane Hand Book of Systematic Reviews.[19] Data were reported as 95% confidence intervals (CI). The median was used to estimate the mean if the value was not reported. Whenever standard error of mean (SEM) was reported, SD was obtained as SD = SEM®N. If data were provided as % of change over a BL numerical value, they were converted to numbers. To account for drop out cases over time or termination of BCP before the time specified in the meta-analysis, subject numbers were approximated to the nearest values for pooled data estimation. If the exact time point was not specified in the manuscript, then the approximated time point was considered by the authors' judgment.

We used individual definitions for defining events of bradycardia, HBE, hypotension and CDEs as described by authors of each study. Dichotomous data like bradycardia, hypotension, CDEs, etc. were converted into incidence (n/N) for a given time interval. The single highest incidence was used to capture the proportion of subjects who experienced a certain adverse response at least once. Data from SGs receiving more than one intervention or different anaesthetic agent or a technique (within a SG) were combined into a single group as per Cochrane Hand Book.[19] Data were clubbed together into a single group whenever the primary authors grouped the study subjects on the basis of an event. Finally, 'intention to treat' basis was used for analysing complications-related data in some SGs. Subjects were repositioned back to supine following BCP-induced haemodynamic disturbances.[20]

Data synthesis specific to HR

Incidence of bradycardia/HBE was considered whenever the events were reported either individually or synonymously in the subject at least once. To differentiate isolated bradycardia from the broader term, HBE, we considered the use of atropine (n/N) for defining the former. Data relating to HR-variability over time were again sub-divided into immediate/early (~10 minutes, EHR), mid (11-30 minutes, MHR) and delayed (after 30 minutes till the end of BCP, DHR). The magnitude of changes over time was represented by mean differences (MDs).

Data synthesis specific to blood pressures (BPs)

MBP was considered for data evaluation and the data synthesis was similar to that followed for HR. We excluded pooled data of systolic or diastolic blood pressures. Whenever SDs were not reported for nadir values, they were imputed from pooled SDs of the same group. All analysis was done presuming no incidence of hypotension in the supine position under anaesthesia. Subjects who were excluded prior to surgery, after BCP, owing to severe hypotension were also included (intention to treat).

Data synthesis specific to CDEs

For analysing CDEs, two types of rSO2 values (MDs) were considered; (1) MDs of pre and post-BCP (pooled), as 'absolute' values; (2) MDs of pre-BCP and 'lowest' achieved post-BCP rSO2 values. Lowermost of lowest was considered whenever right and left cerebral hemispheres were recorded separately (with single or two different methods). Whenever SDs were not reported for nadir rSO2 values, they were imputed from pooled SDs of the same group. All analysis was done presuming no incidences of CDEs in supine position under anaesthesia.

Pre-defined sources of heterogeneity

To explore the potential causes of heterogeneity in our results that could influence primary outcome results, we pre-identified certain clinical aspects of individual SGs. These included (1) randomisation technique; (2) anaesthetic technique; (3) induction agent; (4) maintenance anaesthetic agent; (5) use of opioids; (6) use of vasoactive agents. Equivalent doses of ephedrine and phenylephrine were considered for vasopressor consumption, converting ephedrine doses to their phenylephrine equivalence using a potency ratio of 81.2: 1.[21]

The degree to which some of these additional factors predict EHRs, MHRs and DHRs was evaluated using a meta-regression analysis. To examine the influence of different anaesthetic agents, opioids, vasoactive drugs or eligibility criteria on HR-variability, we performed a sensitivity analysis. Sub-group analysis was considered based on: (1) type of anaesthesia; (2) predisposing or preventing agent; or (3) the maintenance agent for both incidences of bradycardia/HBE and serial HR measurements. Additional analyses ('leave-one-out' analysis, correlation statistics and meta-correlation analysis) were considered as necessary (for primary outcomes).

Meta-analysis was conducted with Review Manager (RevMan) 5.3 (Cochrane Collaboration, Copenhagen, Denmark, 2014). The random effects model was used for all analyses. Heterogeneity was measured and expressed as I2.[22] Meta-regression was performed using JASP software (Version 0.9.2, BibTeX, Amsterdam).[23] This analysis excluded subjects administered with ISB ± sedation since the anaesthetic agent influences on HR are largely absent. Meta-regression (Restricted-Maximum-Likelihood method, random effects) was performed for EHR with priori defined factors, induction agents, opioids and use of PVIs. For MHR and DHR, maintenance anaesthetic agents and opioids were considered.

For continuous variables (HR, absolute and lowest achieved cerebral saturations), MDs were compared using the inverse-variance (I-V) method. For dichotomous variables (incidences of bradycardia, HBEs, CDEs, hypotension), odds ratio (OR), risk ratio (RR) or risk differences (RD) were computed by the Mantel-Haenszel (M-H) or I-V methods. Natural log-transformation was adopted[24] as the outcomes for incidences were expected to be non-normal. Publication bias was checked using regression test for funnel plot asymmetry and Egger's test (JASP software, version 0.9.2).[25] Correlations were attempted for those SGs which mentioned statistical averages of consecutive measurements of HR, rSO2 and SjvO2 on the one hand and for MBP and rSO2 on the other. Meta-correlation analysis was performed after obtaining a series of correlation coefficients for various SGs using MedCalc® Version 14.8.1, MedCalc Software bvba, 2014. For all, statistical significance was set at P < 0.05 (2-tailed).

RESULTS

Summary of results for various outcomes are provided in Table 1.

Table 1

Summary results

Parameter analysis	n	Outcome	Comments (GRADE recommendation)
Definition of bradycardia/HBE	1121	Definition of bradycardia/HBE varied much between authors; therefore, the diversified incidence reporting.	Majority of authors used definition by Liguori et al.a
Incidence of bradycardia and anaesthetic influences	712	9.1% of subjects are reported with bradycardia with RR of 9.8, after positioning to BCP.ISB had no excessive risk of developing bradycardia over GA.	Limited data available for GA subjects(⨁⨁⨁⚪-moderate, for overall and all subgroups)f
Incidence of HBE and anaesthetic influences	1121	15% of ISB and 23% of GA+ISB subjects are reported with HBE with odds of 30, after positioning to BCP. It appears that GA was associated with higher (excessive) risk over ISB.	Limited data available for GA subjects; since anaesthetic causes of hypotension incidences are simultaneously included, may over-estimate the true incidences.(⨁⨁⚪⚪-low, ⨁⚪⚪⚪-very low, ⨁⨁⨁⚪-moderate; for overall, ISB±sedation, GA±ISB subgroups.)f
Timing of bradycardia/HBE	848	Varied significantly in literature; 70% of study groups report the mean timing of adverse HR responses occurring after 30 minutes. Pooled data average timings are 33.6±24 minutes	SDs are high for the pooled data
Effect of β-agonists (epinephrine) on bradycardia/HBE incidences	988	No evidence of excessive risk of developing HBEs with use of β-agonists compared to subjects without its use.	Epinephrine was used either during ISB local anaesthetic block placement or for saline irrigation fluid of arthroscopy(⨁⨁⚪⚪-low; for overall or subgroups analysis)f
Effect of fentanyl on bradycardia/HBE incidences	775	No evidence of excessive risk of developing HBEs with use of fentanyl compared to subjects who did not receive it.	Only the studies which have used fentanyl in every subject, were included for analysis (⨁⨁⚪⚪-low, ⨁⨁⨁⨁-high, ⨁⨁⨁⚪-moderate; for overall, and for subgroup analysis, respectively)
Effect of prophylactic ondansetron and β-blockers on bradycardia/HBE incidences	395	Evidence of lower risk of developing HBE with the use; ondansetron may decrease the incidence by 4 times	Limited number of trials available for β-blocker prophylaxis (⨁⨁⨁⚪-moderate; for both outcomes)f
Serial HR measurements and effect of type of anaesthesia	1453, 1315, 802b	Administering GA or GA+ISB is associated with progressive fall of HR over time and this is maximum after 30 minutes under anaesthesia at BCP. Addition of ISB did not cause additional fall in HR.	Pooled measurements were considered
Serial HR measurements and effect of maintenance anaesthetic agent	1363, 1163, 580c	Subjects with TIVA-propofol and ISB subjects had least fall of HR, over time, in BCP.	Pooled measurements were considered
Serial HR measurements and effect of intraoperative pharmacological agent	Variable	Evidence of highest fall of HR with the use of fentanyl alone (for mid and delayed HR) or for concomitant use of fentanyl and PVIs (for early) is observed.	Limited data is available for fentanyl-PVIs concomitant effects.
Incidence of hypotension and type of anaesthesia (number of subjects)	2366	Evidence of higher ‘excessive’ risk for number of subjects who developed hypotension at BCP for subjects administered with GA over GA+ISB or ISB±sedation.	Incidences of ‘HBE’ were considered for ISB subjects
Incidence of hypotension and maintenance anaesthetic agent (number of subjects)	1251	Use of TIVA-propofol was not associated with excessive risk of developing hypotension over inhaled anaesthetics	Few of the TIVA-propofol group subjects had concomitant use of PVIs at the beginning of BCP
CDEs and maintenance anaesthetics	684	Maintenance anaesthetics can influence the CDEs; TIVA-propofol was associated with higher ‘excessive’ risks for number of subjects who experienced CDEs than inhalational agents.
CDEs and ISB anaesthesia	30	Shoulder surgeries done under ISB alone was associated with least incidences of CDEs.	Only one SG of this meta-analysis has been considered for CDE evaluation.
rSO2 and maintenance agents (absolute fall)	849	Absolute fall of rSO2 was not influenced by different maintenance anaesthetics; however, a non-statistically significant higher desaturation values were recorded for TIVA-propofol compared to inhaled anaesthesia subjects.	The immediate corrective therapy during a rSO2 fall may not reflect the actual differences
rSO2 and maintenance agents (lowest achieved)	599	Lowest achieved rSO2 was not influenced by different maintenance agents. However, a non-statistically significant higher desaturation values were recorded for TIVA-propofol compared to inhaled anaesthesia subjects.	The immediate corrective therapy during a rSO2 fall may not reflect the actual differences
rSO2 - HR relationships	381	Meta-correlations reveal that HR measurements from serial recordings of several study groups statistically correlated well with the respective rSO2 measurements	Statistical correlations were derived from consecutive, serial measurements.
SjvO2 - HR relationships	186	Meta-correlations reveal that HR measurements from serial recordings of few study groups statistically correlated well with respective SjvO2 values	Statistical correlations were derived from consecutive measurements but the strength of correlation was weak.
Influence of PVIs on HR	165	PVIs did not influence HR fall with in study subjects; however, the magnitude of HR fall was higher compared to control subjects.	Limited data available
Influence of PVIs on HR -rSO2/SjvO2 relationships	90	PVIs have not influenced the CDEs and HR-rSO2/SjvO2 relationships.	Limited data available
rSO2 - MBP relationships	457	Meta-correlation analysis revealed a statistically significant correlation between MBP and rSO2 values	Predictable outcome
Vasopressor consumptiond	503	Pooled averages of ephedrine requirements were higher for GA±ISB than GA alone to maintain the desired BP.	Limited data and non-parametric data comparisons.
HR of physiological matched controls	199e	HR increased or remained same in subjects after positioning to BCP.	Physiological controls are those who did not receive any pharmacological agents.

BCP – Beach chair position; BP - Blood pressure; CDE – Cerebral desaturation event; GA – General anaesthesia; GRADE – Grading of Recommendations Assesment, Development and Evaluation; HBE – Hypotension bradycardia episode; HR – Heart rate; ISB – Interscalene block; MBP – Mean blood pressure; PVI – Prophylactic vasopressor infusion; rSO2 – Regional oxygen saturation of brain; SD – Standard deviation; SjvO2 – Jugular venous oxygen saturation; TIVA – Total intravenous anaesthesia; aLiguori et al., defined HBE as HR <50 beats/min at anytime or <30 beats in <5 min compared to pre-anaesthetic state with or without hypotension, and/or decrease in SBP >30 mmHg in <5 min compared to pre-anaesthetic values, or any SBP decrease <90 mmHg; necessarily treated by ephedrine, epinephrine or atropine. b,cdata for early, mid and delayed heart rate dequivalent doses edata not included for total n of meta-analysis. fGRADE for primary outcomes

Literature identification

From 2306 studies that were initially screened, 661 potentially relevant manuscripts were selected based on the abstract. The details pertaining to literature identification are provided in the flow chart (Supplementary Digital Content File 2). Finally, 47 trials provided the data for analysis (from year 2000 to 2019).

Study characteristics

We included all SGs of manuscripts that provided HR data. Hence, the majority of manuscripts had two or more SGs. Supplementary Digital Content File 3 summarises the characteristics of SGs including Jadad scores. In total, there were 91 SGs for this review (n = 3107), 70 SGs detailed about serial HR measurements, additional to the adverse HR-responses. There were 67 randomised SGs (RCTs, n = 29). Supplementary Digital Content File 4 depicts the RoB graph and summary. Thirty-nine SGs were considered as study control groups and four as physiological study controls. One trial (year 1998)[26] was included against the PRISMA protocol, as the same was used by the rest of the authors to define HBE.

First analysis

Bradycardia and/or HBE

Bradycardia/HBE was reported in 24 SGs.[46121326272829303132] For defining 'bradycardia/HBE', primary authors used their own criteria for 8 SGs. The rest followed the definition by Liguori et al.[26] The incidence of isolated bradycardia[121327282930] varied from 0 to19% (n = 65 of 712, 9.1%) and that of HBE,[461213262829303132] 5 to 28% (n = 147 of 988, 14.9% in ISB subjects and n = 255 of 1121, 22.7% in ISB and GA subjects).

Meta-analysis of the incidence of bradycardia revealed risk ratio of 9.8 [(RR, 95%CI; 4.4, 21.9), I2= 0%, P < 0.0001] and HBE, RR of 19.6 [(95%CI; 10.7, 35.8), I2= 0%, P < 0.00001] in BCP. There was evidence of higher observed 'excessive risk' of developing adverse responses for GA subjects over ISB (RD P < 0.05, Figure 1].

Figure 1

Bradycardia (A) and HBE (B) meta-analysis forest plots. All hypotension incidences were included. BCP – Beach chair position; CI- Confidence interval; GA – General anaesthesia; HBE - Hypotension-bradycardia episode; ISB – Interscalene block; IV- Inverse variance; SE -Standard error

Primary authors proposed the possible associations of adverse HR-responses with various factors (epinephrine, fentanyl, ISB, norepinephrine, ondansetron or β-adrenergic blockers). Very low evidence was observed to confirm their effects on adverse HR-responses in ISB subjects. However, further analysis revealed that the use of β- adrenergic agonists[462632] and fentanyl[121328] did not increase risk of HBEs without its use [test for sub-group difference, P = 0.29, I2= 11.4% and P = 0.45, I2= 0%, respectively (Figure 2)]. Effect of prophylactic ondansetron (4-8 mg) in prevention of HBE was analysed in 2 trials;[1328] meta-analysis revealed OR (non-event, 95%CI) of 4.13 (1.89, 9.02, P = 0.0004). Effect of prophylactic use of β-blocker was used in one study[26]; meta-analysis revealed OR (non-event, 95%CI) of 5.8 [1.65, 20.36, P = 0.006 (Figure 3)]. In 17 SGs, the timing of bradycardia/HBE was documented. Pooled data showed the timing of occurrence as 33.6 ± 24 minutes.[461213262831] All BCP surgery subjects received midazolam, fentanyl or propofol sedation alone or in combination in ISB group at different doses and timings.

Figure 2

Effect of β-agonists (epinephrine) on bradycardia/HBEs. A. Forest plot for the use of epinephrine. B. Subgroup analysis forest plots for sub-groups using epinephrine and for those without. GA subjects are not included in this analysis. BCP - Beach chair position; CI - Confidence interval; GA – General anaesthesia; HBE - Hypotension-bradycardia episode; IV - Inverse variance. M-H - Mantel-Haenszel

Figure 3

Effect of drugs that can modify incidence of bradycardia/HBEs. Forest plots for fentanyl (C, D), ß-blockers (E) and ondansetron (F) on bradycardia/HBEs. GA subjects are not included in this analysis. BCP - Beach chair position; CI - Confidence interval; GA – General anaesthesia; IV - Inverse variance. M-H - Mantel-Haenszel

Second analysis

Post-BCP HR-responses analysed from serial HR measurements [Figure 4]

Our meta-analysis of HR-responses over time considered two sub-groups based on the type of anaesthesia and maintenance agents used. BL-HR was reported in 48 SGs (n = 1334); 12 used TIVA-propofol[163334353637] (73.7 ± 13.4 beats/min, n = 451), 33 received inhaled anaesthetics[16293538394041424344454647484950] (73.6 ± 13.6 beats/min, n = 744) and 139 subjects had ISB.[650] MDs between HR-values at supine (Pre-BCP) and post-BCP status are depicted in Figure 4.

Figure 4

Fall of HR over time, for pooled serial measurements under anaesthesia. The mean differences (MDs) are studied for the first 10 minutes, 11-30 minutes and after 30 minutes of beach-chair position from pre-BCP levels. For different types of anaesthesia (a) and maintenance agents (b), the trends are shown. GA - General anaesthesia; ISB - Interscalene block; TIVA - Total intravenous anaesthesia

Sensitivity analysis revealed that various anaesthetic agents significantly influenced fall in HRs. However, it made little difference to the overall results when study controls[293033343537394041424348495051525354555657585960616263] and randomised trials[461213151626272829303334353738394142444546475055606164] were analysed separately [Table 2]. Meta-regression was performed since primary outcomes, characterised by significant heterogeneity, yielded statistically significant omnibus P values for statistical models considering different maintenance agents and opioids.

Table 2

Sensitivity analysis

Factor/Covariate	Time	Number of subjects	Mean difference, HR (95% confidence interval)	I2	P* for χ2 heterogeneity	P, overall effect	Number of study groups
Use of PVIs	Early	107	3.67 (-2.4, 9.7)	83%	<0.00001	0.23	7
	Mid	58	-0.81 (-4.39, 2.78)	0%	0.94	0.66	3
Inducing agent, Propofol	Early	1333	2.72 (1.21, 4.23)	79%	<0.00001	0.0005	46
Inducing agent, Thiopentone	Early	120	2.18 (-0.95, 5.3)	0%	0.98	0.17	5
Remifentanil	Early	811	2.16 (0.35, 3.96)	79%	<0.00001	0.02	27
	Mid	444	5.62 (1.99, 9.25)	83%	<0.00001	0.002	18
	Delayed	229	8.8 (5.41, 12.18)	50%	0.05	<0.00001	8
Fentanyl	Early	111	6.02 (2.28,9.76)	78%	<0.00001	0.002	11
	Mid	296	6.98 (4.95,9.01)	0%	0.74	<0.00001	11
	Delayed	110	16.61 (13.01,20.21)	61%	0.03	<0.00001	5
Alfentanil	Early	80	4.59 (-1.03,10.22)	61%	0.05	0.11	4
	Mid	80	3.13 (-0.45,6.70)	0%	0.66	0.09	4
	Delayed	40	2.93 (-1.53,7.39)	0%	0.63	0.2	2
Sufentanil	Early	53	-0.8 (-5.41, 3.81)	NA	NA	0.73	1
	Mid	117	0.63 (-1.96,3.21)	0%	0.61	0.63	4
TIVA, propofol	Early	480	0.59 (-0.93,2.10)	0%	0.88	0.45	13
	Mid	340	2.72 (-1.64,7.07)	85%	<0.00001	0.22	15
	Delayed	100	3.55 (0.46,6.64)	0%	0.86	0.02	3
Sevoflurane	Early	826	3.55 (1.56, 5.53)	84%	<0.00001	0.0005	33
	Mid	736	5.84 (4.23, 7.45)	39%	0.02	<0.00001	27
	Delayed	420	9.54 (5.9, 13.2)	83%	<0.00001	<0.00001	16
Desflurane	Early	87	2.04 (-1.18, 5.25)	0%	0.89	0.21	4
	Mid	106	7.32 (3.64, 10.99)	40%	0.16	<0.0001	5
	Delayed	60	13.3 (9.72, 16.88)	0%	0.5	<0.00001	3
Randomised trials	Early	1072	2.41 (0.69, 4.14)	80%	<0.00001	0.006	41
	Mid	649	4.33 (2.81, 5.84)	27%	0.08	<0.00001	32
	Delayed	530	9.35 (6.09, 12.61)	82%	<0.00001	<0.00001	20
Study Controls	Early	709	1.73 (0.03, 3.43)	32%	0.07	0.05	23
	Mid	776	6.09 (3.54, 8.65)	75%	<0.00001	<0.00001	26
	Delayed	361	8.84 (5.78, 11.89)	61%	0.003	<0.00001	12

PVI – Prophylactic vasopressor infusion; TIVA – Total intravenous anaesthesia; HR – Heart rate; χ2-Chi-square; *critical P=0.05; significant P are bold and italicised

With regard to EHR,[615162933343537383940424344454647495152545556575865] MHR[162933343536384143464849505153545556575859606162636466] and DHR[616293038434647495155566065] responses, meta-analysis showed a statistically significant fall in HR in subjects with GA (GA or GA + RA, P < 0.0001). Sensitivity analysis and meta-regressions confirmed that fentanyl significantly influenced the HR drop over time (meta-regression, estimates, 14.8, 9.8 and 16.9; standard error (SE) 5.3, 4.3 and 2.8; P = 0.007, 0.024 and <0.001; for early, mid and delayed periods, respectively) in GA subjects (Omnibus P < 0.001. Also, refer 'publication bias', Supplementary Digital Content File 5).

Secondary outcomes

BP responses

BP responses were analysed from 67 SGs.[461213151628293031323334353637384041444546495051525354555660616466] Seven subjects were excluded from the primary study[16465866] even before surgery due to severe hypotension after BCP. For treatment of hypotension, ephedrine,[46121316283235364445495260] phenylephrine[5356606266] or combination of both[293940414246485459616365] were used. Less frequently used were cafedrine/theodrenaline,[3037] epinephrine,[263143] norepinephrine[64] and metaraminol.[1550] Number of subjects showing drop in BP was a better predictor for hypotension than absolute values. Supplementary Digital Content File 6 describes the details of hypotension with respect to type of anaesthesia or maintainance agent used at BCP.

CDEs

CDEs were evaluated in 33SGs.[15163334353945464853585960616465] Meta-analysis of pooled estimates showed statistically significant fall in absolute values of rSO2 with both TIVA-propofol[3335525961] and inhalational[151635394546485358606465] maintenance anaesthetics (P < 0.00001). There were no differences between sub-groups with respect to the type of maintenance agent used (P = 0.05). Lowest recorded values of CDEs[3334353945464753596065] and data on number of subjects who experienced CDEs[633343943465960616465] are detailed in Supplementary Digital Content File 7.

Relationship between rSO2, SjvO2 and HR

Seventeen SGs[33343546566065] evaluated the HR and rSO2 at specific intervals over the entire BCP period. Data were recorded as statistical averages for absolute values of consecutive timings. Meta-correlation-analysis showed correlation between the HR and rSO2 values (r = 0.608, 95%CI, 0.439 to 0.735, P < 0.001). Correlation was attempted between HR and SjvO2 absolute values from 12SGs.[16333435] Meta-correlation analysis revealed statistically significant but weak parallel correlation (r = 0.397, 95%CI, 0.151 to 0.597, P < 0.001) indicating an association between HR and SjvO2 values [Figure 5].

Figure 5

Meta-correlation-analysis depicting the relationship between rSO2(a), SjvO2(b) and HR. 95% confidence intervals are shown. HR - Heart rate; rSO2- Regional cerebral oxygen saturation; SjvO2- Jugular venous oxygen saturation

Use of PVIs and effect on HR, rSO2 and HR-rSO2 relationships,[15333435] details of physiological controls,[31535962] vaso-active drugs consumption[16333540444854566062] are detailed elsewhere (Footnote of Supplementary Digital Content File 3).

DISCUSSION

In our meta-analysis, we attempted to find the incidence and associations of adverse HR-responses during shoulder surgeries done in BCP. We observed the incidence of isolated bradycardia and HBE to be 9.1% and 14.9%, respectively. Current literature provides no concrete evidence linking different anaesthetic techniques, β-agonists or fentanyl with adverse HR-responses. Trials confirming the protective effects of ondansetron and β-blockers against HBEs are few in number. Our meta-analysis unequivocally confirms the influence of fentanyl on HR-drop over time in BCP-GA subjects. Furthermore, HR-rSO2/SjvO2 relationships in GA subjects are clearly elucidated.

The interpretations of adverse HR events may differ between GA and ISB subjects. The seemingly excessive risk of adverse events for GA over ISB subjects could be fallacious for several reasons. Anaesthetic or sedation related events, differences in incidence reporting among the included studies significantly influenced the data. Several authors have followed the definition of Liguori and colleagues,[26] where hypotension in isolation is considered an 'adverse event'. Sub-group meta-analysis has excluded ones that may have reported hypotension but not as a 'true' event of adverse HR-response. To avoid overlapping terms of these cardiovascular events, we analysed them separately. Any conclusion as to whether the hypotension/HBE event was directly linked to BCP or anaesthetic/non-anaesthetic agents remained elusive after this analysis, since every individual received a pharmacological agent in one form or the other. Inclusion of ISB subjects alone to account for adverse HR-responses was likely to reflect the true incidences. Presuming that every event was not the 'true' bradycardia/HBE among all, the actual incidence of bradycardia/HBE therefore, could be less than estimated.

The adverse HR-events were observed approximately between 10 to 50 minutes. Mechanisms related to peak plasma levels of local anaesthetics after ISB or blockade of cardiac sympathetic nerves via stellate ganglion were described. However, these mechanisms do not explain the adverse HR-responses in GA subjects. The claim in few trials regarding the augmentation of HBE risk by epinephrine has been with very low evidence. Epinephrine was administered either through skin infiltration, saline irrigation, concomitant to local anaesthetic or intra-articular injections. One study compared epinephrine to norepinephrine to study HBEs without a control group.[6] The paucity of data with respect to number of studies or type of drug (local anaesthetics, beta-agonists etc.) poses a limitation to any conclusion regarding risk modifying drugs. The factors like variable plasma levels with different routes of administration, short half-life etc., will not favour the specific timings of adverse events. Furthermore, we could not demonstrate higher incidences of adverse HR-responses for the fentanyl SGs over no-fentanyl in ISB subjects. Earlier studies have reported a dose-dependent increase in bradycardia/HBE incidences with fentanyl in BCP-cohorts.[812] The effects of fentanyl on HR were further validated by our second analysis of this study as we observed the highest HR-fall occurring with the use of fentanyl. Fentanyl acts on μ-opioid receptors on cardiac vagal neurons in the nucleus ambiguus and neurons preceding them to reduce GABAergic neurotransmission and induce bradycardia.[12] We believe, therefore, that adverse HR-response could be easily augmented with fentanyl use.

Association between CDEs and HR is as yet unreported. While HR is believed to be influenced by hypoxic events, defining HR-rSO2 relationship is not easy. Cerebral oxygenation may involve regional differences. The near-infrared reflectance spectroscopy is usually applied to frontal areas for convenience while actual rSO2 at the medullary vasomotor centre (VMC) is un-monitored. We have demonstrated a HR-rSO2/SjvO2 association through meta-correlation analysis. There is a dearth of literature on monitoring rSO2 during the ISB-BCP surgery with none reporting any adverse HR-responses. CDEs in ISB-BCP patients have been reported as incidences of 10%,[67] 3.3%[56] or lower absolute values of rSO2.[68] Higher partial pressures of oxygen during controlled ventilation may decrease the CDEs compared to spontaneously breathing (but sedated) ISB subjects. CDEs reported by Yadeau and colleagues[67] in RA patients showed no correlation with all hypotensive events. All ISB studies reporting bradycardia/HBE received intravenous fentanyl and midazolam singly or in combination. Furthermore, propofol infusion (sedation), β-blockers and oxygen (discretional) were randomly used in ISB subjects of this meta-analysis. Adverse HR-responses observed in ISB subjects, therefore, could be secondary to sedation and its CDE effects.[69]

We have limitations for our meta-analysis. From the available studies, we were unable to describe emergent strategy for preventing and managing adverse HR-responses during BCP-surgery, which is needed to inform practice. Non-availability of raw patient data or lowest achieved HR data for many trials precluded conducting individual patient meta-analysis or correlations. Heterogeneity is high in our study but we consider this acceptable since the pre-defined eligibility criteria for the meta-analysis are sound and the data are correct. While included trials might have allocated treatment randomly, their SGs inclusion in this review has not been random. Publication bias was minimal. However, inclusions of studies to this review were not based on Jadad scores.

CONCLUSIONS

Amalgamating the diverse and selective reporting of HR-responses in literature on shoulder surgeries in BCP, we observed lack of enough evidence for definitive associations of adverse HR-responses with different pharmacological agents like β-agonists or opioids. However, fentanyl can significantly influence HR-fall in BCP. Since HR-variations correlate well with monitored brain saturation values, the adverse HR-responses may also be induced by regional oxygenation of VMC in the brain, independent of anaesthetic agents. Close monitoring for CDEs could free the anaesthesiologist from concerns regarding the type of anaesthesia as well as intra-operative maintenance anaesthetic agents and ancillary drugs employed. However, further studies are essential to derive a cause-effect relationship with respect to adverse HR-responses. The key may lie in cerebral oxygenation levels at the VMC, and monitoring this parameter could set the direction for future research in this field.

Financial support and sponsorship

Grants, sponsors, and funding sources that provided direct financial support to the research work contained in the manuscript: None declared for all above mentioned authors.

Conflicts of interest

There are no conflicts of interest.

Note (Supplementary Digital Content File 3):

Analysis details: Data of studies which include both analysis; (1) HR data of adverse HR-responses ( first analysis); (2) HR data of HR-variability (second analysis). To define an 'adverse event', authors' own definitions have been used. Details of bradycardia/HBE or hypotension are shown in separate columns.

During second analysis of HR variabilities, we found no publication bias for EHR and DHR (Egger's test, P = 0.836 and 0.976, respectively) for included studies. However, funnel plot showed that the study by Meex et al.,[59] influenced the analysis (Egger's test, P < 0.001). Excluding this study resulted in non-significant Pvalue (P = 0.06) for MHR responses. However, inclusion of this study did not alter the overall outcomes for MHRs. Please see Supplementary Digital Content File 5 for 'Publication bias'.

Use of prophylactic vasopressor infusions (PVIs) and effect on HR, rSO: PVIs were used in 10 SGs. The certainty of the effects of PVIs on HR necessitated additional analyses on the control groups of each trial. Meta-analysis clearly demonstrated lower HR in BCP among SGs using PVIs as compared to those not using them (P = 0.004, within trials).[15333435] When SGs using PVIs were compared in pre- and post-BCP, lower HRs were not observed (P= 0.23, within SGs).[15333435] The overall association of PVIs vis-à-vis HR changes in BCP was non-significant (sensitivity analysis and meta-regression). Since we considered rSO2 values for the entire duration of surgery, no attempt was made to establish relationships between the two. Further, meta-correlation analyses were considered on HR-rSO2 relationships with and without use of PVIs. The use of PVIs did not make a difference (with PVIs use, r = 0.693, 95%CI, 0.391 to 0.860, P < 0.001, random effects, I2 = 72.5%, n = 90 and without PVIs use, r = 0.560, 95%CI, 0.332 to 0.727, P < 0.001, random effects, I2 = 80.81%, n = 291).

Relationship between rSO: Twenty SGs[1633343546],5660,[65] evaluated MBP and rSO2 at specific intervals over the entire BCP period. Data were recorded as statistical averages for absolute values of consecutive timings. Meta-correlation analysis showed statistically significant correlation between MBP and rSO2 values (r = 0.597, 95% CI, 0.432 to 0.723, P < 0.001, random effects, I2 = 79.9%) confirming the predictable relationship between the two.

Physiological controls and HR: Four SGs[31535962] evaluated HR responses over time. Meta-analysis demonstrated no change of HR after positioning to BCP (P = 0.58).

Vaso-active drugs consumption: Pooled averages of ephedrine requirements (mgs) were higher for GA ± ISB (n = 83) subjects[5456] than GA alone[1633354044486062] (n = 390) to maintain the desired BP (23.1 ± 32.1 vs 15.4 ± 27.3, per subject, respectively, P = 0.026). Ephedrine consumptions in inhalation anaesthesia[1635404448545660] (n = 396) and TIVA-propofol[163335] (n = 77) were 17.8 ± 31.1 and 12.4 ± 5.9 respectively (P = 0.236). CDEs, rSO2 and HR measurements were not analysed for vaso-active drug consumptions as the timings of administration were inadequately available.

Jadad scores: Variable Jadad scores (-2 to 5) were observed for included studies as inclusion of studies to this meta-analysis was not set for minimum scores. Inclusion of all studies would not change the incidences of adverse HR-responses. This is because all subjects of BCP-surgery were analysed pre– and post-BCP status in addition to comparative controls during analysis.

BCP - Beach chair position; BP - Blood pressure; CDE - Cerebral desaturation event; CI - Confidence intervals; DHR - Delayed heart rate; EHR - Early heart rate; GA - General anaesthesia; HBE - Hypotension bradycardia episode; HR - Heart rate; ISB - Interscalene block; MBP - Mean blood pressure; MHR - Mid heart rate; PVI - Prophylactic vasopressor infusion; rSO2 - regional cerebral oxygen saturation; SG - Study group; TIVA - Total intravenous anaesthesia.