A systematic review was conducted to inform the clinical effectiveness and safety of second-line treatments that would be available for prescription by dermatologists and minor surgical treatments. The protocol included all treatments for hyperhidrosis prescribed in secondary care. However, screening and selecting the relevant literature revealed that ETS, although used as part of the treatment pathway for hyperhidrosis, could not be included in a comparative review as the position of ETS in the treatment pathway is uncontestable (ETS is considered only as an intervention of last resort because of its significant risks). Furthermore, the scoping searches identified that the rapid review of ETS for the NICE interventional procedures guidance 487 captured the relevant evidence on ETS.19 They also revealed that because ETS is an established therapy, recent studies of ETS have focused on the details of the surgical procedure, addressing a question that is beyond the remit of the current project. This was included as a protocol amendment in our progress report.
The systematic review was conducted according to the general principles recommended in Centre for Reviews and Dissemination (CRD)’s guidance on the conduct of systematic reviews30 and is reported according to the general principles of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement.31 The protocol was written in accordance with the new Preferred Reporting Items for Systematic Review and Meta-Analysis – Protocols initiative32 and registered on PROSPERO, the international database of prospectively registered systematic reviews in health and social care (URL: www.crd.york.ac.uk/prospero/), as PROSPERO CRD42015027803.
Literature searches
To identify studies of effectiveness, an exhaustive systematic search of electronic databases was undertaken in January 2016 using the following databases: Allied and Complementary Medicine Database (AMED), British Nursing Index, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effects (DARE), EMBASE, Health Technology Assessment (HTA) database, MEDLINE, NHS Economic Evaluation Database (NHS EED), PsycINFO and PubMed.
The search strategy combined relevant search terms with indexed keywords (such as medical subject headings) and text terms that appear in the titles and/or abstracts of database records. Search strategies included appropriate search terms for ‘hyperhidrosis’ combined with search terms for treatment types (e.g. ‘botulinum toxin’, ‘iontophoresis’, ‘curettage’).
No date or language limits were applied.
Additional searches of EMBASE, MEDLINE and NHS EED were carried out to identify studies of cost-effectiveness. A recognised ‘costs’ search filter was used in conjunction with topic terms when the searches of the EMBASE and MEDLINE databases were undertaken.
Additional searches of AMED, British Nursing Index, CINAHL, CENTRAL, CDSR, DARE, EMBASE, HTA database, MEDLINE, NHS EED, PsycINFO and PubMed were carried out to identify quality-of-life studies. The search strategies used combined topic terms for hyperhidrosis with a recognised search filter for ‘quality of life’.
The search strategies are presented as Appendix 1.
Clinical advisors were consulted for additional potentially relevant studies; Julie Halford (nurse specialist and patient representative) provided an extensive bibliography of publications and articles relating to hyperhidrosis. Reference lists of relevant reviews were manually searched.
In addition, information on studies in progress, unpublished research or research reported in the grey literature was sought by searching relevant resources in July 2016 [including Conference Proceedings Citation Index: Science (ISI), ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform Portal].
Two researchers undertook the screening of titles and abstracts obtained through the search, although the library was split between the researchers, rather than each record being double screened. A sample of just over 10% of records was double screened in order to assess the level of agreement between the researchers; it was planned to undertake full double screening if the level of agreement was poor, but this was not necessary. Any record for which a researcher was unsure of their decision was marked and screened by an additional reviewer, in addition to the 10% of records double screened for quality assurance purposes.
Full manuscripts of potentially relevant studies were obtained and independently screened by two researchers, using pre-defined eligibility criteria. Disagreements were resolved through discussion and, when necessary, consultation with a third researcher. Relevant foreign language studies were translated and included in the reviews.
A separate library was created for the records identified by the searches of conference proceedings and trial registers (in July 2016) and all trials identified were deduplicated against the original library. The records were then assessed for relevance by two researchers and any selected for full-text screening were manually compared with the included studies from the original library to identify those that offered additional data. The extra studies were then separated by trial status as ongoing or completed. An online search of pharmaceutical websites and trial registration databases was performed for the completed studies to identify any published reports or trial data. When contact details were available, an e-mail was sent to the principal investigator to request research findings or published reports.
Inclusion and exclusion criteria
Population
Patients with primary hyperhidrosis (including adults and children). Patients with hyperhidrosis secondary to other conditions, such as overactive thyroid or spinal cord injury, or social anxiety disorder, were not eligible for inclusion.
Interventions
Treatments for hyperhidrosis that would be available for prescription by dermatologists and minor surgical treatments for hyperhidrosis.
Comparators
A different active treatment for hyperhidrosis, placebo or no treatment.
Outcomes
Any of the following:
disease severity (e.g. measured with the HDSS, or patient reported)
sweating (e.g. measured by gravimetry, or iodine starch test)
patient quality of life [e.g. assessed using Hyperhidrosis Impact Questionnaire (HHIQ), HDSS (which may be used to measure patient quality of life as well as disease severity) or the Dermatology Life Quality Index (DLQI)]
patient preference
patient satisfaction
patient compliance/adherence to treatment
social functioning (e.g. measured by Social Functioning Questionnaire)
adverse events [such as compensatory sweating (CS)]
resource use.
In addition, the duration of treatment effect was also assessed when adequate data were available.
Study designs
For each intervention, we planned to include good-quality, up-to-date, directly relevant systematic reviews, if they were available. In the absence of such a review, RCTs were included, when available. For interventions for which RCT evidence was lacking, non-RCTs were included. In the absence of controlled trials, large (> 100 patients) prospective case series (single-arm trials) were included.
Data extraction strategy
Data were extracted directly into a standard spreadsheet, which was initially piloted on a sample of studies and refined. Data extraction was conducted by one researcher and checked by a second researcher for accuracy, with any discrepancies resolved by discussion, or consultation with a third researcher, when necessary. Authors of studies were contacted for clarification and missing data, as necessary. In cases of multiple publications of the same study, the publication with the largest sample or longest follow-up was treated as the main source. Data extracted included details of study methods (including study design, country and year of publication), patient characteristics (including age, sex, body treatment site, previous treatments and baseline severity), interventions (including treatment type, dose, frequency and duration), relevant outcome measures (including outcome domain, measurement tool used and follow-up time points) and results. When possible, we sought to include intention-to-treat data. When intention-to-treat data were not available, we extracted and analysed the data as reported in the paper. When results data were missing or limited (e.g. conference abstracts), authors were contacted and, when relevant, manufacturer trials registers were consulted for further data. Where outcome data were presented only in graphical format, authors were contacted to provide further information. If the authors did not respond, data from graphs were extracted using Graph Grabber (version 1.5; Quintessa, Henley-on-Thames, UK) software. All data extraction was performed with Microsoft Excel® (Microsoft Corporation, Redmond, WA, USA).
Quality assessment strategy
The quality assessment of studies was conducted as part of the data extraction process using criteria relevant to the study designs included. The quality of RCTs and non-RCTs was assessed using the Cochrane risk of bias tool, which focuses on the domains shown to have an impact on the trial results in particular (selection, performance and detection biases and attrition).33 An additional question relating to the similarity of treatment groups at baseline was added.34 In addition, a question about ‘within-patient’ study designs was added, owing to concerns about the validity of certain outcome measures in ‘within-patient’ study designs, in which patients receive different interventions on different sides of the body (i.e. the left axilla vs. the right axilla).
Studies without a control group were not formally quality assessed; however, study details are presented and their impact on the reliability of the results is discussed in the relevant sections of this report.
Each controlled trial was given an overall risk of bias judgement: trials that were rated as having a low risk of bias for all key domains (i.e. have a ‘yes’ response for each key domain) were judged to have a low overall risk of bias; trials that were rated as having a high risk of bias for one or more key domains (i.e. have a ‘no’ response) were judged to have a high overall risk of bias; and trials that were rated as having an unclear risk of bias (and no high risk of bias) for one or more key domains were judged to have an unclear overall risk of bias. When relevant, the overall risk of bias judgement was made separately for different outcomes, as certain outcomes are subjective (e.g. the HDSS), whereas other outcomes can be measured more objectively (e.g. gravimetry).
Sequence generation and allocation concealment were both considered to be ‘key domains’; therefore, non-randomised trials were all considered to have a high overall risk of bias. When prognostic factors were considered to differ between treatment groups at baseline, this was also considered to be a key source of bias. When prognostic factors were not reported, an ‘unclear’ response was not considered to be a key source of bias. It is not always possible to blind care providers to different hyperhidrosis interventions; therefore, blinding of care providers was not considered to be a ‘key domain’, nor was blinding of outcome assessors. Blinding of patients was considered to be a ‘key domain’; this is particularly relevant for subjective outcomes, which were assessed in many of the studies included in this review. Although it is acknowledged that it is not possible to blind patients in studies comparing minor surgical interventions with non-surgical interventions, or minor surgical interventions with more major surgery, blinding of participants was still considered a ‘key domain’ in these studies. Whether or not missing outcome data were balanced across groups or adjusted for was also considered to be a ‘key domain’. Whether or not the report appeared to be free from selective outcome reporting was not considered to be a key source of bias; the study protocol was unavailable for the majority of studies.
The final quality assessment question was about ‘within-patient’ study designs, in which patients received different interventions on different sides of the body. Clinical advisors to the project considered that this type of study was of limited use in hyperhidrosis research, particularly for outcomes, such as the HDSS, for which patients are asked to judge the tolerability and impact of their hyperhidrosis on different parts of their body separately. Therefore, ‘within-patient’ study designs were considered to be a key source of bias when the HDSS was used to assess disease severity or for assessments of quality of life. However, ‘within-patient’ study designs were not considered to be a key source of bias for more objective outcomes, such as gravimetry.
The quality of the body of evidence identified was classed according to a modified version of the Grading of Recommendations Assessment, Development and Evaluation (GRADE) classification (very low, low, moderate, high or insufficient) for each intervention and comparison, taking into account the following criteria: study risk of bias, directness of evidence, heterogeneity, precision of effect estimates, risk of publication bias and magnitude of effect.
Data analysis
The results for each of the different treatment types are presented in separate sections within Chapter 3, Results of studies included in the review. Study characteristics and results are presented in a series of structured tables in Appendix 2. Quality assessment results are discussed and tabulated in Chapter 3, Quality of studies included in the review. The results were interpreted in the context of the quality of the individual studies.
Results were pooled in a pairwise meta-analysis if at least two studies reported the same outcome and were considered sufficiently similar for analysis to be appropriate and feasible. Otherwise, the results were summarised in a narrative synthesis. Although a network meta-analysis (NMA) was required to conduct the VOI analyses (see Chapter 6), the evidence was considered too heterogeneous and limited to perform a NMA to address the clinical review questions.
When meta-analyses were performed, dichotomous outcomes were combined to estimate pooled risk ratios (RRs) using standard random-effects DerSimonian–Laird meta-analyses35 and continuous outcomes were combined to estimate pooled mean differences (MDs) using standard random-effects inverse variance meta-analyses. Heterogeneity was assessed using the I2-statistic and visual inspection of forest plots.
For studies that included two intervention groups with two different doses and used one control group, data from each intervention group were entered separately to explore any dose response effect and the number of participants in the control group was divided by two to reduce the risk of double-counting data.36 Although this approach may artificially reduce the power of the study in the meta-analysis and does not account for potential correlation between the two active treatment groups, a separate analysis combining the two arms showed no significant difference in results.
Studies using different units of analysis (i.e. axilla in half-side comparisons vs. patients in between-patient comparisons) were pooled where deemed appropriate and reported in separate subgroups. Meta-regressions and other subgroup analyses were considered inappropriate due to the small number of studies. All analyses were conducted using Review Manager 5.3 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark).
