Topical benzoyl peroxide for acne
Acne is a common, economically burdensome condition that can cause psychological harm and, potentially, scarring. Topical benzoyl peroxide (BPO) is a widely used acne treatment; however, its efficacy and safety have not been clearly evaluated.
To assess the effects of BPO for acne.
We searched the following databases up to February 2019: the Cochrane Skin Specialised Register, CENTRAL, MEDLINE, Embase, and LILACS. We also searched five trials registers and checked the reference lists of relevant randomised controlled trials (RCTs) and systematic reviews.
We included RCTs that compared topical BPO used alone (including different formulations and concentrations of BPO) or as part of combination treatment against placebo, no treatment, or other active topical medications for clinically diagnosed acne (used alone or in combination with other topical drugs not containing BPO) on the face or trunk.
Data collection and analysis
We used standard methodological procedures as expected by Cochrane. Primary outcome measures were 'participant global self‐assessment of acne improvement' and 'withdrawal due to adverse events in the whole course of a trial'. 'Percentage of participants experiencing any adverse event in the whole course of a trial' was a key secondary outcome.
We included 120 trials (29,592 participants randomised in 116 trials; in four trials the number of randomised participants was unclear). Ninety‐one studies included males and females. When reported, 72 trials included participants with mild to moderate acne, 26 included participants with severe acne, and the mean age of participants ranged from 18 to 30 years.
Our included trials assessed BPO as monotherapy, as add‐on treatment, or combined with other active treatments, as well as BPO of different concentrations and BPO delivered through different vehicles. Comparators included different concentrations or formulations of BPO, placebo, no treatment, or other active treatments given alone or combined. Treatment duration in 80 trials was longer than eight weeks and was only up to 12 weeks in 108 trials. Industry funded 50 trials; 63 trials did not report funding. We commonly found high or unclear risk of performance, detection, or attrition bias. Trial setting was under‐reported but included hospitals, medical centres/departments, clinics, general practices, and student health centres. We reported on outcomes assessed at the end of treatment, and we classified treatment periods as short‐term (two to four weeks), medium‐term (five to eight weeks), or long‐term (longer than eight weeks).
For 'participant‐reported acne improvement', BPO may be more effective than placebo or no treatment (risk ratio (RR) 1.27, 95% confidence interval (CI) 1.12 to 1.45; 3 RCTs; 2234 participants; treatment for 10 to 12 weeks; low‐certainty evidence). Based on low‐certainty evidence, there may be little to no difference between BPO and adapalene (RR 0.99, 95% CI 0.90 to 1.10; 5 RCTs; 1472 participants; treatment for 11 to 12 weeks) or between BPO and clindamycin (RR 0.95, 95% CI 0.68 to 1.34; 1 RCT; 240 participants; treatment for 10 weeks) (outcome not reported for BPO versus erythromycin or salicylic acid).
For 'withdrawal due to adverse effects', risk of treatment discontinuation may be higher with BPO compared with placebo or no treatment (RR 2.13, 95% CI 1.55 to 2.93; 24 RCTs; 13,744 participants; treatment for 10 to 12 weeks; low‐certainty evidence); the most common causes of withdrawal were erythema, pruritus, and skin burning. Only very low‐certainty evidence was available for the following comparisons: BPO versus adapalene (RR 1.85, 95% CI 0.94 to 3.64; 11 RCTs; 3295 participants; treatment for 11 to 24 weeks; causes of withdrawal not clear), BPO versus clindamycin (RR 1.93, 95% CI 0.90 to 4.11; 8 RCTs; 3330 participants; treatment for 10 to 12 weeks; causes of withdrawal included local hypersensitivity, pruritus, erythema, face oedema, rash, and skin burning), erythromycin (RR 1.00, 95% CI 0.07 to 15.26; 1 RCT; 60 participants; treatment for 8 weeks; withdrawal due to dermatitis), and salicylic acid (no participants had adverse event‐related withdrawal; 1 RCT; 59 participants; treatment for 12 weeks). There may be little to no difference between these groups in terms of withdrawal; however, we are unsure of the results because the evidence is of very low certainty.
For 'proportion of participants experiencing any adverse event', very low‐certainty evidence leaves us uncertain about whether BPO increased adverse events when compared with placebo or no treatment (RR 1.40, 95% CI 1.15 to 1.70; 21 RCTs; 11,028 participants; treatment for 10 to 12 weeks), with adapalene (RR 0.71, 95% CI 0.50 to 1.00; 7 RCTs; 2120 participants; treatment for 11 to 24 weeks), with erythromycin (no participants reported any adverse events; 1 RCT; 89 participants; treatment for 10 weeks), or with salicylic acid (RR 4.77, 95% CI 0.24 to 93.67; 1 RCT; 41 participants; treatment for 6 weeks). Moderate‐certainty evidence shows that the risk of adverse events may be increased for BPO versus clindamycin (RR 1.24, 95% CI 0.97 to 1.58; 6 RCTs; 3018 participants; treatment for 10 to 12 weeks); however, the 95% CI indicates that BPO might make little to no difference. Most reported adverse events were mild to moderate, and local dryness, irritation, dermatitis, erythema, application site pain, and pruritus were the most common.
Current evidence suggests that BPO as monotherapy or add‐on treatment may be more effective than placebo or no treatment for improving acne, and there may be little to no difference between BPO and either adapalene or clindamycin. Our key efficacy evidence is based on participant self‐assessment; trials of BPO versus erythromycin or salicylic acid did not report this outcome.
For adverse effects, the evidence is very uncertain regarding BPO compared with adapalene, erythromycin, or salicylic acid. However, risk of treatment discontinuation may be higher with BPO compared with placebo or no treatment. Withdrawal may be linked to tolerability rather than to safety. Risk of mild to moderate adverse events may be higher with BPO compared with clindamycin.
Further trials should assess the comparative effects of different preparations or concentrations of BPO and combination BPO versus monotherapy. These trials should fully assess and report adverse effects and patient‐reported outcomes measured on a standardised scale.
Zhirong Yang, Yuan Zhang, Elvira Lazic Mosler, Jing Hu, Hang Li, Yanchang Zhang, Jia Liu, Qian Zhang
Plain language summary
Topical benzoyl peroxide for acne
We reviewed the evidence showing effects and safety of topical benzoyl peroxide (BPO), used alone or in combination. Eligible comparisons included placebo (an identical but inactive treatment), no treatment, or other active (medical) topical medications for treating acne (used alone or in combination with other topical drugs not containing BPO) (evidence is current to February 2019).
The main outcomes of interest in this review were participant‐reported acne improvement and withdrawal from the study due to any side effects. More generally, we also considered the percentage of participants experiencing any side effects over the whole course of a trial.
As a common skin disease, acne vulgaris affects the physical, mental, and social well‐being of millions of adolescents and young adults. A wide range of treatments for acne vulgaris are available, and topical BPO has been recommended as a priority therapy to be used alone or combined with other topical or oral treatments, depending on acne severity. However, the benefits and harms of BPO have yet to be evaluated.
We included 120 studies (which comprised 29,592 people randomised in 116 trials; in four trials the number of randomised participants was unclear). Through our search, we found studies assessing different concentrations of BPO, BPO delivered through various means, or BPO used alone or given with other treatments that may or may not be considered as the primary treatment. These studies compared treatments against different concentrations or formulations of BPO, placebo, no treatment, or other medical treatments given alone or in combination.
Most of the studies included male and female participants with mild to moderate acne; only 67% of studies reported participant age, which ranged between 18 and 30 years. Participants were treated for over eight weeks in nearly two‐thirds of the trials. Industry financially supported approximately two‐fifths of the trials, and more than half of the trials did not report their funding sources. Few studies reported where they were set, but locations included hospitals, medical centres, national medical institutes, clinics, medical departments, and general practices.
We found low‐certainty evidence to suggest that long‐term (i.e. given for longer than eight weeks) BPO treatment may increase self‐reported treatment success compared to placebo or no BPO treatment (three studies), but there may be little to no difference when BPO treatment is compared to adapalene (five studies) or clindamycin (one study). This outcome was not reported by studies comparing BPO treatment to erythromycin or salicylic acid.
Long‐term BPO may lead to an increased chance of treatment discontinuation compared to placebo or no treatment (24 studies), with the most common causes being redness, itch, and skin burning (low‐certainty evidence). When medium‐ to long‐term BPO was compared to adapalene (11 studies), clindamycin (eight studies), erythromycin (one study), or salicylic acid (one study), we found only very low‐certainty evidence, meaning that although there may be little to no difference in withdrawal between these groups, we are not sure of the results. It should be noted that participant withdrawal may be linked to issues around treatment acceptability (dermatitis, rash, facial swelling, sensitivity) rather than to safety.
Very low‐certainty evidence means that we are uncertain if BPO leads to more side effects among participants receiving medium‐ to long‐term BPO than among those given no treatment/placebo (21 studies), adapalene (seven studies), erythromycin (one study), or salicylic acid (one study). Medium‐term treatment with BPO may lead to increased risk of side effects when compared against clindamycin, but the effects of this treatment vary, so the treatment chosen may make little to no difference (six studies; moderate‐certainty evidence). Side effects reported in these studies were usually mild to moderate, and the most common were local dryness, irritation, eczema, redness, pain at the site of application, and pruritus.
Certainty of the evidence
For our key comparisons, we rated the certainty of evidence for 'participant‐reported acne improvement' as low. For the outcomes 'withdrawal due to adverse effects' and ‘percentage of participants having any adverse events', the evidence was mainly of very low certainty.
Included trials were at high or unclear risk of bias, participant numbers were small, results were not consistent across trials, and we suspected publication bias.
Zhirong Yang, Yuan Zhang, Elvira Lazic Mosler, Jing Hu, Hang Li, Yanchang Zhang, Jia Liu, Qian Zhang
Implications for practice
Our review's primary efficacy evidence is based on participant self‐assessment. Low‐certainty evidence suggests that BPO (as monotherapy or as add‐on treatment) may lead to better participant self‐reported acne improvement when compared to no treatment or placebo, and low‐certainty evidence suggests there may be little to no difference between BPO compared to clindamycin or adapalene for this outcome. This indicates that BPO could potentially be used as a replacement for clindamycin in combination treatments to reduce the risk of antibiotic resistance. Our primary efficacy outcome was not assessed by studies comparing BPO with erythromycin or salicylic acid.
Regarding safety, low‐certainty evidence suggests that compared with placebo or no treatment, BPO (as monotherapy or as add‐on treatment) may be associated with higher risk of discontinuation of treatment due to adverse events, which may negatively influence treatment adherence. However, these events may be related to tolerability, mostly cutaneous irritation, such as erythema, pruritus, and skin burning.
As we found only very low‐certainty evidence for withdrawal due to adverse effects for the comparisons BPO versus adapalene, clindamycin, erythromycin, or salicylic acid, we cannot draw conclusions regarding these treatment comparisons for that particular outcome.
More broadly, BPO may increase the risk of adverse events when compared against clindamycin, based on moderate‐certainty evidence; although the 95% confidence interval indicates that BPO might make little to no difference. Adverse events tended to be mild to moderate and well tolerated; the most common were local dryness, irritation, dermatitis, erythema, application site pain, and pruritus.
Very low‐certainty evidence regarding the risk of adverse events for BPO compared against adapalene, clindamycin, erythromycin, or salicylic acid means we cannot draw conclusions regarding these treatment comparisons for that particular outcome.
The 43 studies in Studies awaiting classification may alter the conclusions of the review if assessed.
Implications for research
This systematic review highlights the need for large‐scale well‐conducted RCTs to assess the benefits and harms of BPO for treating acne, with focus on the comparative effects of different preparations or concentrations and combination versus monotherapy.
No RCTs included in our review were regarded as being at low risk of bias in all assessment domains. To improve the quality of clinical trials in this field, FDA 2005 has presented guidelines for assessing drugs for treating acne vulgaris. However, our review suggests that further improvement of study quality is needed. Future trials need to ensure that allocation concealment and blinding are properly conducted.
Our review suggests three limitations in outcome assessment. First, it was uncommon to assess patient‐reported efficacy outcomes in the included trials, which could be helpful in informing both participants and clinicians (FDA 2005). Only 31% of trials reported patient self‐assessed acne improvement. These outcomes should be consistently included in future acne trials. Second, standardisation of outcome measurements needs to be improved for acne trials. Likert‐like scales used for patient self‐assessed acne improvement, for example, varied in different studies, which challenges synthesis of data for this outcome. Although the importance of standardisation of outcome measurements for acne trials has been acknowledged (Barratt 2009), substantial improvement in standardisation is still needed to facilitate comparisons between studies and to promote evidence synthesis. The work on standardisation of outcome measurement may significantly benefit from ongoing collaborative initiatives, including the Acne Core Outcomes Research Network (ACORN) (ACORN), the Core Outcome Measures in Effectiveness Trials (COMET) (COMET), and the Cochrane Skin Group Outcomes Research Initiative (Schmitt 2016). Finally, long‐term outcome assessment (at least six months) was rare, with only four included trials following participants for at least 24 weeks (Dreno 2016; Dréno 2018; Iftikhar 2009; Kawashima 2017a; Korkut 2005). Future trials need to strengthen the long‐term evidence base.
Although BPO as monotherapy or used in combination with topical retinoid or antibiotics has been consistently recommended for first‐line treatment of acne by recent clinical guidelines (Goh 2015; Le Cleach 2017; Zaenglein 2016), available evidence supporting this recommendation is not of high quality in terms of participant self‐evaluation, as suggested in our review. BPO is available in various concentrations and vehicles; however, insufficient evidence suggests the comparative efficacy of its different formulations. The quality of evidence regarding safety (most adverse effects were cited as cutaneous intolerability) is mostly low or even very low. These findings in our review highlight the need for high‐quality evidence for comparisons between BPO monotherapy and other active topical treatments, between BPO combination and monotherapy, and between different formulations.
Future trials need to enhance reporting quality by following the CONSORT statement (Schulz 2010). Incomplete reporting, which impairs risk of bias assessment, is very common in acne trials (Ingram 2010). Lack of sufficient reported information leads to substantial uncertainty in the risk of bias for most included trials. Poor reporting also challenges evidence synthesis. For example, a majority of the trials assessed change in acne lesion counts but seldom reported essential statistics for data synthesis and results interpretation, such as standard deviation, 95% confidence interval, and P value for estimates of effect size. Although adverse events were observed in most trials, the number of participants having adverse events in each group was not generally reported. These problems should be properly tackled in future research.