Effectiveness of Whitening Strips Use Compared With Supervised Dental Bleaching: A Systematic Review and Meta-analysis

Nowadays, people have become more concerned about the appearance of their smile¹  and often seek out dental bleaching to improve their self-esteem, social and professional relationships, and quality of life.^2-4  A study conducted in Brazil in 2018 reported that after in-office bleaching treatment, there was a 78% increase in dental esthetic perception by patients.⁵ 

Although dentist-supervised bleaching is used worldwide,^6,7  limited access to dental treatment has prompted the development of over-the-counter tooth-bleaching systems in some countries, mostly known as whitening strips (WS).⁸  Manufacturers claim that these products have good efficacy and lower cost, and they can be easily found online, in markets, and in pharmacies.⁹  However, these unsupervised treatments can be harmful for patients in cases of abuse or use in nonindicated cases, as reported by the American Dental Association.¹⁰  Dental bleaching also has some adverse effects, such as tooth sensitivity (TS), gingival irritation (GI),¹¹  and increased susceptibility to demineralization, as shown by in vitro studies, as well as decreased cellular viability due to the cytotoxic effects of H₂O₂.^6,12,13 

A systematic review published by Serraglio and others¹⁴  in 2016 compared the use of over-the-counter WS vs 10% carbamide peroxide gel applied at-home using a customized tray. They concluded that there was no significant difference in color change or TS, but there was a higher rate of GI for the group treated at-home. However, some articles from previous years are missing^15-20  in this systematic review, and most of the studies included in the meta-analysis had a high risk of bias (RoB), which suggests uncertainty about the estimates provided by the authors.

Therefore, the purpose of the present systematic review of the literature was to perform a systematic review and meta-analysis to evaluate bleaching efficacy, TS, GI, and patient satisfaction with bleaching protocols performed using over-the-counter WS and dentist-supervised bleaching performed at-home or in the dentist's office. To this end, we aimed to answer the following focused research question based on the PICO acronym (Population, Intervention, Comparison, and Outcome): Does the use of over-the-counter WS lead to the same color change compared with dentist-supervised bleaching techniques in patients who undergo dental bleaching?

This study protocol was registered at the International Prospective Register of Systematic Reviews (PROSPERO, CRD42017070562) and followed the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) statement for report.²¹ 

Controlled vocabulary (MeSH terms) and free keywords were combined with the Boolean operator “OR” for the concepts “participant” and “intervention” from the PICO question addressed at the end of the Introduction section. Then, the concepts were combined with the Boolean operator “AND” to find randomized clinical trials (RCTs) that compared over-the-counter WS and dentist-supervised bleaching techniques. The primary outcome evaluated was color change in ΔSGU (shade guide units) and in ΔE* (measured with a spectrophotometer colorimeter, or chromometer). The secondary outcomes were risk and intensity of TS and GI as well as patient satisfaction.

Searches were performed in MEDLINE (via PubMed), Cochrane Library, Brazilian Library in Dentistry (BBO), Latin American and Caribbean Health Sciences Literature (LILACS), as well as in the citation databases Scopus and Web of Science (Table 1) using free keywords and a controlled vocabulary (Medical Subject Heading terms [MESH]). The reference lists and the first page of linked related articles in PubMed for all primary studies were hand searched for additional relevant publications. No restrictions were placed on the publication date or language. The grey literature was investigated by searching abstracts of the annual conference of International Association for Dental Research (IADR) and its regional divisions (2001-2019), the System for Information on Grey Literature in Europe database, and dissertations and theses (ProQuest Dissertations and Thesis full-text database; Periódicos Capes thesis database).

Clinical trial registries were searched to find unpublished and ongoing trials: Current Controlled Trials (www.controlled-trials.com), the international clinical trials registry platform (http://apps.who.int/trialsearch/), ClinicalTrials.gov (www.clinicaltrials.gov), Rebec (www.rebec.gov.br), and the EU Clinical Trials Register (https://www.clinicaltrialsregister.eu).

We included parallel and split mouth RCTs that compared over-the-counter WS and dentist-supervised at-home and in-office bleaching techniques in patients with permanent dentition. All retrieved studies were initially scanned for relevance (according to the eligibility criteria) by title, followed by abstract reading when the title was not clear enough to check if the text met the inclusion criteria. Finally, when the abstract was not clear enough, the full texts were assessed, which were read by two reviewers. The eligible articles received a study identification (ID), combining the first author and year of publication.

Two reviewers (GRVR and BMM) independently extracted data from included articles, such as study design, participants, interventions and outcomes. In cases of disagreement, a decision was reached by consulting a third reviewer (FSN). If there were multiple reports of the same study (ie, reports with different follow-ups), data from all reports were extracted directly into a single data collection form to avoid overlapping data.

Quality assessments of the selected trials were carried out by two independent reviewers, using the Risk of Bias tool from the Cochrane Collaboration.²²  The assessment criteria contained six items: selection bias (adequate sequence generation and allocation concealment), performance bias (blinding of patient and operators), detection bias (blinding of outcome assessors), attrition bias (incomplete outcome data), reporting bias (free of selective reporting), and other possible sources of bias. Disagreements between the reviewers were solved through discussion and, if needed, by consulting a third reviewer FSN).

For each aspect of the quality assessment, the RoB was scored following the recommendations described in the Cochrane Handbook for Systematic Reviews of Interventions 5.2.0 (http://handbook.cochrane.org). Each domain level was judged as low risk, high risk, or unclear RoB. At the study level, a study was considered as low RoB if all key domains for each outcome had a low RoB. If one or more domains were judged as unclear RoB, the study was considered to have an unclear RoB. If at least one domain was considered as high RoB, the study was considered to have a high RoB.

For the outcomes of risk and intensity of TS, GI, patient satisfaction, and color change in ΔE*, adequate sequence generation and allocation concealment were the key domains. For color change in ΔSGU, blinding of outcome assessors was also considered a key domain, because the lack of blinding could affect the color measure when done with SGU. Incomplete outcome data and selective reporting were not considered key domains, as these are not problems for this type of research. Blinding of the patient and operator was not considered a key domain, because they could easily identify the different bleaching protocols.

We collected data on color change within periods ranging from immediately after to 4 weeks after bleaching. This variation was due to the differences in the assessment periods reported in the studies. When the study reported more than one follow-up, we took data from the most immediate one. Regarding TS, GI, and patient satisfaction, the worst mean value of TS reported for the group was collected.

Data were analyzed using Revman 5.3 (Review Manager version 5.3, The Cochrane Collaboration, Copenhagen, Denmark). Independent meta-analysis comparing over-the-counter WS vs at-home bleaching and over-the-counter WS vs in-office bleaching were performed in case of available data. Data from eligible studies were summarized by calculating the risk ratio (RR) along with the 95% confidence interval (CI) for the dichotomous data ratio (risk of TS and GI). For the ΔE*, ΔSGU, intensity of TS, and patient satisfaction, the mean difference (MD) was calculated when studies used the same evaluation instrument or the standardized mean difference (SMD) was employed when at least one study used a different evaluation instrument. For all meta-analyses, the random effects model was chosen to summarize the mean effect size of the primary studies. Heterogeneity was assessed using Cochran's Q test and I²  statistics in accordance with the Cochrane Handbook for Systematic Reviews of Interventions. The Cochran's Q test was calculated as the weighted sum of the squared differences between the individual study effects and the pooled effect across studies, with the weights being those used in the pooling method. The I²  statistics describe the percentage of variation across studies that is due to heterogeneity rather than chance. The heterogeneity was classified as follows: <40% = low, 30% to 60% = moderate, 50% to 90% = substantial, and 75% to 100% = considerable. Overlap in the category percentages was due to the fact that other factors such as the magnitude and direction of effects as well as the strength of the evidence were taken into consideration when evaluating heterogeneity.

When more than half of the studies included in the meta-analysis did not report the standard deviation (SD), the missing data were imputed based on the average of the coefficient of variance of the remaining articles.²³ 

Sensitivity analyses were conducted to investigate if assumptions made during data collection would affect the results and could be the reason for high heterogeneity, whenever detected. Analyses were also performed on studies with a high RoB.

We graded the quality of the evidence for each outcome across studies (body of evidence) using the Grading of Recommendations: Assessment, Development and Evaluation (GRADE; http://www.gradeworkinggroup.org/) system to determine the overall strength of evidence for each outcome.²⁴  The GRADE approach is used to contextualize or justify intervention recommendations with four levels of evidence quality, ranging from high to very low.

For RCTs, the level of evidence is high in the GRADE approach, but it can be downgraded due to five reasons (RoB, imprecision, inconsistency, indirectness of evidence, and publication bias) in one or two levels. Each one of these topics was assessed as “no limitation,” “serious limitations,” or “very serious limitations” to allow for categorization of the quality of the evidence for each outcome into high, moderate, low, and very low. “High quality” suggests that we are very confident that the true effect lies close to the estimate of the effect. On the other extreme, “very low quality” suggests that we have very little confidence in the effect estimate, and the reported estimate can be substantially different from what was measured. The GRADEpro Guideline Development Tool, available online (www.gradepro.org), was used to create a summary of findings table, as suggested in the Cochrane Handbook for Systematic reviews of Interventions 5.2.0 (http://handbook.cochrane.org).

The search strategy was conducted initially on August 10, 2017, and was updated on March 22, 2019. After database screening and duplicate removal, 4604 studies were identified (Figure 1). After title screening, 307 studies remained, and this number was reduced to 35 studies after careful examination of the abstracts. From the 35 articles, 15 studies were excluded for the following reasons: 1) two studies compared dentist-supervised dental bleaching with dentifrice^25,26  and 2) 13 studies compared different over-the-counter WS groups.^27-39  In summary, a total of 20 studies were selected for the qualitative evaluation.

The characteristics of the 20 selected studies are listed in Table 2. The parallel study design was predominantly used in 19 studies,^{9,15,17,19,20,40-53}  and only one study used the split mouth design.¹⁶ 

Eighteen studies had already been published,^{9,16,17,19,40-53}  and two studies were abstracts of the IADR.^15,20 

The Vita Classical shade guide (Vita Zahnfabrik, Bad Säckingen, Germany) was used in 11 studies.* Nine studies used an objective instrument for color assessment (spectrophotometer, colorimeter, or chromometer).^{15,16,40-43,46,50,51}  Photography was used in 11 studies^† (Table 2). Four studies^16,42,43,52  used the Vita Bleachedguide 3D-Master scale (Vita Zahnfabrik, Bad Säckingen, Germany).

The intensity of self-reported TS was evaluated using a 0-10 or 0-100 visual analogue scale (VAS) in four studies,^9,42,43,50  while two studies used a 0-4⁴³ or 0-7⁴⁰  numeric rating scale (NRS). The risk of TS was evaluated in 15 studies.^{16,17,19,41-49,51-53} 

The intensity of self-reported GI was evaluated in two studies using a 0-10 VAS scale⁹  or 0-7 NRS scale.⁴⁰  The risk of GI was evaluated in 15 studies.^{16,17,19,41-52} 

Patient satisfaction was evaluated in five studies.^{9,16,40,42,48} 

The mean age of all studies that reported this information was approximately 31.5 years. Five studies did not report this information^{15,16,20,42,53}  (Table 2). In 10 studies that reported the gender of the sample, females were more prevalent.^{19,40,41,44,45,47,49-52} 

The product employed was hydrogen peroxide, with a concentration varying from 2.9% to 14%. In each clinical session, the product remained in contact with the dental structure from 30 to 60 minutes. The WS were used from 7 to 28 days (Table 2).

The products carbamide peroxide (10% to 35%) or hydrogen peroxide (7.5% to 10%) were employed. The bleaching trays were used from 7 to 28 days, with a daily use time that varied from 30 minutes to 10 hours (Table 2).

For in-office bleaching, hydrogen peroxide gels with concentrations that varied from 15% to 38% were employed. In each clinical session, the product remained in contact with the dental structure from 15 to 45 minutes, and one to three clinical sessions were performed.

The RoB of the eligible studies is presented in Figure 2. Ten studies were classified as high RoB in the domain sequence generation,^{17,19,41,44-49,51}  and one study⁴⁰  was classified at high RoB in blinding of outcome assessor during the evaluation of color in SGU. These studies were not used in the meta-analysis. The study by Aka and Celik⁴⁰  was employed in all meta-analyses, except for the outcome color change in ΔSGU.

Seven studies^{17,44,45,47-49,51}  were classified as having a high RoB in the domain of other possible sources due to possible conflicts of interest in the study.

All meta-analyses were performed on studies classified as at low or unclear RoB in the key domains and from which information about the outcome was reported and could be extracted. In this phase, the studies by Brito and others¹⁵  and Rodrigues and others²⁰  were removed, because the data could not be extracted as they were IADR abstracts.

No study comparing WS vs IO bleaching remained at this phase. At the end, only seven studies that compared WS vs at-home bleaching were included in the meta-analysis of primary and secondary outcomes.^{9,16,40,42,43,50,53} 

As these studies performed different comparisons, we meta-analyzed the data in two subgroup analyses, based on the concentration of active hydrogen peroxide used in the dentist-supervised at-home group. Studies that only used 10% carbamide peroxide (3.6% active hydrogen peroxide) in the dentist-supervised group were analyzed separately^9,40,42,50  from those that used a higher active hydrogen peroxide concentration (approximately 10%) in the dentist-supervised group.^{16,42,43,52,53} 

Both subgroups showed that the dentist-supervised at-home group had an increase in color change (p=0.0007). The SMD was −0.50 (95% CI −0.79 to −0.21]. We did not detect heterogeneity in the data (p=0.61; I²=0%; Figure 3).

The studies by Auschill and others,⁹  Carlos and others,⁴²  and Kim and others⁵⁰  reported the final SGU color, whereas the other two studies^43,53  reported the change from baseline. The comparison of the final measurements in an RCT, in theory, estimates the same quantity as the comparison of changes from baseline; in these cases, the outcome can be summarized in the meta-analysis only as the MD. For this reason, two studies^16,52  had to be removed from the meta-analysis, as the authors employed a different shade guide unit for color assessment (Bleachedguide), and its inclusion could only be done using the SMD (not possible in this case, as the present meta-analysis mixed final SGU measurements with change from baseline values). Thus, no significant difference was observed between the subgroups of low and high peroxide concentration (p=0.32). The MD was −0.39 (95% CI −1.16 to 0.37). We detected high heterogeneity in the overall data (p=0.0002; I²=79%), caused by the studies that used a low hydrogen peroxide concentration (p<0.0001; I²=91%; Figure 4).

The RR was 0.78 (95% CI 0.65 to 0.93), showing significant differences between the groups (p=0.006), favoring the over-the-counter WS group. We did not detect any heterogeneity in the data (p=0.98; I²=0%; Figure 5).

The overall SMD for the intensity of TS was −0.30 (95% CI −0.56 to −0.04) and was statistically significant (p=0.02) in favor of the over-the-counter WS group. We did not detect heterogeneity in the overall data (p=0.62; I²=0%). The subgroup analysis revealed no significant differences between the groups (p=0.85 and p=0.13) with the low and high hydrogen peroxide concentrations, respectively (Figure 6).

The RR was 0.90 (95% CI 0.40 to 2.06), showing no significant differences between the groups (p=0.81). We detected high heterogeneity in the data (p=0.03; I²=71%; Figure 7).

Only two studies could be included in the subgroup^9,40  with the low peroxide concentration. The subgroup analysis revealed no significant difference between the groups (p=0.20). The SMD for the intensity of GI was 0.27 (95% CI −0.15 to 0.70; Figure 8).

Carlos and others⁴²  did not report the SD of patient satisfaction, so the SD was imputed. The SMD was −0.32 (95% CI −1.00 to 0.37), showing no significant differences between the groups (p=0.37). We detected high heterogeneity in the overall data (p=0.007; I²=75%), caused by the studies that used a low hydrogen peroxide concentration (p = 0.004; I²=82%; Figure 9).

A meta-analysis was also performed, including studies classified at high RoB, and the results were not different (data not shown), except for the risk of TS, which in this case showed no difference between the groups.

We analyzed whether the concentration of hydrogen peroxide was responsible for the heterogeneity in the meta-analysis of color change in ΔSGU. The findings of the single study⁵⁰  that employed a low concentration of hydrogen peroxide in the over-the-counter WS group generated the heterogeneity in the aforementioned meta-analysis. Also, through a sensitivity analysis, we analyzed whether gel viscosity could cause heterogeneity in the meta-analysis of gingival irritation. The single study⁴²  that employed a high-viscosity gel generated the heterogeneity mentioned earlier. We also analyzed the scale used for the evaluation of patient satisfaction; a single study⁴⁰  employed a different scale (NRS 0-7) for this evaluation and was responsible for the heterogeneity in the meta-analysis.

The body of evidence regarding color change in ΔSGU, risk of GI, and patient satisfaction was graded as very low; for this outcome, we observed that most of the RCTs were at unclear RoB, and we observed data inconsistency because of the high heterogeneity. The quality of the evidence of ΔE*, risk and intensity of TS, and intensity of GI was graded as moderate because of the unclear RoB of most articles (Table 3).

Unfortunately, many studies of this systematic review did not perform adequate sequence generation^{17,19,41,44-49,51}  and thus were classified as having a high RoB in this key domain. They performed random sequence by balancing patients by age, color of the teeth, or other characteristics of the patients in a nonrandom process. These articles were not included in the meta-analysis because of the lower reliability of their data.^54,55  An adequate stratification process is very important in clinical trials, because it is used to balance differences between the control and treated groups and seeks to prevent selection bias as the result of subconscious actions.⁵⁶  On the other hand, if the researchers responsible for patient recruitment have prior knowledge of the randomization of the groups, they can choose participants with a better prognosis for the experimental group and those with a worse prognosis for the control group, or vice versa.^57,58 

An adequate sequence-generation process is essential for the randomization process. The authors of the most eligible studies did not reveal how the allocation concealment was performed. When inappropriately performed, larger estimates of treatment effects are observed compared with studies that had an adequate allocation concealment.^58,59 

Another important domain to reduce the RoB is the blinding of outcome assessors; because, on the contrary, the evaluator may tend to deviate from the truth because of predispositions or expectations.⁶⁰  Examiner blinding is much more important for subjective outcomes, which involve a personal judgment, for example, the assessment of color change in ΔSGU. Examiner blinding may have little practical importance when evaluators use objective tools to collect the results, such as color change in ΔE*, or for patient-centered outcomes, such as risk and intensity of TS or GI. For these latter outcomes, patient blinding is much more essential. In this way, adequate sequence generation, allocation concealment, and blinding of outcome assessors have been proposed as the most important methodological components of controlled trials.^59,61 

We also performed a meta-analysis with all eligible studies regardless of their RoB. We observed that for all cases (except the risk of TS, which showed no difference between the groups), no differences were observed when comparing the meta-analysis of studies with a low and unclear RoB.

This study revealed a statistically significant difference in color change for ΔE* measurements in favor of dentist-supervised bleaching. The quality of the evidence for this outcome was graded as moderate, downgraded only by the fact that most RCTs had an unclear RoB. However, no significant difference in color change by ΔSGU was observed. Although the conclusions of the meta-analysis on color change in ΔE* and ΔSGU may seem contradictory, we perform a careful analysis of the SMD of −0.5 obtained for ΔE*.

Reexpressing the SMD as MD (by multiplying the SD of the control groups by the pooled SMD) resulted in an MD in ΔE* of 1 to 1.5 (taking the SD of control groups from eligible studies). If two objects are placed side by side in a controlled environment, the smallest difference in color detected by human observers is a ΔE* value of 1.⁶²  However, under clinical conditions, a ΔE* of 3.3 has been shown to be the upper limit for human eyes to detect color differences.⁶³ 

Color matching with a shade guide scale depends on the subjective perspective of the examiner. Although this is a valid method,^64-68  with good reliability to differentiate between light and dark colors, the classical Vita shade guide is not specifically designed for dental bleaching assessments.^69,70  It is subjective to changes in the illumination of the clinical offices, weather, time of day, and season. In addition, other external factors such as observer conditions also play a role in the accuracy of color perception. However, when using devices such as chromometers or spectrophotometers, objective and precise measurements are obtained, and they are less sensitive to these sources of variation.⁷⁰  A small difference indeed exists between over-the-counter products and dentist-supervised at-home bleaching, but the pooled magnitude of color change (MD [ΔE*] = 1 to 1.5) detected in the meta-analysis can be clinically detected only by highly calibrated eyes in a controlled environment. Under clinical conditions, over-the-counter products can provide an almost equivalent color change to the dentist-supervised bleaching technique.

Regarding TS, lower sensitivity levels were observed for the over-the-counter WS group. The amount and viscosity of the product in contact with the dental substrate may account for this difference.⁷¹  Some studies^72,73  have reported that over-the-counter WS have approximately 15% of the amount of the gel placed in customized bleaching trays. This proportion may vary depending on the dentist prescription and presence or not of reservoirs in the customized bleaching trays. The higher amount of the gel in the bleaching tray leads to a higher amount of hydrogen peroxide that reaches the pulp, causing an inflammatory reaction^74,75  and consequently increasing the risk of TS.^72,76 

Although we expected that a higher risk of GI would occur in the over-the-counter WS group because of the lack of supervision and consequent extravasation of gel into the gingival tissue,⁸  no differences were detected between the groups. Under the conditions of RCTs, patients are automatically supervised by the dentist, even when using over-the-counter products, in contrast to what happens in a clinical scenario; this may explain the similarity between the GI results.

Another important aspect is patient satisfaction after treatment, which is the main objective of dental whitening. If there is a clinically significant color change, the patient will probably be satisfied.³  In this systematic review, there were no differences between groups.

The studies included in an earlier systematic review of the literature that addressed the same research question¹⁴  are different from those included in the present systematic review. This earlier study included only studies that used 10% carbamide peroxide gel in the dentist-supervised at-home group, irrespective of their RoB. In addition, other recent studies have been published in the meantime that were included in the current study^{16,40,42,43,50,53}  but not in the earlier systematic review. These factors together explain the differences between the conclusions of the present review and the earlier one.¹⁴ 

At the moment, the Scientific Committee on Cosmetic Products and Non-Food Products recommends that tooth-whitening products should contain between 0.1% and 6.0% hydrogen peroxide, as this concentration is safe and proper if supervised by a dentist.⁷⁷  Therefore, special care should be taken for individuals with periodontal diseases, defective restorations, many fillings, crowns, and extremely dark stains. Also, conditions such as preexisting tissue injury or concurrent use of tobacco and/or alcohol may exacerbate the toxic effects of hydrogen peroxide.⁷⁸ 

Finally, more studies with rigorous methodology are needed to increase the reliability of the published data.

Although TS was lower for the WS, dentist-supervised at-home bleaching led to a greater color change when evaluated with a spectrophotometer. However, the color alteration was undetectable by unaided human eyes.

The authors of this article certify that they have no proprietary, financial, or other personal interest of any nature or kind in any product, service, and/or company that is presented in this article.

This study was partially supported by the National Council for Scientific and Technological Development (CNPq) under grants 303332/2017-4 and 308286/2019-7.