Article Text
PDF
Abstract
Background Variation in clinical practice is often considered unwarranted when it does not reflect patient preference or evidence-based medicine. Complications from regional anesthesia such as nerve injury and systemic toxicity are dose dependent. It is currently unclear if there is significant variation with the dosing of local anesthetics (LA) in the context of the modern practice of peripheral regional anesthesia.
Methods We analyzed data from the International Registry of Regional Anesthesia that include prospective data on peripheral regional anesthesia procedures from 21 centers located around the world. Using data from years 2011 to 2017, our primary aim was to characterize the degree of variation in dosing of LA for the top 10 most commonly performed single injection peripheral nerve blocks. Our secondary aim was to identify potential drivers of this variation.
Results Among the 26 457 peripheral blocks performed, mean (±SD) LA dose per block in ropivacaine equivalents was 125.1±51.2 mg and 1.6±0.7 for mg/kg. There was large variation across all block types, with the highest variation (measured by interdecile range) in axillary blocks (143.8 mg) and lowest in interscalene blocks (83.3 mg). In a regression analysis, dose was primarily associated with the hospital (Cohen’s f=0.37) where the block was administered and block type (f=0.38), less so with age (f=0.02), weight (f=0.12), gender (f=0.05) or LA (f=0.17) used. Hospital site had strong impact on variation in LA dose (f=0.88). Variation was not significantly associated with number of blocks performed by hospital site.
Conclusions Large variation in dosing for regional blocks exists within and among hospitals, which is unlikely to be warranted. For many blocks, the variation of dosing is larger than the mean dose. Hospital site had strong impact on variation in LA dose and moderate impact on mean LA dose.
- Regional Anesthesia
- Quality
- Safety
Statistics from Altmetric.com
Introduction
Variations in the delivery of healthcare and health services is an important public health topic. Since the classic article by Glover1 in 1938 documenting the drastic differences in tonsillectomy rates in British schoolchildren, serious questions have arisen among policymakers regarding the efficiency and effectiveness of healthcare. Variations in care can be considered unwarranted or warranted.2 3 For instance, when clinical care has proven to be scientifically effective with little risk (eg, aspirin therapy following a myocardial infarction), any variation is unwarranted and in need of correction. Often, unwarranted variation is driven by care models that are known as supply sensitive, such as visits, diagnostic tests, imaging exams, hospitalizations, intensive care unit stays, and so on, where frequency of use is strongly influenced by capacity of system. Variation in care may be considered warranted if it reflects patient preference care, such as the decision to undergo joint replacement surgery.
Little is known about variations in the use of local anesthesia for peripheral regional anesthesia. Large volume and dose peripheral nerve blocks (PNB) have been the norm within the regional anesthesia community for decades, despite calls for dose reduction based on innovations of image guidance.4 Analysis of large clinical registries examining both nerve injuries and systemic toxicity indicates that adverse patient outcome following PNBs is associated with high doses of local anesthetic (LA).5–7 Systemic toxicity is a dose-dependent phenomenon, and, therefore, using the smallest effective dose would seem ideal.
If variation in LA dose is limited, it would perhaps suggest professional consensus regarding effective care. If significant variation exists, it would likely be unwarranted and reflect on factors embedded within individual health systems and/or providers, rather than on specific patient preferences.
Given that there is a paucity of scientific agreement regarding the relationship between LA dosing, efficacy, and safety, we hypothesized that we would be able to demonstrate significant variation for dosing of LA during PNBs. Using the International Registry of Regional Anesthesia (IRORA), our primary aim was to characterize this variation (or lack thereof) across different health systems and block types. Our secondary aim was to identify potential drivers of this variation.
Methods
The IRORA is a prospective multicenter clinical registry that collects information on contemporary PNB practice. The study period for this current investigation is from 1 June 2011 to 31 December 2017, based on submissions from 21 contributing institutions (see online supplementary appendix 1 for participating institutions). Features of the previously described clinical registry3 include a waiver of the consent for data collection so that all eligible patients can be captured minimizing selection bias, systematic postoperative contact with all patients and a data quality assurance process that includes site visits, interviews, and chart reviews.
Supplemental material
Selection of data entries
As the purpose of this investigation was to explore variation in dose, and an individual block dose could be influenced by performing multiple blocks at the same time, we restricted the analysis to unitary PNBs. A total of 28 783 unitary blocks were identified, and the 10 most common blocks used for the variation analysis made up 26 457 of the blocks.
Data preparation for analysis
Given the goal of comparing dosing and variation we converted and summarized all LA given to ropivacaine equivalents (REq) in milligrams. Bupivacaine was converted to REq by a multiplication factor of 1.4, lidocaine and prilocaine divided by 3, and levobupivacaine by 1.1.8 9
Statistical analysis
Data are presented for categorical variables as counts and percentages, continuous variables are shown as mean and SD or medians and interpercentile ranges. Categorical data were analyzed with χ2 test and continuous data with analysis of variance (ANOVA) and Kruskal-Wallis tests. All continuous variables were analyzed using Hartigan’s dip test, and when showing multinodal distributions (total LA and LA dose per kilogram) results were reported as medians versus means and SD.
Primary outcome
As a measure of the primary outcome, variation in LA dose, we used the interdecile range (IDR) to capture the range that encloses 80% of the LA total dose in REq milligrams for the block. Calculated similar to the IQR, IDR is the 90th percentile minus the 10th percentile. As no standard metrics have been established to describe dose variation, we chose IDR over IQR as the latter only contains the middle 50% of all data points and appeared inadequate for describing variation. The central 80% is a good compromise between IQR and total range that overemphasizes potential individual outliers. For the variation model, the a priori chosen covariates were hospital site, block type, median dose and number of blocks administered by site.
Secondary outcome
To evaluate the impact of potential drivers of dose and variation, we calculated Cohen’s f as effect size estimates for independent variables. Effect sizes of small, medium and large correspond to Cohen’s f in one-way ANOVA regressions as 0.1, 0.25 and 0.4. In the dose model, the dependent variable was the total dose in REq and the covariates were a priori chosen to include block type, hospital site, type of the primary LA used, patient age, weight and gender.
Sensitivity analysis
As some centers contributed smaller numbers to the database, we performed a sensitivity analysis using data from the five centers contributing most data (over 1400 blocks each) to investigate if variation or effect sizes are impacted by smaller sites with fewer blocks submitted.
All analyses were done using R (V.3.51, Vienna, Austria) as distributed by Debian Linux using the sjstats (V.1.4.1) library. For all analyses, we set the p value for statistical significance to 0.05 (two sided). Data were cleaned, aggregated and summary statistics calculated with Python (V.3.7) using the Pandas (V.0.2.3.4) library. Images were generated with Python Seaborn (V.0.9.0.1) image library.
Results
Hospitals performed between 98 and 7260 blocks, with a median of 563. The mean age of patients was 57.4±18.6, with fewer females, 12 615 (47.7%) than males, 13 842 (52.3%). Average weight was 83.3±21.6 kg. The most commonly used LA was ropivacaine (n=21 141, 80%), followed by bupivacaine (n=2601, 9.8%), lidocaine (n=2074, 7.8%), levobupivacaine (n=407, 1.5%) and prilocaine (n=207, 0.8%). Mean±SD LA doses (mg) in REq were 125.4±51.0 per block with an REq per kg of 1.6±0.7. Mean doses decreased from 2011 (137±60.4) to 2016 (120.4±49.3) annually. Similar data are displayed in table 1 by regional block type. The 10 most common blocks in the database were, in order, interscalene (n=4939), femoral (n=4877), supraclavicular (n=3513), adductor canal (n=2849), popliteal (n=3175), axillary (n=2084), fascia iliaca (n=1748), ankle (n=1280), paravertebral (n=1146) and forearm (n=846). Age, gender and dose were different for block types (p<0.001). The block associated with the oldest average age was the fascia iliaca block (68.7±18.2) versus the forearm block with the youngest mean age (44.3±18.4). Highest average block dose was for axillary blocks, 170.4±56.9 in REq, and lowest for forearm blocks, 90.3±45.7 in REq (illustrated as boxplots with median and IDR by block and site in figure 1).
Characteristics of blocks by block type
Boxplot of local anesthetic (LA) ropivacaine equivalents by block type for each site. The box contains the 25th −75th percentiles, the horizontal line in the box represents the median. Whiskers have been set to represent the interdecile range containing the central 80% (10th–90th percentiles). Sites are anonymized by letter coding. AC, adductor canal; Ax, axillary plexus; FA, forearm; Fem, femoral; FI, fascia iliaca; IS, interscalene; Pop, popliteal sciatic; PV, paravertebral; Supra, supraclavicular.
Variation in dosing (IDR of total block dose in REq) was highest (143.8) for axillary blocks and lowest for interscalene blocks (83.3). When considering IDR in relation to median LA dose the forearm (1.4), paravertebral (1.1), and adductor canal (0.98) blocks had the highest ratio, whereas the interscalene block had the least (0.67) (table 1).
Total block dose was significantly associated with hospital site (p<0.0001, Cohen’s f=0.37), block type (p<0.0001, f=0.37), primary LA (p<0.0001, f=0.17) and weight (p<0.0001, f=0.12) but not with age or gender.
IDR was strongly associated with hospital site (p<0.0001, Cohen’s f=0.88) and median dose (p<0.0001, Cohen’s f=0.27) while block type and number of blocks performed at that hospital were not significant.
Discussion
Considerable variation of LA dosing exists within and among IRORA hospital sites (figure 1 provides a strong visual impression of the variation). The primary determinants of total block LA dose are block type and hospital site, while variation was primarily impacted by hospital site. It did not appear to be related to number of blocks performed at a particular site.
Although the paucity of evidence-based recommendations for LA dosing may drive a lack of agreement in clinical strategies, practice variation in each health system is a contributing factor. This contribution of each hospital/health system to dosing variation is likely explained by unique cultural factors and local traditions. Apart from some extremes of low-weight adults, most block dosages are not limited due to weight-based toxicity concerns, suggesting that dosing decisions may be influenced by what is perceived to the be the optimal dose to achieve the desired effect. As illustrated in figure 2, many blocks have a spike in the dosing distribution at around 150 mg ropivacaine which happens to be the dose contained in the manufacturer provided 30 mL vial of 0.5% ropivacaine. For roughly half the blocks, the range (IDR) of doses actually equals or exceeds the median dose, demonstrating a dramatic variation in practice.
Distribution of local anesthetic total doses (in ropivacaine equivalents (mg)) for the 10 most common blocks.
The described variation in block dose varies by block type, pointing towards varying levels of evidence and agreement regarding dosing (which is not a patient-driven choice) depending on regional block administered. Axillary blocks (showing the greatest absolute variation of 143.8 mg REq and a mean dose of 170.4 mg REq) have been shown to be successful with a 5 min onset time with less than 30 mg REq.10 This discrepancy of practice and published data may come from a lack of concern regarding LA toxicity, as larger doses have been shown to have LA plasma levels below what is considered the maximum allowable dose.11 Maximum allowable doses though are a flawed concept, as they lack scientific evidence, are extrapolated from animal data, and are unspecific for type of blocks.12 Rosenberg et al stated, ‘only clinically adequate and safe doses (ranges) that are block specific are justified, taking into consideration the site of LA injection and patient-related factors.’12 Indeed, in the setting of intravenous regional anesthesia (Bier blocks), seizures have been reported with 1.4 mg/kg of lidocaine or 1.3 mg/kg of bupivacaine (or about 33% and 50% of the ‘safe’ dose) and cardiac arrest with 2.5 mg/kg of lidocaine or 1.6 mg/kg of bupivacaine (50% of the ‘safe’ dose).13 As in other fields of medicine, the minimum effective dose to achieve the desired effect ought to be given versus administering what is provided in the manufacturer’s vial or what is considered to be non-toxic based on inadequate evidence. While our data do not provide evidence to base new guidelines upon, blocks with the highest variation of dosing identified here within ought to be the primary targets for guidelines or research efforts.
Clinical registries in regional anesthesia, most prominently IRORA5–7 and the Pediatric Regional Anesthesia Network,14 15 were created as prospective data collections with web-based data entry including data validation and review to ensure reliable and accurate information. They have in the past been primarily used to detect rare adverse events and provide insight into their frequency, and by doing so have impacted practice recommendations.15–17 This manuscript is taking peripheral regional anesthesia data a step further in order to detect unwarranted variation in care. Our data suggest a knowledge gap in evidence, as variation to this large degree likely reflects uncertainty in optimal clinical strategies.
The results of this study are mostly limited by the lack of efficacy data of block success and other associated quality metrics such as measures of physical function. The correlation of variation and mean dose with such outcomes is highly desirable, but not within the scope of this manuscript. Results may be further confounded by the lack of data regarding the intended goal of the block as doses for postoperative analgesia may differ from a block placed for surgical anesthesia. As sites may have different rates of blocks for surgical anesthesia, this could have contributed to dose variation. Furthermore, results may have been impacted by the geographic representation, approximately 70% of the block data originated from Australia, New Zealand and Malaysia; 30% from the USA.
In summary, large variation in dosing for regional blocks exists within and among hospitals, which is unlikely to be warranted. For many blocks, the variation of dosing is larger than the mean total dose. Dose is as expected dependent on block type, but to a similar degree on the hospital where the block was administered. Hospital site had strong impact on variation in LA dose and moderate impact on mean total LA dose.
Acknowledgments
We thank all contributing member hospitals of IRORA (see online supplementary appendix 1).
References
Footnotes
Contributors All authors contributed in the planning, design, data analysis and interpretation as well as writing of this manuscript.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval This study was approved by the Dartmouth College Committee for the Protection of Human Subjects (IRB No STUDY00030887, 17 February 2018) and St Vincent’s Hospital, Melbourne, Human Research and Ethics Committee (Low-Risk Research 107/17).
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article.