Safety of Botulinum Toxin A Injections for Facial Rejuvenation: A Meta-Analysis of 9,669 Patients

Abstract

Purpose: 

To quantitatively evaluate safety profile for botulinum toxin A (BTX-A) injections among patients undergoing treatment for cosmetic indications is produced, with special attention to clinically relevant covariates and their relative impact on safety.

Methods: 

A systematic literature search was performed using PubMed (1996-January 2020) and Embase (1947-January 2020) to identify all randomized controlled trials (RCTs) that reported safety data for patients receiving BTX-A for cosmetic indications compared to placebo. A meta-analysis was performed to determine pooled risk ratios (RR) for treatment-related adverse events (TRAEs) and for specific adverse events. Meta-regression and additional analyses were performed for significant and/or clinically relevant covariates.

Results: 

Following the review of 8,690 studies, 32 RCTs involving 9,669 patients were included. The pooled RR of any TRAE occurring after BTX-A injection compared to placebo injection was 1.53 (95% CI, 1.33–1.77; p < 0.001). Statistically significant covariates included individual injection volume and total injection volume. The type of BTX-A formulation, treatment site, total BTX-A units, and BTX-A units per injection were not significant. Specific adverse events more likely to occur following BTX-A injection rather than placebo injection included eyelid/eyebrow malposition (RR 3.55; p < 0.001), facial paresis (RR 2.42; p = 0.316), and headache (RR 1.45; p = 0.003). Injection site reactions and injection site bruising occurred at similar rates in both groups.

Conclusions: 

The overall safety profile of BTX-A is acceptable and consistent with previous publications. The authors’ additional analyses provide a relative comparison of the impact of various treatment parameters on safety.

In Brief

This meta-analysis indicates that botulinum toxin A injections for cosmetic indications carry a minimal risk of treatment-related adverse events and identifies potentially modifiable covariates that may help improve the safety profile.

Botulinum toxin (BTX), a neurotoxin produced by the anaerobic spore-forming Clostridium botulinum, is the most poisonous substance known.^1,2 There are seven biochemically and serologically distinct botulinum toxins, classified as type A through G, and the human nervous system is susceptible to five serotypes (BTX-A, B, E, F, and G).² A therapeutic use for botulinum toxin type A (BTX-A) emerged in the early 1970s based on the work of an ophthalmologist, Alan B. Scott, who demonstrated the efficacy of the toxin as a nonsurgical treatment for strabismus by injecting it directly into extraocular muscles.^3,4 Since then, a multi-million-dollar industry has emerged for BTX-A injections for various medical and cosmetic indications, spanning across multiple disciplines of medicine.^5,6

There are three predominant formulations used in North America with approval from regulatory bodies.⁷ These include onabotulinumtoxinA (Botox, Vistabel; Allergan Inc., Irvine, California), incobotulinumtoxinA (Xeomin, Bocouture; Merz Pharma, Frankfurt, Germany), abobotulinumtoxinA (Dysport, Azzalure; Galderma Laboratories, Fort Worth, Texas), and prabotulinumtoxinA (Jeuveau, Evolus, Inc; Newport Beach, California).⁷ While each is purported to have unique features, it is unclear whether there are clinically important differences between the formulations, even with the results of a recent comparative meta-analysis of randomized controlled trials (RCTs).⁷

Given the widespread use of BTX-A for cosmetic indications, particularly in facial and periocular regions, it is important to understand the risk of adverse events for patients who are interested in undergoing treatment. While previous meta-analyses have investigated the safety profile of BTX-A, several years have passed since their publication. Importantly, the direct and relative impact of modifiable variables on the safety profile remains unanswered.^7–11 Thus, the objective of this study was to produce an up-to-date quantitative risk profile for this intervention, with special attention to the impact of various clinically relevant covariates such as the BTX-A formulation used and the injection technique. We hypothesized that BTX-A would have an acceptable risk profile that showed some variation based on the covariates of interest.

METHODS

This meta-analysis of randomized controlled trials (RCTs) was conducted and reported according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The protocol has been registered to the International Prospective Register of Systematic Reviews (PROSPERO) (ID: 2020 CRD42020167129).

Eligibility Criteria.

Eligible studies included RCTs of patients who received facial BTX-A injections for cosmetic indications. Studies were included in the meta-analysis if they met all of the following inclusion criteria: (1) original data pertaining to adverse events following BTX-A injection; (2) a comparison group which received a placebo (e.g. saline or vehicle); and (3) adult patients (>17 years old). The following exclusion criteria were used in the selection process: (1) noncosmetic indications for BTX-A treatment; (2) nonfacial locations of treatment; (3) coadministration of other cosmetic treatments and/or procedures (e.g., filler); (4) fewer than 5 patients; (5) language other than English; (6) unpublished studies (e.g., conference abstracts); and (7) no original data reported (e.g., systematic reviews, reanalysis of previously reported data).

The primary and secondary outcomes of this meta-analysis were the total number of treatment-related adverse events (TRAEs) and specific adverse events, respectively. Studies that reported safety data in inadequate detail to be incorporated into the analysis of these outcomes were also excluded.

Information Sources and Search Strategy.

Systematic literature searches were completed using the PubMed (1996-January 2020) and Embase (1947-January 2020) databases. Various search terms were constructed using synonyms of “botulinum toxin type A” and the location (e.g., “facial”) or indication (e.g., “cosmetic”) of treatment, which were then combined using Boolean operators within the search builder of each database. The full electronic search strategy for each database is available in Appendix 1, Supplemental Digital Content 1, available at https://links.lww.com/IOP/A329. The reference lists of included articles were also reviewed for additional eligible studies that may not have been captured with the search strategy.

Study Selection.

The study selection process was conducted using Covidence^TM (Covidence Systemic Review Software; Veritas Health Innovation, Melbourne, Australia). The initial title and abstract screen were performed in parallel by all three authors, with two votes required per decision. A full-text review was then conducted for the remaining articles, again with two votes required per decision. Articles were included in the meta-analysis if a consensus was reached with both votes.

Data Collection and Handling.

A data extraction form based on the Cochrane Consumers and Communication Review Group’s Data Extraction Template was developed using Microsoft Excel software. Data extraction was performed by one author (MG) and repeated to ensure accuracy. The following data was collected from each study: study characteristics (year of publication, location, number of centers, design, sample size, duration of follow up); patient demographics (age, sex, history of BTX-A treatment, treatment indication); injection parameters (BTX-A formulation, volume, units, number of injections); and adverse event data (type and frequency). Authors of studies were not contacted directly for additional information or data, although appendices and supplemental materials were reviewed to collect all available safety data.

The frequency of adverse events by study group was collected as a raw value rather than a summary measure if both were reported within a given study. However, there was variability in how this was reported, with only some studies making the distinction between treatment-related adverse events (TRAEs) and treatment-emergent adverse events (TEAEs). The latter refers to any event that was not present prior to treatment or any preexisting event that worsened following treatment, whereas the former specifically refers to adverse events which are deemed to be related to the treatment by investigators. Since the primary outcome of interest was best represented by TRAEs, this value was used when specified. If not specified, the total number of adverse events, as reported by investigators, was assumed to represent TRAEs as this approach was less likely to underestimate safety outcomes. When only TEAEs were reported in a given study, this value was used to represent TRAEs. Some studies also made the distinction between the number of adverse event occurrences and number of patients experiencing adverse events (i.e., one patient may have experienced the same adverse event multiple times or multiple events). When specified, the number of patients was collected rather than the number of events as this value was clinically applicable to scenarios in which a patient is considering BTX-A treatment. If not specified in an included study, the reported frequency was assumed to represent the number of patients.

Due to the variability in the naming of specific adverse events (i.e., the secondary outcome) among the included studies, similarly named and/or closely-related adverse events were grouped together into the following categories: injection site bruising, hematoma, or hemorrhage; injection site reaction (edema, erythema, pruritis); eyelid and/or eyebrow malposition; headache (including migraine and other descriptors of headache); and facial palsy or paresis. The frequency of an adverse event was only collected if it was specifically reported such that the absence of any given event within a manuscript was not assumed to mean that the event did not occur. Adverse events representing distal paralysis or paresis due to systemic extension of BTX-A were also collected.

Publications reporting multiple studies with separate sets of data, such as those with two study arms, were treated as separate studies in the meta-analysis. For studies with more than one BTX-A treatment group and only one placebo group, the BTX-A groups were combined into a single group to overcome unit-of-analysis error. In cases when the BTX-A groups received the same formulation in different doses, the variables detailing the injection technique were combined using weighted averages based on sample size. In cases when BTX-A groups received different BTX-A formulations, the variables detailing the treatment technique were not incorporated into the analysis.

Risk of Bias Assessment.

Risk of bias assessment for each individual RCT was performed using the Cochrane Risk of Bias Tool which critiques the following domains: (1) Random Sequence Allocation, (2) Allocation Concealment, (3) Blinding of Participants and Personnel, (4) Blinding of Outcome Assessment, (5) Incomplete Outcome Data, (6) Selective Reporting, and (7) Other Bias. The “Other Bias” domain was specified a priori as whether safety events were monitored in an active (i.e., patients were specifically prompted or instructed to report safety outcomes) or passive (i.e., events only recorded if patients voluntarily reported them) fashion. The risk of bias assessment was performed at the outcome level rather than at the study level. This was because the outcome of interest for this meta-analysis (i.e., safety data) was treated as a secondary outcome in many of the included studies, such that they were designed and blinded for efficacy assessments rather than safety assessments. The results of the risk of bias were considered in the interpretation of this meta-analysis and were also included in the statistical analysis as a binary covariate based on whether a study was deemed to have a low risk of bias in ≥4 of the 7 domains.

Additional analyses were performed to determine the risk of bias across studies. These included a funnel plot along with both Egger’s and Begg’s tests for small-study effects to quantitatively assess the level of publication bias. A chronologically sorted cumulative meta-analysis and an influence analysis were also performed to determine whether there were any study outliers, and if so, whether they represented a risk of bias across studies.

Data Synthesis and Analysis.

A random-effects meta-analysis was performed to obtain a pooled effect estimate for the primary and secondary outcomes. The principal summary measure used in this analysis to quantify these outcomes was a risk ratio. Statistical heterogeneity was assessed via the I² statistic, which estimates the percentage of observed variation in study-specific estimates that can be explained by between-study heterogeneity. The traditional thresholds of 30–60%, 50–90%, and 75–100% were used to designate moderate, substantial, and considerable levels of between-study heterogeneity. The 0–40% range was used to designate low and unimportant levels of heterogeneity.

A meta-regression was also performed to identify any study variables that had a statistically significant association with the primary outcome. If present, a subgroup meta-analysis was performed to assess the impact of the significant covariate on the pooled effect estimate. Subgroup meta-analyses were also performed for covariates of interest, even if not statistically significant. Sensitivity analyses (influence analysis and cumulative meta-analysis) were also performed to identify outlier studies. All analyses were performed using Stata 14.0 (Texas, USA) software and a p value of less than 0.05 was considered statistically significant.

RESULTS

Study Selection.

The study screening and selection process is outlined, according to PRISMA guidelines, in Fig. 1. The cumulative search results from Pubmed and Embase yielded 8,690 potentially eligible studies. Following the removal of 663 duplicates and 7,734 nonrelevant titles or abstracts, 293 full-text articles were reviewed for inclusion. The most common reason for exclusion following full-text review was the absence of a control group that received a placebo injection (n = 148). Several conference abstracts (n = 56) were also removed, suggesting that there may be a considerable amount of unpublished safety and/or efficacy data pertaining to BTX-A injections for cosmetic indications. The study selection process yielded a total of 32 studies (29 manuscripts; 3 manuscripts reported results from 2 studies) that were included in the meta-analysis.

FIG. 1.: Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flow chart outlining the study selection process with reasons for exclusion of full-text articles.

Study Characteristics.

The characteristics of the included studies are presented in Table 1.^12–43 All 32 of the included studies were randomized controlled trials with parallel designs, published between 2004 and 2020. Most studies (n = 30) were multi-center in nature and conducted primarily in North America (n = 16) and Europe (n = 9). The studies collectively assessed 9,669 patients with a mean follow-up duration of 156 days (range 28–252 days).

TABLE 1. - Study characteristics of all articles included in meta-analysis

Study details								Treatment characteristics & injection parameters				Control/Placebo group			BTX-A group
Study	Centers	Continent	Design	Follow Up (Days)	Patients	Females (%)	Age (Mean ± SD)	Indication	Treatment Naïve?	Needle Gauge	Number of Injections	Patients	Volume Per Injection (mL)	Total Volume (mL)	Formulation	Patients	Volume Per Injection (mL)	Units Per Injection (U)	Units Per Injection, Wt Avg. (U)	Total Volume (mL)	Total Units Per Session (U)	Total Units Per Session, Wt. Avg.(U)
Alimohammadi et al. 12	Single	Europe	Parallel RCT	168	16	16 (100)	NR	GL	Yes	26	5	4	0.1	0.5	ONA	12	0.1	1/2/4	2.33	0.5	5/10/20	11.67
Ascher et al. 16	Multi	Europe	Parallel RCT	180	119	114 (95.8)	NR	GL	Yes	29	5	17	0.05	0.25	ABO	102	0.05	2		0.25	50
Ascher et al. 15	Multi	Europe	Parallel RCT	109	100	94 (94)	49.8 ± 8.2	GL	Yes	29	5	50	0.05	0.25	ABO	50	0.05	2		0.25	50
Ascher et al. 14	Multi	Europe	Parallel RCT	180	185	160 (86.5)	NR	GL	No		5	60	0.05	0.25	ABO	125	0.05	10		0.25	50
Ascher et al. 13	Multi	Europe	Parallel RCT	113	176	176 (100)	47.3	GL	Yes		5	35	0.05	0.25	ABO	141	0.05	5/10/15	9.71	0.25	20/50/75	48.55
Beer et al. 13 (EV-001)	Multi	NA	Parallel RCT	150	330	306 (92.7)	NR	GL	No		5	84	0.1	0.5	PRABO	246	0.1	4		0.5	20
Beer et al. 13 (EV-002)	Multi	NA	Parallel RCT	150	324	290 (89.5)	NR	GL	No		5	78	0.1	0.5	PRABO	246	0.1	4		0.5	20
Bertucci, et al. 18	Multi	NA	Parallel RCT	252	609	532 (87.4)	NR	GL	No		5	204	0.1	0.5	DAXI	405	0.1	8		0.5	40
Brandt et al. 19	Multi	NA	Parallel RCT	180	158	23 (15)	42.9 ± 9.9	GL	Yes	30	5	53	0.05	0.25	ABO	105	0.05	10		0.25	50
Carruthers et al. 22	Single	NA	Parallel RCT	28	40	40 (100)	NR	GL, CFL, FHL	No		16 (GL: 5, FHL: 5, CFL: 6)	20	0.06	0.96	ONA	20	0.06	4		0.96	64
Carruthers et al. 25 (SAKURA 1)	Multi	NA	Parallel RCT	252	275	262 (86.5)	NR	GL	No		5	102	0.1	0.5	DAXI	201	0.1	8		0.5	40
Carruthers et al. 25 (SAKURA 2)	Multi	NA	Parallel RCT	252	306	270 (88.2)	NR	GL	No		5	102	0.1	0.5	DAXI	204	0.1	8		0.5	40
Carruthers et al. 20	Multi	Multiple	Parallel RCT	150	445	384 (86.3)	46.4 ±	CFL	Yes		6	223	0.1	0.6	ONA	222	0.1	4		0.6	24
Carruthers et al. 23	Multi	NA	Parallel RCT	252	191	167 (87.4)	NR	GL	No		5	35	0.1	0.5		214	0.1			0.5
Carruthers et al. 21	Multi	NA	Parallel RCT	120	267	238 (86.2)	NR	GL	No		5	92	0.1	0.5	INCO	184	0.1	4		0.5	20
De Boulle et al. 18	Multi	Multiple	Parallel RCT	180	787	702 (89.2)	46.9 ±	GL, CFL, FHL	Yes		16 (GL: 5, FHL: 5, CFL: 6)	156	0.1	1.6	ONA	631	0.1	4 (GL, FHL), 6 (CFL)	3.26	1.6	40 (FHL: 20, GL: 20) /64 (FHL: 20, GL: 20, CFL: 24)	51.95
Fagien et al. 28	Multi	Multiple	Parallel RCT	180	391	336 (85.9)	44 ±	GL, FHL	Yes		10 (GL: 5, FHL, 5)	101	0.1	1	ONA	290	0.1	4		1	40
Hanke et al. 29	Multi	NA	Parallel RCT	120	271	254 (93.7)	46.5 ±	GL	No		5	89	0.1	0.5	INCO	182	0.1	4		0.5	20
Harii et al. 30	Multi	Asia	Parallel RCT	112	140	126 (90)	45.7 ± 9.1	GL	Yes		5	49	0.1	0.5	ONA	90	0.1	2/4	3.01	0.5	10/20	15.05
Kane et al. 31	Multi	NA	Parallel RCT	150	816	719 (88.1)	NR	GL	No	30	5	272		0.51*	ABO	544				0.51 *	50/60/ 70/80	61.16
Kerscher et al. 32	Multi	Europe	Parallel RCT	120	156	135 (86.5)	NR	GL, CFL, FHL	No	30, 32	16 (GL: 5, FHL: 5, CFL: 6)	51	0.092*	1.47	INCO	105	0.031	2 (CFL), 2-4 (FHL), 4 (GL)	3.69	1.47	54 to 64	59
Lowe et al. 33	Multi	Europe	Parallel RCT	180	162	144 (88.9)	NR	CFL	No		6	32	0.1	0.6	ONA	130	0.1	1/2/4/6	3.24	0.6	6/12/ 24/36	19.43
Moers et al. 34	Multi	NA	Parallel RCT	210	917	809 (87.6)	49.5 ± 9.5	CFL, GL	Yes		11 (GL: 5, CFL: 6)	306	0.1	1.21	ONA	611	0.1	4 (GL), 6 (CFL)	3.09	1.1	24 (CFL: 24) /44 (CFL: 24, GL: 20)	33.94
Monheit et al. 35	Multi	NA	Parallel RCT	120	373	313 (83.9)	42 ± 10	GL		30	5	94	0.05	0.25	ABO	279	0.05	4/10/15	9..75	0.25	20/50/75	48.75
Monheit et al. 36	Multi	NA	Parallel RCT	150	300	260 (87)	44.2	GL	Yes		5	100	0.05	0.25	ABO	200	0.05	10		0.25	50
Rivers et al. 37	Multi	NA	Parallel RCT	120	125	98 (83.8)	NR	GL, CFL	Yes		11 (GL: 5, CFL: 6)	57	0.1	1.1	ONA	60	0.1	4		1.1	44 (GL: 20, CFL: 24)	44
Rzany, et al.39	Multi	Multiple	Parallel RCT	150	540	476 (88.1)	49	GL	No		5	49	0.1	0.5		491	0.1			0.5
Rzany et al. 38 (Study Arm 1)	Multi	Europe	Parallel RCT	112	110	98 (89.9)	46.6 ± 9.2	GL			3	37	0.05	0.15	ABO	73	0.05	10		0.15	30
Rzany et al. 38 (Study Arm 2)	Multi	Europe	Parallel RCT	112	111	100 (90.1)	46.4 ± 8.1	GL			5	38	0.5	0.25	ABO	73	0.05	10		0.25	50
Solish et al. 40	Multi	NA	Parallel RCT	180	175	152 (56.9)	46.8 ± 9.8	FHL, GL	No		10 (GL: 5, FHL, 5)	59	0.1	1	ONA	116	0.087	2/4	3.49	0.75	30 (FHL: 10, GL: 20)/40 (FHL: 20, GL: 20)	34.92
Wu et al. 43	Multi	Asia	Parallel RCT	120	227	195 (85.9)	42.3	GL	Yes	30	5	57	0.1	0.5	ONA	170	0.1	4		0.5	20
Wu et al. 42	Multi	Asia	Parallel RCT	150	417	360 (86.3)	46.4 ± 9.57	CFL	No		6	101	0.1	0.6	ONA	316	0.1	4		0.6	24

^*Weighted Average.

ABO, AbobotulinumtoxinA; CFL, Crow’s Feet Lines; DAXI, DaxibotulinumtoxinA; FHL, forehead lines; GL, Glabellar Lines; NA, North America; NR, not reported; ONA, OnabotulinumtoxinA; PRABO, PrabobotulinumtoxinA; RCT, randomized controlled trial; SD, standard deviation.

The BTX-A formulations that were assessed among the included studies were onabotulinumtoxinA (n = 12), abobotulinumtoxinA (n = 10), prabobotulinumtoxinA (n = 2), daxibotulinumtoxinA (n = 3), and incobotulinumtoxinA (n = 3). There were two studies in which multiple BTX-A formulations were used.^23,39 Three main cosmetic indications for treatment were studied: glabellar lines (GL), crow’s feet lines (CFL), and forehead lines (FHL). Most study designs included treatment of only one indication (n = 25) in isolation. The number and location of injection sites for each indication was essentially identical among all included studies (5 distinct sites for GL, 5 distinct sites for FHL, and 3 distinct sites per side for CFL). Treatment approaches between studies differed mostly commonly in the volume, concentration, and formulation of BTX-A used.

Synthesis of Results.

The results of individual studies (i.e., group sizes and effect estimates) are summarized in the forest plot shown in Fig. 2 along with the results of the primary meta-analysis. The overall relative risk of any TRAEs occurring after BTX-A injection compared to placebo injection was 1.53 (95% CI, 1.33–1.77; 95% PI, 0.95–2.46; p < 0.001). The degree of between-study heterogeneity was minimal and unlikely to be important (I², 36.3%; 95% CI, 2–59%).

FIG. 2.: Forest plot representing the RCTs included in the primary meta-analysis along with the pooled effect estimate. The overall relative risk of TRAEs among patients receiving BTX-A was 1.53 (95% CI, 1.33–1.77; 95% PI, 0.95–2.46; p < 0.001). GL, glabellar lines; CFL, crow's feet lines; FHL, forehead lines.

A meta-regression was performed to identify potential sources of between-study heterogeneity. The year of publication, study location, number of centers, and risk of bias assessments were all found to be nonsignificant. The indication for treatment (i.e., CFL, GL, FHL) was also nonsignificant and interestingly, the number of indications treated in one session did not impact the overall risk of TRAEs (see Figures, Supplemental Digital Content 1–2, available at https://links.lww.com/IOP/A330). This suggests that patients who had multiple areas treated were just as likely to experience a TRAE as those who only had a single area treated.

The duration of follow-up was found to have a statistically significant impact on the primary outcome (p = 0.001). A subgroup analysis based on study duration showed a higher risk of TRAEs among studies lasting 150 days or more (RR 1.65; 95% CI, 1.43–1.91) compared to studies shorter than 150 days (RR 1.28; 95% CI, 1.02–1.60), which is an expected sequela of study duration (see Figure, Supplemental Digital Content 3, available at https://links.lww.com/IOP/A330).

Of the treatment parameters that were considered, the needle gauge, number of injections, total volume injected, number of units per injection, and total number of units injected did not have a statistically significant association with the risk of TRAEs. However, the volume per injection was found to be significant (p = 0.028). The risk of TRAEs after BTX-A injection compared to placebo was higher in studies which used individual injection volumes of 0.1 mL or more (RR 1.77; 95% CI, 1.48–2.11) than in those which used volumes of less than 0.1 mL (RR1.19; 95% CI, 1.03–1.38).

Among the included studies, the mean volume per individual injection was 0.08 mL, with a range of 0.05–0.10 mL of reconstituted BTX-A solution. Figure 3 shows the unadjusted relative risk estimates associated with a single unit increase in each of the treatment parameters. For example, with each 0.05 mL increase in the volume per individual injection, the unadjusted relative risk of TRAEs increased by approximately 37% (RR 1.37; 95% CI, 1.00–1.87; p = 0.047). While the unadjusted risk estimate for total volume was not significant, each 0.50 mL increase in total injection volume increased the risk of TRAEs increased by approximately 8.6% (RR 1.09; 95% CI, 0.91–1.30; p = 0.347). The impact of the number of BTX-A units per injection and the number of BTX-A units per entire treatment session was also nonsignificant (Units per Injection: RR 0.98; 95% CI, 0.93–1.04; p = 0.552; Total Units: RR 1.00, 95% CI 0.99–1.01; p = 0.713), as was the number of injections (RR 1.01; 95% CI 0.97–1.05; p = 0.650). While these results are limited to hypothesis-generating purposes, they represent the relative impact of each treatment parameter on the overall risk of TRAEs. Namely, the relative impact was highest for individual injection volume, followed by total injection volume, number of injections, total number of BTX-A units, and finally the number of BTX-A units per individual injection.

FIG. 3.: Bubble plot with fitted meta-regression line representing the exponentiated linear trend in the relative risk of TRAEs for select covariates pertaining to injection technique. The p values indicate the statistical significance result of the univariate meta-regression for each covariate.

Risk of Bias.

The risk of bias assessment for all included RCTs, based on the Cochrane Risk of Bias Tool, is summarized in Table 2. Most of the included studies were considered to have a low risk of bias for the following domains: selection bias (random sequence generation and allocation concealment), performance bias (blinding of participants and personnel), attrition bias (incomplete outcome data), and reporting bias (selective reporting). However, while most studies were appropriately blinded for treatment and efficacy assessments of their primary outcomes (i.e., efficacy of BTX-A injection), most were considered to have an unclear and potentially high risk of bias for outcome assessment of adverse events given the likelihood that the study blind was broken in this phase of the study.

TABLE 2.: Results of the risk of bias assessment for all included studies, based on the Cochrane Risk of Bias Tool

The risk of publication bias was determined to be low based on the funnel plot shown in Fig. 4. Studies with a small sample error (i.e., larger sample sizes) were distributed with good symmetry across the Funnel Plot, suggesting that there was no publication bias among the larger studies. Conversely, there was a slight amount of asymmetry among studies with a large sample error (i.e., small sample sizes), suggesting the possibility of publication bias due to small-study effects. Subsequent statistical tests for publication bias due to small-study effects were contradictory, with Egger’s test result suggesting the presence of bias (p = 0.001) and Begg’s test suggesting the opposite (p = 0.230). Egger’s test is generally considered to be a more sensitive test and it was therefore postulated that the degree of publication bias due to small-study effects, if present, was minimal.

FIG. 4.: Funnel plot displaying the standard error of the log risk ratio by the log risk ratio of treatment-related adverse events. This plot offers a visual depiction of the risk of publication bias among the studies included in this meta-analysis.

Additional Analyses

BTX-A Formulation.

The type of BTX-A formulation was not found to have a statistically significant impact on the relative risk of TRAEs (p = 0.092.). However, the relative risk of TRAEs determined by the subgroup meta-analysis was 1.61 for onabotulinumtoxinA (95% CI, 1.31–1.98; p < 0.001), 1.21 for abotulinumtoxinA (95% CI, 1.00–1.46; p = 0.049), 1.21 for prabobotulinumtoxinA (95% CI, 0.73–2.00; p = 0.460), 2.16 for daxibotulinumtoxinA (95% CI, 1.54–3.05; p < 0.001), and 2.35 for incobotulinumtoxinA (95% CI, 0.68–8.14; p = 0.177) (see Figure, Supplemental Digital Content 4, available at https://links.lww.com/IOP/A330). These trends suggest that onabotulinumtoxinA has a similar but slightly higher risk of TRAEs compared to abobotulinumtoxinA. The effect estimates for the remaining BTX-A formulations had limited precision based on the 95% confidence intervals. Of note, “conversion ratios” for BTX-A formulations were not accounted for in this meta-analysis due to the lack of consensus on a single ratio for each possible conversion.

Treatment Indication.

The indication of treatment did not have a statistically significant association with the relative risk of TRAEs (p = 0.848). Additionally, there was no statistically significant difference in the risk of TRAEs between patients who were treated for a single indication (e.g., GL alone) and those who were treated for multiple indications (e.g., GL and CFL) simultaneously (p = 0.518). Trends suggested that TRAEs were relatively more likely to occur for GL treatments (RR 1.49; 95% CI, 1.25–1.77; p < 0.001) than for CFL treatments (RR 1.41; 95% CI, 0.79–2.50; p = 244) when compared to placebo (see Figure, Supplemental Digital Content 1, available at https://links.lww.com/IOP/A330).

Treatment Naïve Patients.

History of previous BTX-A treatment was not found to be a statistically significant covariate in the meta-regression (p = 0.889). The relative risk was similar among treatment naïve patients (RR 1.55; 95% CI, 1.26–1.91; p < 0.001) and those with a history of BTX-A treatment (RR 1.58; 95% CI, 1.29–1.94; p < 0.001) (see Figure, Supplemental Digital Content 5, available at https://links.lww.com/IOP/A330).

Specific Adverse Events.

Subgroup analyses were performed to determine the relative risk of specific adverse events of interest (see Figures, Supplemental Digital Content 6–10, available at https://links.lww.com/IOP/A330). Compared to patients who received placebo injections, those who received BTX-A injections were 0.99 times more likely to experience injection site bruising, hematoma, or hemorrhage (95% CI, 0.66–1.49; p = 0.979). The overall relative risk for injection site reactions was 1.18 (95% CI, 0.80–1.73; p = 0.397). For eyelid and/or eyebrow malposition, the relative risk among patients receiving BTX-A was 3.55 (95% CI, 1.84–6.85; p < 0.001). The relative risk was 1.45 for headaches (95% CI, 1.13–1.84; p = 0.003) and 2.42 for facial palsy or paresis (95% CI, 0.43–13.56; p = 0.316). There were no reports of systemic extension of BTX-A manifesting as distal paresis or paralysis among the included studies.

Sensitivity Analyses.

The cumulative meta-analysis presented in Fig. 5 demonstrates increasing precision in the pooled relative risk estimate as more evidence from studies accumulated over time. Thus, there were no studies that reported drastically different effect estimates than others and no important trends over time. This suggests that the pool of studies included in this meta-analysis reported safety data that were consistent and accumulated towards a more precise effect estimate over time.

FIG. 5.: Cumulative meta-analysis, sorted by year. The risk ratio (RR) in each line represents the pooled RR for the study indicated in the same line as well as all preceding studies in chronological order. This analysis represents how the pooled risk ratio changed over time as more studies emerged and, in this case, additional evidence confirmed the existing date over time and increased the precision of the risk ratio estimate. No obvious study outliers are apparent.

The influence analysis presented in Fig. 6 shows that excluding any study from the meta-analysis would produce an overall risk estimate that would remain within the 95% confidence interval of the pooled estimate of all studies combined. This suggests a low likelihood of study outliers and further supports a low risk of bias across the included studies.

FIG. 6.: Influence analysis representing the impact of each study on the overall pooled risk ratio (RR). The RR provided for each study indicates what the overall pooled risk ratio (i.e., for all studies) would be if that specific study was removed from the analysis. No obvious study outliers are apparent. GL, glabellar lines; CFL, crow's feet lines; FHL, forehead lines.

DISCUSSION

The results of this meta-analysis suggest that facial BTX-A injections for cosmetic indications carry a statistically significant albeit fairly low risk of TRAEs. More specifically, adult patients receiving BTX-A injections were 1.53 times more likely to experience a TRAE compared to those who received placebo injections (95% CI, 1.33–1.77; p < 0.001). There was a reliable amount of precision in this estimate based on the sensitivity analyses and the minimal degree of between-study heterogeneity (I², 36.3%; 95% CI, 2–59%).

Several covariates were assessed in this meta-analysis for their respective impact on the risk of TRAEs following BTX-A injections and these hypothesis-generating results may help guide future studies and potentially clinical practice. Most notably, the type of BTX-A formulation used was not found to be significant, as well as the indication for treatment (i.e., CFL, GL, FHL), number of indications treated in one session, and whether patients had a history of previous treatments.

Our meta-regression provided interesting results on the relative impact of variables related to treatment technique. Individual injection volume was found to have a significant association with the overall risk of TRAEs, whereas injection concentration, total BTX-A units, total injection volume, and number of injections were not. These trends suggest that TRAEs might be more likely to result from tissue distension due to solution volume rather than the pharmacodynamics of BTX-A within subcutaneous tissue. Furthermore, these results may provide a relative comparison of the importance of these potentially modifiable injection parameters for future studies and clinical practice. Namely, future studies should explore whether using lower volumes of more concentrated solutions or alternatively, whether a higher number of injections of lower volumes, yields a safer risk profile for patients. Existing publications on this topic have emphasized efficacy over safety outcomes. Hsu et al. compared two solutions containing equivalent doses of BTX-A (5 U in 0.25 mL vs. 5 U in 0.05 mL) in patients with FHL and reported a larger area of effect with the more dilute solution.⁴⁴ Carruthers et al. similarly compared 4 different dilutions of BTX-A in patients with GL and ultimately found no significant differences in efficacy or safety outcomes. Punga et al. assessed two solutions in patients with GL and reported similar times of onset, improvement in wrinkle severity, and duration off effect. However, the data trends in this study tended to show higher efficacy in the twofold volume group efficacy.⁴⁵ Although both studies reported no difference in overall safety between the two groups, the results of our meta-analysis suggest a possible difference based on pooled data from larger studies. Several factors are important to consider in determining the optimal strategy, including the size of the target group being targeted, the impact of the pattern of muscle contraction on BTX-A spread, and the impact of diffusion to nearby structures (e.g., eyelid musculature). Patient factors are also important to consider, such as whether the patient is especially averse to injection punctures. Ultimately, the relationship between various aspects of treatment technique and their impact on overall safety and efficacy is not yet fully understood and our results suggest that further characterization of these variables may add a higher level of sophistication to BTX-A treatment.

Specific adverse events that were more likely to occur among patients treated with BTX-A rather than placebo included eyelid and/or eyebrow malposition, facial palsy and/or paresis, and headaches. These adverse events are consistent with the pharmacodynamics of BTX-A and the nature of treatment (i.e., subcutaneous facial injections). Injection site reactions and injection site bruising, hematoma, or hemorrhage occurred at similar rates in both patient groups and yielded a risk estimate that was not statistically significant. None of the included studies reported any occurrences of distal paresis or paralysis. Collectively, these results provide a generally reassuring risk profile for BTX-A injections.

There were several limitations inherent to this meta-analysis. All the included RCTs were designed and powered for efficacy assessments rather than safety assessments. Consequently, the reporting of safety outcomes was relatively less detailed for many studies, which limited the amount of data that could be contributed to the meta-analysis and in some cases even led to the exclusion of an otherwise well-designed study. Furthermore, the blinding of outcome assessments was maintained only for the assessment of BTX-A efficacy and was likely broken in the context of safety outcomes, thus indicating a high risk of detection bias for several studies. The risk of bias assessment also identified a potential bias due to passive monitoring of adverse events among several studies. Nonetheless, the overall risk of bias within and across the included studies was relatively low beyond these considerations, which are themselves difficult to overcome as high-quality RCTs in the context of BTX-A are unlikely to be designed specifically for safety assessments.

Another limitation was the lack of standardized “conversion ratios” for the different BTX-A formulations. While some conversion ratios have been published in the literature and may be utilized by some clinicians to estimate equivalent BTX-A doses when switching between formulations, the lack of a standardized ratio for each formulation precluded their inclusion as a covariate in this meta-analysis. Thus, while our analysis ultimately showed no significant differences between the different BTX-A formulations, this may have been masked because conversion ratios were not available.

Compared to previously published meta-analyses on this topic, this meta-analysis incorporated nearly double the number of studies despite using relatively stricter selection criteria.^7–11,46 One meta-analysis published in 2004 had a similarly high number of included studies (n = 36), although some of the included studies focused on noncosmetic indications for BTX-A treatment. Nonetheless, the rate of adverse events for the entire dataset in this study was similar to that determined by our meta-analysis.¹¹ The total adverse event rates in the remaining meta-analyses were also statistically significant and similar to those determined in our study,^9,10 with the exception of one study by Guo et al. in which no significant differences were found in the frequency of adverse events.⁸ This may have been the result of their study selection process, which exclusively included studies of patients receiving a specific dose of BTX-A.⁸ It is also unclear whether this conclusion was based on a pooled meta-analysis or the observation that no significant differences were found in each individual study.⁸ Brin et al. assessed the safety of onabotulinumtoxinA and reported a similar risk ratio (i.e., approximately 1.5) to our study for both onabotulinumtoxinA and for all formulations of BTX-A cumulatively.¹⁰

The added value of this meta-analysis stems from the secondary analyses which offer clinically relevant information and provide hypothesis-generating data. Our meta-analysis found no significant differences in the risk of TRAEs among the available BTX-A formulations, but trends suggested that relative risk of TRAEs in decreasing order is as follows: incobotulinumtoxinA, daxibotulinumintoxinA, onabotulinumtoxinA, abobotulinumtoxinA, and prabobotulinumtoxinA. We also found no significant differences in risk based on treatment indication (i.e., GL or CFL), including in patients who had multiple indications treated in one session. Our results also indicate that there was no difference in the risk of TRAEs between treatment naïve patients and those who have received previous BTX-A treatments.

In summary, this meta-analysis determined that the risk of TRAEs following BTX-A injections was statistically significant when compared to that of placebo injections, albeit within an acceptable range that is unlikely to discourage most patients from undergoing this therapy. Our meta-regression and additional analyses provided additional data regarding the relative importance of certain covariates. In addition to these trends from nonsignificant results, certain variables were found to have a statistically significant association with the risk estimate and included study duration, individual injection volume, and total volume injected per treatment session. Lastly, specific adverse events that were more likely to occur following BTX-A injection were mild-to-moderate, transient, and consistent with the expected pharmacodynamics and nature of BTX-A injections. The concern for systemic extension of BTX-A was minimal or nonexistent based on the included studies.