Sublingual immunotherapy for asthma
Abstract
Background
Asthma is a common long‐term respiratory disease affecting approximately 300 million people worldwide. Approximately half of people with asthma have an important allergic component to their disease, which may provide an opportunity for targeted treatment. Sublingual immunotherapy (SLIT) aims to reduce asthma symptoms by delivering increasing doses of an allergen (e.g. house dust mite, pollen extract) under the tongue to induce immune tolerance. However, it is not clear whether the sublingual delivery route is safe and effective in asthma.
Objectives
To assess the efficacy and safety of sublingual immunotherapy compared with placebo or standard care for adults and children with asthma.
Search methods
We identified trials from the Cochrane Airways Group Specialised Register (CAGR), ClinicalTrials.gov (www.ClinicalTrials.gov), the World Health Organization (WHO) trials portal (www.who.int/ictrp/en/) and reference lists of all primary studies and review articles. The search is up to date as of 25 March 2015.
Selection criteria
We included parallel randomised controlled trials (RCTs), irrespective of blinding or duration, that evaluated sublingual immunotherapy versus placebo or as an add‐on to standard asthma management. We included both adults and children with asthma of any severity and with any allergen‐sensitisation pattern. We included studies that recruited participants with asthma, rhinitis, or both, providing at least 80% of trial participants had a diagnosis of asthma.
Data collection and analysis
Two review authors independently screened the search results for included trials, extracted numerical data and assessed risk of bias, all of which were cross‐checked for accuracy. We resolved disagreements by discussion.
We analysed dichotomous data as odds ratios (ORs) or risk differences (RDs) using study participants as the unit of analysis; we analysed continuous data as mean differences (MDs) or standardised mean differences (SMDs) using random‐effects models. We rated all outcomes using GRADE (Grades of Recommendation, Assessment, Development and Evaluation) and presented results in the 'Summary of findings' table.
Main results
Fifty‐two studies met our inclusion criteria, randomly assigning 5077 participants to comparisons of interest. Most studies were double‐blind and placebo‐controlled, but studies varied in duration from one day to three years. Most participants had mild or intermittent asthma, often with co‐morbid allergic rhinitis. Eighteen studies recruited only adults, 25 recruited only children and several recruited both or did not specify (n = 9).
With the exception of adverse events, reporting of outcomes of interest to this review was infrequent, and selective reporting may have had a serious effect on the completeness of the evidence. Allocation procedures generally were not well described, about a quarter of the studies were at high risk of bias for performance or detection bias or both and participant attrition was high or unknown in around half of the studies.
One short study reported exacerbations requiring a hospital visit and observed no adverse events. Five studies reported quality of life, but the data were not suitable for meta‐analysis. Serious adverse events were infrequent, and analysis using risk differences suggests that no more than 1 in 100 are likely to suffer a serious adverse event as a result of treatment with SLIT (RD 0.0012, 95% confidence interval (CI) ‐0.0077 to 0.0102; participants = 2560; studies = 22; moderate‐quality evidence).
Within secondary outcomes, wide but varied reporting of largely unvalidated asthma symptom and medication scores precluded meaningful meta‐analysis; a general trend suggested SLIT benefit over placebo, but variation in scales meant that results were difficult to interpret.
Changes in inhaled corticosteroid use in micrograms per day (MD 35.10 mcg/d, 95% CI ‐50.21 to 120.42; low‐quality evidence), exacerbations requiring oral steroids (studies = 2; no events) and bronchial provocation (SMD 0.69, 95% CI ‐0.04 to 1.43; very low‐quality evidence) were not often reported. This led to many imprecise estimates with wide confidence intervals that included the possibility of both benefit and harm from SLIT.
More people taking SLIT had adverse events of any kind compared with control (OR 1.70, 95% CI 1.21 to 2.38; low‐quality evidence; participants = 1755; studies = 19), but events were usually reported to be transient and mild.
Lack of data prevented most of the planned subgroup and sensitivity analyses.
Authors' conclusions
Lack of data for important outcomes such as exacerbations and quality of life and use of different unvalidated symptom and medication scores have limited our ability to draw a clinically useful conclusion. Further research using validated scales and important outcomes for patients and decision makers is needed so that SLIT can be properly assessed as clinical treatment for asthma. Very few serious adverse events have been reported, but most studies have included patients with intermittent or mild asthma, so we cannot comment on the safety of SLIT for those with moderate or severe asthma. SLIT is associated with increased risk of all adverse events.
PICOs
Plain language summary
Sublingual immunotherapy for asthma
Review question
We assessed the evidence on the use of sublingual immunotherapy (SLIT) for people with asthma compared with placebo or with normal treatment for asthma. We focused on whether SLIT is a good treatment for asthma and whether it is safe.
Background
Asthma is a long‐term condition that causes breathing problems and cough, which sometimes develop into asthma attacks. This may lead to the need for patients to take extra medication, visit a clinic or a hospital for treatment or even be admitted to the hospital. Approximately 300 million people worldwide have asthma, and allergies may be an important trigger of asthma symptoms in about half of these people (e.g. house dust mites, pollen). The aim of SLIT is to reduce the body's allergic response that causes asthma symptoms; this is done by giving repeated doses of what the person is allergic to in liquid or tablet form under the tongue. Currently, it is not clear whether SLIT is more helpful or safer for people with asthma, when compared with placebo or just continuation of normal asthma treatments.
Study characteristics
We included 52 studies involving 5077 people. These studies lasted between one day and three years. Most of the people included in the studies had mild asthma. Both males and females were included, and about half of the studies included only children.
Most studies involved people with house dust mites or pollen allergy. The evidence presented here is current to 25 March 2015.
Key results
Very few studies recorded the number of people who had asthma attacks leading to a hospital visit or the need for additional medication, so we do not know if SLIT reduces asthma attacks, possibly because most of the patients included in these studies had mild asthma. A few studies reported quality of life, but they used different scales, so we could not really tell if SLIT had a positive effect. Some studies reported that people taking SLIT had fewer asthma symptoms and had a reduced need for asthma medication compared with controls, but studies measured this in different ways, some of which may not be accurate.
People receiving SLIT were no more or less likely to experience serious unwanted side effects, but these were generally very rare. We are not confident that this finding would apply to people with more severe asthma. People receiving SLIT were more likely to experience any unwanted side effect, but many of these were mild.
Guidelines for asthma treatment suggest that SLIT should be used only for people with asthma that is difficult to control with standard treatments. However, many of the studies in this review included people with mild asthma, so trials looking at the effects of SLIT for people with more severe asthma are needed. It would be helpful if these studies used standard scales to report their findings, so that results can be combined in the future.
Quality of the evidence
The evidence presented in this review is generally of moderate or low quality, and very few studies have reported outcomes that are important to people with asthma, such as asthma attacks and quality of life. Most studies did not clearly explain how investigators decided which people would receive SLIT and which individuals would receive placebo or normal care, and in some studies, both participants and trial organisers knew which treatment they were getting. This may have affected the results.
Authors' conclusions
Summary of findings
| Subgroup and sensitivity analyses compared with placebo for asthma | ||||||
| Patient or population: adults and children with asthma Weight mean duration of all included studies: 54 weeks (Fadel 2010 and Rodriguez 2012 excluded, as duration not reported) | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Control | SLIT | |||||
| Exacerbation requiring ED or hospital visit Study duration: | No events | No events | Not estimable | 47 | ⊕⊕⊝⊝ | |
| Quality of life | No meta‐analysis possible | Not applicable | ‐ | (0 RCTs) | Not applicable | 5 studies reported quality of life outcomes but we were not able to perform a meta‐analysis |
| Serious adverse events Weighted mean duration of studies: 49 weeks | 14 per 1000 | 12 per 1000 (0 to 24) | RD 0.0012, (‐0.0077 to 0.0102) | 2560 (22 RCTs) | ⊕⊕⊕⊝ | |
| Exacerbation requiring OCS Weighted mean duration of studies: 25 weeks | No events | No events | Not estimable | 77 | ⊕⊕⊝⊝ | |
| All adverse events Weighted mean duration of studies: 60 weeks** | 222 per 1000 | 327 per 1000 | OR 1.70 (1.21 to 2.38) | 1755 | ⊕⊕⊝⊝ | |
| Bronchial provocation | Mean bronchial provocation in control group was 1020 mcg (PD20) and 5.45 mg/mL (PC20) | Mean bronchial provocation in intervention group was 0.69 standard deviations higher (0.04 lower to 1.43 higher) | ‐ | 139 | ⊕⊝⊝⊝ | 3 studies reported outcome as PC20 and 1 as PD20. We combined the different scales using standardised mean differences |
| ICS use | Mean ICS use in control group was 255 mcgl | Mean ICS use in intervention group was 35.1 higher (‐50.21 to 120.42) | ‐ | 174 | ⊕⊕⊝⊝ | Both treatment and control groups in both studies included in this analysis showed significantly decreased ICS use at end of study compared with baseline but no intergroup difference was detected |
| *The basis for the assumed risk (e.g. median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). **All adverse events was not a prespecified outcome, but we have included it in the 'Summary of findings' table, as substantial data were contributed to this outcome. We have left out the asthma symptom scores outcome, as we were able to perform only a limited narrative analysis. | ||||||
| GRADE Working Group grades of evidence. | ||||||
| aOnly a small number of included studies reported this outcome, suggesting lack of relevance in this study population. Treatment period in Calderon 2006 was just 4 weeks and exacerbations requiring ED/hospital admission/OCS are rare events. Downgrade once. bNo events but could be a product of the asthma severity of the recruited population. No downgrade. cFunnel plot not possible as no one outcome shows > 10 studies contributing events. Many reports are conference abstracts without associated peer‐reviewed full publication. Downgrade once. d5/21 studies included in this analysis were assessed as having high risk of performance and detection bias, but none of the 5 contributing events. No other serious issues. e5/21 studies included a mixed population of participants with asthma and rhinitis (but all > 80% with asthma), but the 5 studies contributing events to this analysis recruited exclusively participants with asthma. fEvents rare. Participants had largely mild to moderate asthma and may have been less at risk of serious adverse events. Downgraded once for indirectness. gTwo studies contributing events assessed as having high risk of performance and detection bias, with 3 others at high risk but not contributing events. Study contributing greatest weight (41%) to the analysis reported only as a conference abstract with uncertainty about attrition bias. Downgrade once. hSix out of 19 studies included mixed rhinitis and asthma populations, and of those contributing events made up approximately 25% of the analysis weight. Most of these events were mild and transient and did not lead to participant withdrawal. Downgrade once. iTwo out of four (contributing > 50% of analysis weight) studies assessed at high risk of performance and detection bias. Downgrade once. jHigh level of heterogeneity (I2 = 76%) and combines PC20 with PD20 scores using SMDs. Downgrade once. kPossibility of benefit in control group not excluded by confidence intervals. Downgrade once. lCalculated as the weighted mean of control group scores of the included studies. mImprecise estimate with confidence intervals including the possibility of a clinically important harm or benefit from SLIT. Downgrade once. nMany participants in included studies had mild asthma and so would be less likely to be using ICS. This was a predefined outcome, which may have less relevance to the study population.. Downgrade once. | ||||||
Background
Description of the condition
Asthma is a common long‐term respiratory disease that affects both adults and children. It is characterised by reversible airflow limitation, typically leading to recurrent wheezing, chest tightness, shortness of breath and cough. Symptoms may vary over time and in intensity and can be triggered by factors including allergens, viral illnesses and exercise (CDC 2013; GINA 2014). Airflow limitation is a result of several factors including bronchoconstriction, airway oedema, bronchial hyper‐responsiveness and airway remodelling, which may become irreversible over time (NAEPP 2007). Asthma therapy generally aims to reduce smooth muscle constriction through the use of inhaled agents such as long‐ and short‐acting beta2‐agonists (LABA and SABA) and to reduce airway inflammation through therapies such as inhaled corticosteroids (ICS) and leukotriene receptor antagonists (LTRA) (BTS/SIGN 2014).
Although estimates vary between populations, it is increasingly recognised that for as many as 50% of those with asthma, their condition has an important atopic component (Agache 2012; Arbes 2007; Normansell 2014; Pearce 1999), defined by a positive skin prick test to a recognised allergen, which may provide a therapeutic target for immunotherapy.
Atopy is defined as the production of specific immunoglobulin (Ig)E in response to common environmental allergens; it can be identified through skin prick testing. Total serum IgE has also been associated with asthma. Up to 95% of adults and children with asthma are skin prick test positive for one or more allergens (Craig 2008), but it should be noted that more than 50% of non‐asthmatic children and adults are also skin prick test positive (Arbes 2007).
Description of the intervention
The aim of immunotherapy is to build up tolerance to an allergen through repeated exposure to the causative allergen. Subcutaneous immunotherapy (SCIT) is well established in the United States, whereas survey data from 2011 suggest that only 11.4% of US allergists prescribe sublingual immunotherapy (SLIT) (Sikora 2013). In Europe, SLIT represents approximately 45% of immunotherapy and up to 80% of new prescriptions for immunotherapy (Cox 2009; Linkov 2014). SLIT is available as tablets or as a solution and is usually taken in the morning, once daily, on alternate days, or twice weekly, according to manufacturer instructions. The drops or tablets are kept under the tongue for one to two minutes before they are swallowed. A build‐up phase of gradually increasing doses is usually followed by a maintenance phase at the maximum dose. It is currently thought that a SLIT course should last for three to five years, which is consistent with evidence derived from trials of SCIT (Passalacqua 2012). Considerable inconsistency can be seen in the literature about safe and effective dosing of SLIT, and a recent World Allergy Organization position paper states that a regimen will have to be established individually for each allergen extract formulation (Canonica 2014).
How the intervention might work
Recognition of the important allergic component for many people with asthma has led to interest in the use of immunotherapy directed against specific allergens; although the efficacy of subcutaneous immunotherapy for asthma has been established, evidence for SLIT is conflicting (Incorvaia 2010; Passalacqua 2012). Allergen‐specific sublingual and subcutaneous immunotherapy is thought to work primarily by inducing T‐cell tolerance and promoting regulatory T‐cells, which secrete the suppressive cytokines interleukin (IL)‐10 and transforming growth factor (TGF)‐beta. This in turn leads to production of the non‐inflammatory immunoglobulins IgG4 and IgA, thus directing the immune response away from the inflammatory, atopic IgE response.(Fujita 2012). The hope is that targeting the dysregulated underlying immune response and thus desensitising the immune system to the specific allergen will permit those with allergic asthma to experience improvement in symptoms (Jutel 2014). The sublingual route of administration may offer advantages over the subcutaneous route, not only in terms of acceptability to patients. The oral cavity is a naturally 'tolerogenic environment', as it frequently encounters foreign proteins without the provocation of a local or systemic immune response and therefore may be an appropriate site for delivery of a treatment intended to produce immune tolerance (Canonica 2014). Pharmacokinetic studies suggest that the allergen extracts are retained for some time in the oral mucosa before they drain to local lymph nodes. This may account for the relative frequency of local reactions and infrequency of serious, systemic reactions (Marcucci 2007).
Why it is important to do this review
Asthma is thought to affect approximately 300 million people worldwide (Partridge 2006)—between 1% and 18% of the population in different countries (GINA 2014). The burden of the disease is considerable; in the United States alone, asthma costs approximately $56 billion a year and in 2009 led to 479,300 hospitalisations and 3388 deaths (CDC 2013); more asthma‐related death is thought to occur in middle‐ and low‐income countries (WHO). Many people with asthma remain inadequately controlled despite treatment and therefore are at high risk of exacerbation (Partridge 2006). Allergen‐specific immunotherapy may represent an important addition to the more established asthma therapies and thus may help to reduce the morbidity and mortality associated with this disease. Indeed, it is the only treatment that specifically targets underlying causes of allergen‐triggered asthma, and it may lead to long‐term desensitisation (Di Rienzo 2003). Moroever, SLIT may represent a more acceptable and safer route of administration than SCIT (Linkov 2014). However, the position of SLIT as a therapeutic option for asthma has yet to be established. Most national and international guidelines do not recommend its routine use for asthma because evidence of efficacy and safety is robust, or they recommend use only in those with symptoms difficult to control with standard treatments (BTS/SIGN 2014; GINA 2014; NAEPP 2007).
Objectives
To assess the efficacy and safety of sublingual immunotherapy compared with placebo or standard care for adults and children with asthma.
Methods
Criteria for considering studies for this review
Types of studies
We included parallel randomised controlled trials (RCTs), blinded and unblinded, of any duration that evaluated sublingual immunotherapy versus placebo or as an add‐on to standard medical management of asthma. We excluded cross‐over trials because of the long‐term effects of treatment. We included studies reported as full text, those published as abstract only and unpublished data.
Types of participants
We included both adults and children with asthma of any severity, diagnosed by a clinician or according to validated national or international guidelines (e.g. BTS/SIGN 2014; GINA 2014). Participants could have any allergen‐sensitisation pattern. We included participants with a dual diagnosis of asthma and allergic rhinitis. As a pragmatic decision, and in a change to our protocol, we chose to exclude studies in which less than 80% of participants were reported to be diagnosed with asthma at baseline, as findings for patients with asthma were rarely presented separately. We excluded patients with other respiratory co‐morbidities.
Types of interventions
We included trials evaluating any type or dose of SLIT (including single‐allergen and multiple‐allergen preparations) versus placebo or as an add‐on to standard medical management of asthma.
We included trials that allowed the use of short‐acting reliever medications such as salbutamol, provided these medications were not part of the randomly assigned treatment. We also included trials that allowed participants to continue their usual preventative asthma medication (e.g. LABA/ICS/LTRA), again provided this was not part of the randomly assigned treatment.
Types of outcome measures
Primary outcomes
-
Exacerbation requiring emergency department (ED) visit or hospitalisation (participants with at least one).
-
Quality of life* (measured on a validated scale, e.g. Asthma Quality of Life Questionnaire).
-
Serious adverse events (all‐cause).
Secondary outcomes
-
Asthma symptom scores* (measured on a validated scale, e.g. Asthma Control Questionnaire).
-
Exacerbations requiring systemic corticosteroids (participants with at least one).
-
Response to provocation tests*.
-
Required dose of ICS.
Reporting by trial authors of one or more of the outcomes listed here was not an inclusion criterion for the review.
*If more than one validated scale measuring the same construct was reported within a study, or if different scales were used across studies, we analysed them together using standardised mean differences.
Outcomes were selected to reflect those most important to people with asthma after a check of the existing literature (Busse 2012; Sinha 2012).
Search methods for identification of studies
Electronic searches
We identified trials from the Cochrane Airways Group Specialised Register (CAGR), which is maintained by the Trials Search Co‐ordinator for the Group. The Register contains trial reports identified through systematic searches of bibliographic databases including the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, the Cumulative Index to Nursing and Allied Health Literature (CINAHL), the Allied and Complementary Medicine Database (AMED) and PsycINFO, and through handsearching of respiratory journals and meeting abstracts (see Appendix 1 for further details). We searched all records in the CAGR using the search strategy provided in Appendix 2. We also conducted a search of ClinicalTrials.gov (www.ClinicalTrials.gov) and the World Health Organization (WHO) trials portal (www.who.int/ictrp/en/) for relevant studies. We conducted the most recent search on 25 March 2015.
Searching other resources
We checked reference lists of all primary studies and review articles for additional references.
We searched on 14 March 2015 for errata or retractions from included studies published in full text on PubMed (www.ncbi.nlm.nih.gov/pubmed).
Data collection and analysis
Selection of studies
Two review authors (RN and KMK) independently screened titles and abstracts to consider inclusion of all potential studies identified as a result of the search and coded them as 'retrieve' (eligible or potentially eligible/unclear) or 'do not retrieve'. We retrieved the full‐text study reports and publications, and two review authors (RN and KMK) independently screened the full texts to identify studies for inclusion. We identified and recorded reasons for exclusion of ineligible studies, resolving disagreements through discussion or, if required, by consultation with a third person. We identified and excluded duplicates and collated multiple reports of the same study, so that each study rather than each report was the unit of interest in the review. We recorded the selection process in sufficient detail to complete a PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‐Analyses) flow diagram (Figure 1) and a Characteristics of excluded studies table.
Study flow diagram.
Data extraction and management
We used a Microsoft Excel data collection form that had been piloted on at least one study in the review to document study characteristics and outcome data. Two review authors (RN, KMK or ALB) extracted the following study characteristics from included studies.
-
Methods: study design, total duration of study, details of any 'run‐in' period, number of study centres and locations, study setting, withdrawals, dates of study.
-
Participants: N, mean age, age range, gender, severity of condition, diagnostic criteria, baseline lung function, smoking history, inclusion criteria, exclusion criteria.
-
Interventions: intervention, comparison, concomitant medications, excluded medications.
-
Outcomes: primary and secondary outcomes specified and collected, time points reported.
-
Notes: funding for trial, notable conflicts of interest of trial authors.
Two review authors (RN, KMK or ALB) independently extracted outcome data from included studies. We resolved disagreements by reaching consensus or by involving the third review author. All three review authors transferred data into the Review Manager (RevMan 2012) file. We double‐checked that data were entered correctly by comparing data presented in the systematic review with data from the study reports.
Assessment of risk of bias in included studies
Two review authors (RN, KMK or ALB) independently assessed risk of bias for each study using the criteria outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We resolved disagreements by discussion or by consultation with a third review author. We assessed risk of bias according to the following domains.
-
Random sequence generation.
-
Allocation concealment.
-
Blinding of participants and personnel.
-
Blinding of outcome assessment.
-
Incomplete outcome data.
-
Selective outcome reporting.
-
Other bias.
We graded each potential source of bias as high, low or unclear and provided a quote from the study report together with a justification for our judgement in the 'Risk of bias' tables within the Characteristics of included studies table. We summarised risk of bias judgements across different studies for each of the domains listed. We considered blinding separately for different key outcomes when necessary (e.g. for unblinded outcome assessment, risk of bias for all‐cause mortality may be very different than for a patient‐reported symptom scale). When considering treatment effects, we took into account risk of bias for studies that contributed data to that outcome.
Assesment of bias in conducting the systematic review
We conducted the review according to the published protocol (Normansell 2014a), and we report deviations from it in the Differences between protocol and review section of the systematic review.
Measures of treatment effect
We analysed dichotomous data as odds ratios and continuous data as mean differences (MDs) or standardised mean differences (SMDs). For rare events, we used risk differences (RDs) to account for trials with no events in either arm. We entered data presented as a scale with a consistent direction of effect. We used change from baseline scores when possible.
We undertook meta‐analyses only when this was meaningful (i.e. if treatments, participants and the underlying clinical question were similar enough for pooling to make sense).
We narratively described skewed data reported as medians and interquartile ranges and explained when meta‐analysis was not considered appropriate.
When multiple trial arms were reported in a single trial, we included only the relevant arms. If two (or more) comparisons (e.g. drug A vs placebo, drug B vs placebo) were combined in the same meta‐analysis, we halved (or divided by the appropriate number to reflect the number of treatment arms) the control group to avoid double‐counting.
If trials reported outcomes at multiple time points, we used the end of treatment time point. As the benefits of immunotherapy are intended to persist beyond the treatment period, we also looked for primary outcomes reported at follow‐up off treatment and described them, when available.
Unit of analysis issues
For dichotomous outcomes, we used participants rather than events as the unit of analysis (i.e. number of participants admitted to hospital at least once rather than number of admissions per participant).
Dealing with missing data
We planned to contact investigators or study sponsors to verify key study characteristics and to obtain missing numerical outcome data when possible (e.g. when a study is identified as an abstract only), but owing to the large number of studies included, we attempted to contact study authors only to clarify whether a study did or did not meet our inclusion criteria.
Assessment of heterogeneity
We used the I² statistic to measure heterogeneity among the trials in each analysis. If we identified substantial heterogeneity, we reported this and explored possible causes by performing prespecified subgroup analysis.
Assessment of reporting biases
We were not able to construct a funnel plot because the only primary outcome that was included in more than 10 trials was serious adverse events (SAEs), and only five studies contributed events.
Data synthesis
We used a random‐effects model for all analyses, as we expected variation in effects due to differences in study populations and methods. We performed sensitivity analyses using a fixed‐effect model when we encountered substantial heterogeneity.
Summary of findings table
We created summary of findings Table for the main comparison using data from seven outcomes. In a change to our protocol, we did not include asthma symptoms as we did not perform a meta‐analysis for this outcome and instead included all adverse events. We used the five GRADE (Grades of Recommendation, Assessment, Development and Evaluation) considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of a body of evidence as it relates to studies that contributed data to the meta‐analyses for prespecified outcomes. We used methods and recommendations described in Section 8.5 and Chapter 12 of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011) with GRADEpro software. We justified all decisions to downgrade or upgrade the quality of studies by using footnotes, and we made comments to aid readers' understanding of the review when necessary.
Subgroup analysis and investigation of heterogeneity
When possible, we intended to carry out the following subgroup analyses for primary outcomes, using the formal test for subgroup differences in Review Manager (version 5.3) (RevMan 2012).
-
Age of participants (adults vs children).
-
Asthma severity (as defined by baseline severity reported in the trial or by review authors' assessment according to the asthma medication used).
-
Type of target allergen for sublingual immunotherapy (e.g. house dust mite (HDM), grass pollen).
-
Study duration (> or < one year).
Sensitivity analysis
We carried out sensitivity analyses while excluding the following.
-
Studies at high risk of bias for blinding.
-
Unpublished data (i.e. no peer‐reviewed full paper available).
Results
Description of studies
Details of methods, participants, interventions and outcomes for all included studies can be found in the Characteristics of included studies tables.
Results of the search
We identified 372 records through initial database searching and a further 61 from searches of clinicaltrials.gov and the World Health Organization (WHO) Clinical Trials Register. After removing duplicates, we screened 401 records, checking title and abstract only, and excluded 177. We assessed the remaining 224 full texts for eligibility and excluded 150 records (referring to 111 unique studies, plus seven ongoing studies and 12 studies awaiting classification) at this stage. Of those excluded, the majority included a mixed study population of participants with asthma, rhinitis or both, and results from participants with asthma were not presented separately (n = 53). As a pragmatic decision, and in a change to our protocol, we chose to exclude studies in which less than 80% of participants were reported to be diagnosed with asthma at baseline. We excluded 12 studies because we were unable to ascertain the percentage of participants with asthma at baseline.
We included 52 individual studies (74 records) in the qualitative synthesis; 34 contributed data to at least one meta‐analysis, but 27 of these appeared only in the adverse or serious adverse events analyses. Three studies appeared only in the narrative synthesis of unvalidated symptom or medication scores (Cooper 1984; Lewith 2002; Reilly 1994). Fifteen studies did not report any data relevant to this review (Almarales 2012; Hanna 2013; Inal 2009; Karakoc‐Aydiner 2011; Keles 2009; Marcucci 2003; Mosges 2010; Muratore 1993; Orefice 2004; Radu 2007; Rodriguez 2012; Rodriguez Santos 2004; Tian 2014; Virchow 2014; Yukselen 2013). A study flow diagram is presented in Figure 1.
Included studies
Fifty‐two studies met our inclusion criteria, given the pragmatic change to the protocol described above. These studies included a total of 5256 participants, and 5077 were randomly assigned to comparisons of interest in this review. The largest included study randomly assigned 834 participants, and the smallest just 15. The median total number of participants across all 52 studies was 56. Thirty‐seven were reported as full peer‐reviewed articles, 14 were published as abstracts only (i.e. we did not identify a linked full‐text article) and one was found only on clinicaltrials.gov.
Methods
As per our protocol, all included trials were RCTs with parallel design and compared SLIT versus placebo plus conventional therapy (n = 39) or conventional pharmacotherapy alone (n = 13). Six studies (Eifan 2009; Hanna 2013; Karakoc‐Aydiner 2011; Keles 2009; Keles 2011; Mungan 1999; Yukselen 2013) included one or more arms that were not relevant to this review, for example, SCIT or SCIT plus SLIT. Trial duration varied greatly across studies, with the shortest lasting just one day and the longest 156 weeks. Several studies included a run‐in period, and 10 included a period of post‐treatment follow‐up ranging from two weeks to two years. Outcomes data were extracted at the last time point reported, which was end of treatment in six studies and post‐treatment in three studies; in one study different outcomes were reported at different time points. Trials were conducted in a variety of countries worldwide, but most were carried out in Europe (including Turkey) (n = 33) and Asia (n = 8). Only one study recruited participants in the USA.
Participants
We included studies involving both children and adults. Eighteen studies recruited only teenagers and adults and 25 studies children only; two studies included mixed populations of adults and children. In seven studies, the age range of participants was not reported. Most studies did not specify the ethnicity of participants.
Most of the included studies (n = 44) recruited exclusively participants with asthma, although severity of the condition ranged from mild and intermittent to moderately severe. Eight studies stated that participants with asthma 'and/or' rhinitis were included, meaning that investigators recruited participants with a diagnosis of asthma or rhinitis or both. As has been mentioned, we included these studies only if we could confirm that more than 80% of participants had an asthma diagnosis at baseline. We excluded 53 studies because less than 80% of participants had asthma; we excluded12 because we were unable to confirm the percentage of participants with asthma at baseline despite attempts to contact the trial authors.
The inclusion criteria of most studies stated that participants must have had a positive skin prick test to the allergen of interest and/or serum allergen‐specific IgE above a specified threshold. Usually, participants were also required to have a clinical history consistent with allergic asthma or rhinitis or both. Some studies stated that they excluded participants sensitised to other common aero‐allergens and those with severe asthma or with other co‐morbidities. Most studies excluded participants who had received immunotherapy in the past.
Interventions
More than half of the included studies (n = 34) targeted house dust mite (HDM) allergy, with the remainder targeting grass pollen (n = 6), birch pollen (n = 3), cockroach (n = 1), cat dander (n = 1), Alternaria (n = 1), Parietaria (n = 1), olive pollen (n = 1), Artemisia (n = 1) and a combination of HDM and Parietaria (n = 1). The remaining two studies involved homeopathic SLIT compared with placebo: One used HDM homeopathic SLIT and the other various allergens according to participant allergic response, with HDM the dominant allergen (84% of participants). As homeopathic SLIT represents a different entity from standard SLIT (with the allergen far more diluted), we intended to exclude these studies in a sensitivity analysis. However, neither study (Lewith 2002; Reilly 1994) contributed data to a meta‐analysis, so this was not necessary. Dosing also varied across studies; when possible, we have extracted this information and presented it in the Characteristics of included studies tables.
Typically, SLIT interventions targeting perennial allergens, such as HDM, were administered continuously, while those targeting seasonal allergens, such as grass pollen, were administered before the start of the pollen season or during the pollen season. Most studies stated that participants were allowed to continue using specified rescue medication for asthma and rhinitis symptoms throughout the study, and in some trials the frequency of use of rescue medication was an efficacy outcome. Most studies made no changes to baseline preventer medication, such as ICS.
Outcomes
Outcomes reported were not consistent across reviews, and validated scales were rarely used. Asthma symptoms were reported by a large majority of included studies (n = 42), as were medication use scores (n = 36). Many studies also reported outcomes not specified in our protocol, including lung function such as peak expiratory flow rate (PEFR) (n = 32) and laboratory immunological outcomes such as serum allergen‐specific IgE and IgG levels (n = 31). Adverse events were reported by just over half of the included studies (n = 27). Outcomes less frequently reported included skin prick tests (n = 16), bronchial provocation tests (n = 11), quality of life (n = 6) and exacerbations (n = 5). Despite the large number of outcomes reported in the included studies, meta‐analysis was somewhat hampered by the wide range of unvalidated measures used; two out of our three primary outcomes of interest were rarely reported (exacerbations and quality of life). We have presented the data extracted for symptom scores and medication use by using unvalidated or incompatible scales in Analysis 1.8 and Analysis 1.9.
Subgroup and sensitivity analyses
Studies contributing data to our primary analyses were insufficient for us to complete the planned sensitivity and subgroup analyses. In a post hoc change to the protocol, we chose to investigate the subgroups of age, target allergen and study duration for all adverse events; these results are presented in Analysis 2.1, Analysis 2.2 and Analysis 2.3. We chose to perform a sensitivity analysis by excluding studies assessed to be at high risk of performance bias for all adverse events (Analysis 2.4).
Summary characteristics of the included trials including information about potential effect modifiers (e.g. age, treatment duration, allergen) are presented in Table 1, and full details of each included study are given in Characteristics of included studies.
| Study ID | Total N | Allergen | Comparator | Age range | Country | Duration | % with asthma |
| 120 | HDM | Placebo | Not reported | Cuba | 52 weeks | 100 | |
| 50 | Cat dander | Placebo | 14‐55 | Spain | 52 weeks | 81.8 | |
| 15 | HDM | Placebo | 7‐18 | Turkey | 26 weeks* | 100 | |
| 85 | HDM | Placebo | 7‐42 | France | 108 weeks* | 100 | |
| 48 | Grass pollen | Placebo | 4‐14 | Italy | 13 weeks* | 89.6 | |
| 43 | Grass pollen | Placebo | 18‐65 | Unclear | 4 weeks* | 100 | |
| 71 | HDM | Placebo | 18‐65 | UK and Denmark | 4 weeks | 100 | |
| 72 | HDM | Placebo | 5‐14 | Spain | 4 weeks | 100 | |
| 19 | Grass pollen | Placebo | 5‐15 | UK | > 8 but < 16 weeks* | 100 | |
| 44 | Alternaria | Pharmacotherapy | 18‐65 | Spain | 52 weeks | 100 | |
| 114 | Timothy grass | Placebo | 18‐65 | Denmark and Sweden | 19.5 weeks | 100 | |
| 48 | HDM | Pharmacotherapy | 5‐10 | Turkey | 52 weeks | 85 | |
| 55 | Grass pollen | Placebo | 18‐50 | Syria | Not reported | 100 | |
| 60 | HDM | Placebo | 13‐45 | Mexico | 26 weeks | 100 | |
| 60 | HDM | Placebo | Not reported | Not reported | 13 weeks | 100 | |
| 32 | HDM | Placebo | Not reported | Turkey | 52 weeks | 100 | |
| 86 | HDM | Placebo | 5‐12 | Italy | 26 weeks* | 100 | |
| 31 | HDM | Pharmacotherapy | 'Children' | Unclear | 156 weeks | 100 | |
| 53 | HDM | Pharmacotherapy | Not reported | Unclear | 17.3 weeks | 100 | |
| 58 | HDM | Pharmacotherapy | 5‐12 | Turkey | 52 weeks* | 100 | |
| 56 | HDM/Parietaria | Pharmacotherapy | 7‐68 | Italy | 52 weeks | 100 | |
| 18 | Artemisia pollen | Placebo | 15‐56 | Unclear | 7.14 weeks* | 100 | |
| 242 | Homeopathic HDM | Placebo | 18‐55 | UK | 16 weeks | 100 | |
| 20 | HDM | Placebo | 6‐12 | Taiwan | 24 weeks* | 100 | |
| 24 | HDM | Placebo | 4‐16 | Italy | 52 weeks | 84.6 | |
| 79 | Birch pollen | Pharmacotherapy | 18‐65 | Italy | 156 weeks* | 100 | |
| 604 | HDM | Placebo | 14+ | Denmark, Germany, Italy, Spain, United Kingdom, Sweden, France, Poland | 52 weeks | 100 | |
| 116 | Ultra‐rush birch pollen | Placebo | 6‐14 | Germany | 0.015 weeks | 100 | |
| 36 | HDM | Placebo | 18‐46 | Turkey | 52 weeks | 88 | |
| 28 | HDM | Placebo | 4‐9 | Italy | 52 weeks | 100 | |
| 124 | HDM | Placebo | 18‐65 | Spain | 104 weeks | 100 | |
| 110 | HDM | Placebo | 6‐12 | Taiwan | 24 weeks* | 100 | |
| 47 | HDM | Pharmacotherapy | Not reported | Italy | 156 weeks | 100 | |
| 24 | HDM | Placebo | 8‐15 | Italy | 104 weeks | 100 | |
| 30 | Parietaria | Placebo | 8‐14 | Italy | 56 weeks* | 100 | |
| 111 | HDM | Placebo | 5‐16 | France | 78 weeks | 100 | |
| 106 | HDM | Placebo | 5‐13 | Romania | 26 weeks | 100 | |
| 28 | Homeopathic HDM/feathers/ mixed moulds | Placebo | 16+ | Scotland | 4 weeks* | 100 | |
| 40 | HDM | Placebo | 'Adults' | Cuba | Not reported | 100 | |
| 50 | HDM | Pharmacotherapy | 6‐15 | Cuba | 104 weeks | 100 | |
| 264 | HDM | Pharmacotherapy | 3‐13 | China | 52 weeks | 82 | |
| 50 | Grass pollen | Placebo | 6‐17 | Poland | 104 weeks | 100 | |
| 60 | HDM | Placebo | 4‐18 | China | 48 weeks | 100 | |
| 24 | Birch pollen | Placebo | Not reported | Unclear | 104 weeks | 100 | |
| 834 | HDM | Plaecbo | Not reported | Austria, Croatia, Denmark, France, Germany, Lithuaina, Netherlands, Poland, Serbia, Slovakia, Spain, United Kingdom | 78 weeks | 100 | |
| 66 | Olive pollen | Placebo | 7‐17 | Greece | 104 weeks | 90.6 | |
| 484 | HDM | Placebo | 16‐50 | China | 52 weeks* | 100 | |
| 89 | Greer German cockroach | Placebo | 5‐17 | USA and UK | 13 weeks | 80 | |
| 32 | HDM | Placebo | Not reported | Turkey | 52 weeks | 100 | |
| 63 | HDM | Placebo | 'Adults' | France | 1.4 weeks | 100 | |
| 128 | HDM | Pharmacotherapy | 4‐14 | Taiwan | 104 weeks | 100 | |
| 106 | HDM | Pharmacotherapy | 4‐14 | China | Outcomes reported at 25 weeks | 100 |
*Studies that included post‐treatment follow‐up periods.
Excluded studies
We excluded studies that did not meet the criteria specified in our protocol or in which less than 80% of participants had received a diagnosis of asthma. We excluded 12 studies because we were unable to ascertain what percentage of the participants had asthma despite an attempt to contact the study authors. Reasons for exclusion of studies after the full text had been retrieved can be found in the Characteristics of excluded studies tables.
Risk of bias in included studies
For details on the risk of bias rating for each study and the reasons for each rating, see Characteristics of included studies. A summary of risk of bias judgements by study and by domain (sequence generation, allocation concealment, blinding, incomplete data and selective reporting) can be found in Figure 2.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Overall, a lot of uncertainty surrounded allocation procedures because of insufficient reporting, and about a quarter of the studies were at high risk of bias for blinding because they applied open‐label designs. Participant attrition was high or unknown in around half of the studies, and selective reporting is likely to have had a serious effect on the completeness of this evidence base.
Allocation
We assessed one study (Lewith 2002) as having low risk of bias for both random sequence generation and allocation concealment. Investigators used sealed envelopes followed by randomisation by minimisation according to age, sex, smoking status and asthma severity.
We considered 11 further studies to be at low risk of bias for random sequence generation. Seven studies (Caffarelli 2000; Eifan 2009; La Grutta 2007; Marcucci 2003; Mosbech 2014; Stelmach 2009; Yukselen 2013) used computer‐generated lists. Keles 2011 used the table randomisation method. Two studies (Pajno 2000; Pajno 2003) used a key code for random sequence generation. Reilly 1994 used a restricted technique of permuted blocks, stratified for intended allergen and daily dose of steroid. For these studies, no details were given on allocation concealment, and they were considered to be at unclear risk of bias in this domain.
Thirty‐nine studies stated that they were randomised but provided no specific details about sequence generation nor allocation concealment and were assessed to be at unclear risk of bias for both domains.
One study (Tian 2014) was at high risk of bias for random sequence generation, as participants were divided into treatment group and control group in order of admission. No details were given about allocation concealment; therefore we assessed risk of bias as unclear in this domain.
Blinding
We assessed most included studies (n = 37) described as double‐blind and placebo‐controlled as having low risk of bias in both performance bias and detection bias domains.
Two studies were assessed as having unclear risk of bias in both domains. Although Mungan 1999 was placebo‐controlled and single‐blind, no details were provided about who exactly was blinded. Radu 2007 was also single‐blind and did not include details on who was blinded.
La Grutta 2007 was rated as having high risk of performance bias, as the study was open‐label. Assessor blinding was not described for some outcomes, so we considered detection bias to be unclear.
We assessed 12 studies as having high risk of bias for both domains. Eight studies (Criado Molina 2002; Eifan 2009; Marogna 2005; Orefice 2004; Rodriguez Santos 2004; Shao 2014; Zhang 2013; Zheng 2012) were open‐label, which may have introduced bias. Hanna 2013 was a prospective study, with participants randomly assigned to three parallel groups with no mention of blinding. We made the assumption that three studies (Karakoc‐Aydiner 2011; Keles 2009; Keles 2011) were open‐label, as participant or assessor blinding was not mentioned.
Incomplete outcome data
Participant attrition was adequately described in 27 included studies, and we considered risk of attrition bias to be low. In 12 of these studies, no dropout was reported. In 14 other studies, withdrawal rates were low (no more than 20%), with similar rates reported in control groups. Pham‐Thi 2007 performed data analysis according to the intention‐to‐treat principle.
Altogether, we considered 16 studies to be at unclear risk for attrition bias. Of these, 13 studies provided no information about withdrawal rates. Cooper 1984 excluded three participants from its treatment group and four from the placebo group who were not included in the analysis. However, the paper does not report whether these exclusions were part of the asthma series and did not attempt to impute results for dropouts. One study (Radu 2007) was stopped after six months and also did not report dropout rates. Shao 2014 had a balanced and low dropout below 20% but did not include these data in the efficacy analysis.
We assessed nine studies (Alvarez‐Cuesta 2007; Bousquet 1999; Criado Molina 2002; Marogna 2005; NCT00633919; Orefice 2004; Pajno 2000; Stelmach 2009; Wood 2014) as having high risk of bias in this domain because of high withdrawal rates and/or unbalanced dropout between treatment and control groups and/or because only completers were analysed. Orefice 2004 also excluded individuals with more severe asthma during the trial; however, it is not clear whether this was baseline exclusion or exclusion during the study.
Selective reporting
Only 14 studies reported all stated outcomes and were assessed as having low risk of reporting bias.
We considered six studies to be at unclear risk of reporting bias. The numerical reporting of Criado Molina 2002 was inconsistent, and data could not be included in the meta‐analysis. Marcucci 2003 reported outcomes well (although mainly non‐clinical) but did not report a trial registration to check whether all prespecified outcomes were included in the write‐up. Pajno 2003 reported several outcomes narratively or gave 'ranges' of P values. Some discrepancies between reports appear to be related to the same participant group. All stated outcomes were reported in Rodriguez Santos 2004, but numerical data were not well presented; within‐group outcomes were reported rather than comparisons with control. Wood 2014 and Zheng 2012 did not clearly report adverse event outcomes.
We assessed 32 included studies as having high risk of bias for this domain. Fourteen studies were provided only as conference abstracts, with minimal information and details regarding the conduct of the study as well as data that could not be meta‐analysed. Fourteen studies did not report data for all outcomes, selectively reported outcome data or lacked numerical supporting data (Bousquet 1999; Calderon 2006; Cooper 1984; Corzo 2014 (a); Corzo 2014 (b); Eifan 2009; Gomez Vera 2005; Mosbech 2014; Mosges 2010; Mungan 1999; Pham‐Thi 2007; Tian 2014; Wang 2014; Yukselen 2013). Most outcomes were reported with a level of statistical significance in only three studies (La Grutta 2007; Lewith 2002; Vourdas 1998) and could not be included in the meta‐analysis. Although Marogna 2005reported all stated outcomes, several were provided only in graphical form or with inexact P values that also could not be meta‐analysed.
Other potential sources of bias
We considered three studies as having other potential sources of bias. Alvarez‐Cuesta 2007 had an unbalanced male‐to‐female ratio, and Radu 2007 was stopped after six months (planned for 36 months) because of statistically significant differences in outcomes that favoured the active treatment. Reilly 1994, a study of homeopathic SLIT, stated that "both doctors (homeopathic and asthma clinic doctor) could also veto any patient they considered unsuitable", which may have introduced bias.
Effects of interventions
See: Summary of findings for the main comparison SLIT vs control
Primary outcomes
Exacerbations requiring ED or hospital admission
Only one short study of 43 participants, involving four different SLIT dosing arms (Calderon 2006), included this outcome and reported no events during the four‐week treatment period nor during the five‐ to six‐week follow‐up period (Analysis 1.1; low‐quality evidence).
Quality of life
Quality of life (QoL) was a stated outcome in five included studies (Bousquet 1999; Inal 2009; Lewith 2002; Mosbech 2014; Pham‐Thi 2007), but none presented data in a manner that allowed for meta‐analysis. Bousquet 1999 reported increased QoL scores using the Short‐Form Health Status Survey (not specific for asthma) in the SLIT group compared with the placebo group after 25 months of treatment, and improvements were statistically significant in several domains, including general mental health, general perception of health and physical pain. Inal 2009, a conference abstract, also reported significant improvement in QoL scores after two years of SLIT treatment when compared with placebo, but the scale used was not reported. Lewith 2002, a study of homeopathic SLIT versus placebo, reported asthma QoL using the 'asthma bother profile' and did not find a statistically significant difference between groups. Mosbech 2014 narratively and graphically reported the Asthma Quality of Life Questionnaire (AQLQ) after a year of SLIT treatment (including three different dosing arms) and did not find a statistically significant difference between active treatment and placebo. Finally, Pham‐Thi 2007 assessed QoL using two forms of the Childhood Asthma Questionnaire (CAQ). The severity dimension showed statistically significant improvement in the SLIT group compared with the placebo group, but in all domains, average changes were not statistically different between groups. We have presented in Analysis 1.2 the numerical data extracted from Bousquet 1999 and Lewith 2002.
Serious adverse events
Occurrence of serious adverse events (SAEs) was a reported outcome for 22 included studies involving 2560 participants, but only five studies (Mosbech 2014; NCT00633919; Niu 2006; Pajno 2000; Wang 2014) observed any events. Although events were infrequent, analysis using risk differences (RDs) suggests that no more than one in 100 are likely to suffer an SAE as a result of treatment with SLIT (Figure 3; Analysis 1.3; RD 0.0012, 95% confidence interval (CI) ‐0.0077 to 0.0102; moderate‐quality evidence).
Forest plot of comparison: 1 SLIT vs control, outcome: 1.3 Serious adverse events.
In total, 22 participants receiving SLIT and 12 in the control groups experienced an SAE. Mosbech 2014 reported that 15 participants receiving active treatment experienced an SAE: six in the 1 standard quality (SQ)‐HDM group, three in the 3 SQ‐HDM group and six in the 6 SQ‐HDM group. Of these events, only two were deemed by investigators to be possibly related to SLIT and were described in detail: One was a case of migraine and the other dizziness. Four participants receiving placebo experienced an SAE. In NCT00633919, two participants receiving active treatment experienced an SAE (one road traffic accident and one femur fracture) and two in the placebo group (one perianal abscess and one a diagnosis of obsessive‐compulsive disorder). Five participants experienced an SAE in Niu 2006 (one in the SLIT group and four in the control group), but these events were not further described. In Pajno 2000, a 'serious asthma attack' led to withdrawal of a participant from the control group. In Wang 2014, six SAEs occurred involving five participants (four in the SLIT group and one in the control group) and included a knee fracture, Arnold‐Chiari Syndrome, contact dermatitis, ovarian cyst rupture, pneumonia and traumatic brain injury. None of these events were thought to be treatment‐related, and none of the included studies reported any deaths.
Secondary outcomes
Asthma symptom scores
Most included studies (n = 42) reported asthma symptoms as an outcome, but a variety of often unvalidated scales were used and numerical data were not always presented. Details of the scoring systems used are presented in Analysis 1.8. We judged that a meta‐analysis using standardised mean differences of those studies presenting numerical data would not be a sound methodological approach, and so the data extracted from the included studies are tabulated in Analysis 1.8 but were not meta‐analysed. In summary, of those presenting numerical data, five studies found no statistically significant differences in asthma symptom scores between groups (Bousquet 1999; Dahl 2006; Lewith 2002; Mungan 1999; Pham‐Thi 2007). Alvarez‐Cuesta 2007 reported a marked reduction in asthma symptoms during cat exposure for the SLIT group, with no significant improvement observed in the placebo group. Caffarelli 2000, Eifan 2009, Ippoliti 2003, Marogna 2005, Niu 2006, Pajno 2000, Reilly 1994, Stelmach 2009 and Zheng 2012 reported statistically significant reductions in asthma symptom scores in the SLIT group compared with the placebo group at the end of treatment. Cooper 1984 reports a 'small advantage' in favour of SLIT in number of days with asthma symptoms and in symptoms graded according to severity. Lue 2006 found improvements in both daytime and nighttime asthma symptom scores in the SLIT group, although the between‐group difference for the former was not statistically significant.
Medication use scores
Similarly, 12 studies reported numerical medication use scores, which were frequently unvalidated aggregate scores including rescue medication use and use of inhaled corticosteroid (ICS) and oral corticosteroid (OCS). Details of the scoring systems used are presented in Analysis 1.9. Although medication scores were not a predefined outcome, as many of them incorporated ICS use (which was an outcome of interest), we extracted the data and have presented them, again without meta‐analysis (Analysis 1.9). Seven studies did not find a statistically significant difference between medication use scores for SLIT and control groups at the end of treatment (Bousquet 1999; Dahl 2006; Lewith 2002; Lue 2006; NCT00633919; Niu 2006; Pham‐Thi 2007). Four studies found a statistically significant difference favouring SLIT when compared with control in asthma medication scores at the end of treatment (Eifan 2009; Marogna 2005; Pajno 2000; Stelmach 2009). Mungan 1999 reports a statistically significant decrease from baseline in asthma medication scores in the SLIT group but no decrease in the control group.
Exacerbations requiring systemic corticosteroids
Both Calderon 2006 and Pajno 2003 included this outcome, but neither study observed any events (Analysis 1.4; low quality).
Response to provocation tests
Response to bronchial provocation using the methacholine challenge test was included as an outcome in 11 studies, and four (Keles 2011; Marogna 2005; Pajno 2003; Stelmach 2009) contributed to the meta‐analysis. Marogna 2005 reported this outcome using provocative dose (PD)20 and the remaining studies provocative concentration (PC)20. Studies targeted a variety of allergens including HDM (n = 1), birch pollen (n = 1), Parietaria (n = 1) and grass pollen (n = 1). All four studies were at least a year in duration. Reilly 1994 reported change from baseline (rather than endpoint) PC20log and for this reason could not be reliably pooled with the other measures in the meta‐analysis. This study reported a small benefit for homeopathic SLIT over placebo, which was not statistically significant.
Heterogeneity between the four studies that contributed to the meta‐analysis was significant for response to bronchial provocation tests, and the confidence intervals were too wide to allow a clear judgement of SLIT benefit (Analysis 1.6; SMD 0.69, 95% CI ‐0.04 to 1.43; participants = 139; studies = 4; very low quality) with a high level of heterogeneity (I2 = 76%). When a fixed‐effect model was used to further investigate heterogeneity, the effect suggested a small benefit from SLIT (SMD 0.72, 95% CI 0.37 to 1.08). If Marogna 2005, the only study reporting PD20, was removed from the analysis, heterogeneity was somewhat lower (I2 = 55%), but the pooled effect remained imprecise and was not statistically significant.
Required dose of ICS
Three studies (Bousquet 1999; Niu 2006; Pham‐Thi 2007) reported ICS use numerically at the end of treatment: Bousquet 1999 in beclomethasone mcg/d, Pham‐Thi 2007 in budesonide mcg/d (equivalent) and Niu 2006 in puffs/d. Differences between groups in puffs per day of ICS were not statistically significant (Niu 2006) at the end of treatment. We have not included these results in the meta‐analysis. Although ICS use significantly decreased from baseline in both treatment and control groups in Bousquet 1999 and Pham‐Thi 2007, pooling of ICS use at the end of treatment yielded an imprecise estimate with wide confidence intervals including the possibility of both benefit and harm from SLIT (Analysis 1.7; MD 35.10, 95% CI ‐50.21 to 120.42, low quality) with no heterogeneity (I2 = 0%).
Both Mosbech 2014 and Virchow 2014 also assessed ICS reduction and reported that participants taking higher‐dose SLIT treatment experienced a significant reduction in ICS use compared with those given placebo at the end of treatment, but neither study presented data in a way that allowed for meta‐analysis.
All adverse events
In a change to the protocol and as a result of the infrequency of SAEs, we chose to include an analysis of all adverse events. We extracted data for all adverse events, not just those deemed to be treatment‐related. Nineteen studies including 1755 participants reported all adverse events, and 11 contributed more than 500 events to the meta‐analysis. Pooled results demonstrated increased risk of experiencing an adverse event in the SLIT group compared with the control group; this finding was statistically significant (Figure 4; Analysis 1.5; odds ratio (OR) 1.70, 95% CI 1.21 to 2.38; low quality) with a low level of heterogeneity (I2 = 13%). This translates into an absolute increase from 222 per 1000 people in the control group to 327 per 1000 (257 to 404) and is presented graphically in Figure 5 by way of a Cates' plot. However, most adverse events were reported to be mild and transient and rarely led to withdrawal from the trial.
Forest plot of comparison: 1 SLIT vs control, outcome: 1.5 All adverse events.
Cates plot illustrating all adverse events (created at: www.nntonline.net)
Subgroup analyses
In a change to our protocol and as described above, we chose to perform subgroup analyses on adverse events, rather than serious adverse events, as so few data contributed to this primary outcome.
Participant age
We examined subgroups of children (mean participant age < 18) versus teenagers and adults (mean participant age ≥ 18) versus mixed age study populations or those for which the age range was not specified. The effect for adults and teenagers was more precise than for children because of the numbers of participants in the trials and the numbers of events observed in either group (Analysis 2.1; OR 1.48, 95% CI 1.06 to 2.06 vs OR 2.13, 95% CI 0.83 to 5.47), but results of tests for subgroup differences were not statistically significant (I2 = 0%, P value = 0.72).
Target allergen
More than half of the included studies targeted SLIT at HDM (n = 34); the next most common target allergen was pollen (n = 13). We chose to examine the subgroups of HDM versus pollen versus other or mixed allergens; no events were observed in studies within the other or mixed allergen subgroup, so this subgroup did not contribute statistically to the analysis. Participants receiving HDM SLIT and pollen SLIT were more likely to experience adverse events than those in the control group (Analysis 2.2; OR 1.47, 95% CI 1.10 to 1.97 and OR 5.48, 95% CI 1.99 to 15.05), and results of the test for subgroup differences were statistically significant (P value = 0.01), suggesting that those receiving pollen SLIT experienced more adverse events than those receiving HDM SLIT. However, we cannot conclude that this finding is a result of the different SLIT target allergen, as additional confounding between studies is likely.
Study duration
We chose to use a cutoff duration of less than 52 weeks versus 52 weeks or longer for this subgroup analysis. As we might expect, a smaller percentage of participants in the shorter studies experienced an adverse event during the study (Analysis 2.3; OR 1.53, 95% CI 0.38 to 6.19 vs OR 1.77, 95% CI 1.22 to 2.58), but results of tests for subgroup differences were not statistically significant (P value = 0.84), so we cannot draw any conclusions from this analysis about the interaction between study length and all adverse events.
Asthma severity
We did not perform the planned subgroup analysis according to baseline asthma severity as the majority of studies included participants with mild or intermittent symptoms, or did not describe baseline asthma severity in sufficient detail.
Sensitivity analyses
We chose to perform only two sensitivity analyses. First, we examined the effect of removing studies at high risk of performance or detection bias, or both, from the adverse events analysis. This analysis demonstrated a consistent direction of effect despite the removal of open‐label and unblinded trials (Analysis 2.4; OR 1.47, 95% CI 1.10 to 1.96).
Second, we removed studies that recruited a mixed population of participants with asthma and rhinitis from the adverse event analysis. As above, this had minimal impact on the pooled effect (Analysis 2.5; OR 1.42, 95% CI 1.06 to 1.91).
Discussion
Summary of main results
Fifty‐two studies met our inclusion criteria, randomly assigning 5077 participants to comparisons of interest in this review. Most of the studies were double‐blind and placebo‐controlled, but studies varied in duration from just one day to three years. The largest study included 834 participants, and the smallest 15. Just over half were conducted in Europe (including Turkey), and half recruited children only. Participants with severe asthma were excluded from most of the included studies, resulting in a study population consisting largely of participants with intermittent or mild symptoms.
With the exception of adverse events, reporting of meaningful clinical outcomes was generally poor. Only 22 studies contributed data to the primary outcome meta‐analyses: 22 to the serious adverse events outcome (with only five contributing events) and one to the analysis of exacerbations requiring hospital visits (no events). Although five studies numerically reported quality of life outcomes, the data were not suitable for meta‐analysis. This scarcity of evidence limits our ability to draw any conclusions about the effect of SLIT on exacerbations or quality of life. It would appear that SLIT is probably safe, at least in the population studied; although events were infrequent, analysis using risk differences suggests that no more than 1 in 100 are likely to suffer a serious adverse event as a result of treatment with SLIT.
Evidence from meta‐analysis is again lacking for secondary outcomes. Although many studies reported asthma symptom scores, a variety of largely unvalidated scales were used, and a narrative synthesis of those studies presenting numerical data did not reveal a consistent effect. However, no study reported statistically significant worsening of asthma symptoms with active treatment.
Similarly, a narrative synthesis of asthma medication use scores did not reveal a consistent effect; some studies reported improvement and others no improvement. Asthma medication use scores were generally unvalidated aggregate scores including, for example, rescue medication use, ICS use and OCS use. Again, no study reported significantly increased asthma medication use in the SLIT group. We were able to pool reduction in ICS use from two studies, which reported this in micrograms per day; no difference was found between active treatment and control, with wide confidence intervals including the possibility of both benefit and harm from SLIT. Two studies reported exacerbations requiring OCS, but no events occurred.
Eleven studies reported response to bronchial provocation testing, and four contributed to the meta‐analysis. The benefit of SLIT over control was not statistically significant, again with wide confidence intervals and a high level of heterogeneity.
All adverse events was not a prespecified outcome in our protocol, but we chose to extract these data because of the very infrequent occurrence of serious adverse events. Meta‐analysis of 19 studies with 11 contributing more than 500 events revealed a significant increase in participants reporting an adverse event on active treatment compared with control. However, the clinical importance of these events is doubtful, as they were usually transient and mild and rarely prevented participants from continuing in the trial. In addition, inclusion of respiratory symptoms as adverse events may have masked or minimised differences between groups, as an expected benefit of SLIT would be reduction of these symptoms.
Subgroup analysis of all adverse events according to participant age and study duration did not reveal significant subgroup differences. Findings suggest that those receiving SLIT for pollen allergy may experience more adverse events than those receiving SLIT for HDM allergy. Similarly, sensitivity analysis excluding those studies at high risk of performance and detection bias did not significantly alter this outcome.
Overall completeness and applicability of evidence
Despite identifying 52 unique studies that met our inclusion criteria, we were able to perform a very limited meta‐analysis. Fourteen of our included studies were reported as abstracts only and therefore provided minimal numerical data. Use of largely unvalidated symptom and medication scores also impeded quantitative synthesis of findings. Although a pooled analysis of composite asthma symptom and medication use scores using standardised mean differences would have been possible, we considered this approach to be not methodologically sound and believed it might result in misleading conclusions. We decided to include exacerbations, serious adverse events and quality of life as our primary outcomes, as these have been identified as important to people with asthma (Busse 2012; Sinha 2012). However, we recognise that this decision is also a limitation of this review, as most study participants had intermittent or mild persistent asthma and therefore were unlikely to be experiencing frequent exacerbations. In addition, we recognise that although treatment‐related adverse events are important in immunotherapy, risk of attribution bias is present if trialists are making this judgement, and unanticipated treatment‐related adverse events might not be identified. For this reason, we chose to include all‐cause adverse events.
Insufficient data contributing to the meta‐analysis also restricted potential subgroup analyses, resulting in difficulties in reaching any conclusions about SLIT efficacy in different age groups, for different allergens or for different treatment durations. As only a small minority of studies (n = 3) reported outcomes during post‐treatment follow‐up, we cannot comment in this review on the lasting benefits of SLIT for asthma.
The position of both SCIT and SLIT as potential therapeutic options for asthma has yet to be clearly established within international asthma guidelines. The Global Initiative for Asthma Guidelines (GINA 2014) state that the efficacy of allergen immunotherapy for asthma is limited, and that potential benefits of immunotherapy must be weighed against risk of adverse reactions, cost and duration of treatment. The UK guidance adopts a similar position and does not routinely recommend immunotherapy for asthma in adults or children (BTS/SIGN 2014). The organisation that advises the National Health Service (NHS) in the UK on cost‐effective treatments (the National Insitiute of Clinical Excellence ‐ NICE) currently does not provide guidance on asthma therapies.
The British National Formulary states that "desensitising vaccines should generally be avoided or used with particular care in patients with asthma" because of the risk of life‐threatening adverse events (BNF), and indeed both SCIT and SLIT are absolutely contraindicated in patients with severe or uncontrolled asthma (Slovick 2014). However, somewhat at odds with this, US Guidelines for the Diagnosis and Management of Asthma (NAEPP 2007) state that immunotherapy (SCIT) should be considered only in patients for whom standard pharmacological methods are insufficient (which implies that symptoms are not well controlled) and for whom a clear relationship between allergen exposure and symptoms is not evident. A more recent task force report in the USA (Cox 2011) further supports this by stating that "candidates for immunotherapy are patients whose symptoms are not controlled adequately by medications and avoidance measures", but does go on to specify that asthma must be controlled at the time of immunotherapy administration. Cox 2011 also highlights the investigational nature of SLIT in the USA at the time of the report and conflicting available evidence regarding its benefits. Until last year, no Food and Drug Admiistration (FDA) approval had been provided for any SLIT products in the USA; the first SLIT product ('Oralair') was approved in April 2014 for the treatment of allergic rhinitis (FDA Press Release 2014).
In the light of all information provided above, the applicability of our findings is somewhat limited, as most participants recruited to the studies included in this review had mild or intermittent symptoms, and so would not be likely candidates for immunotherapy for their asthma symptoms, at least according to current guidance (BTS/SIGN 2014; Cox 2011; GINA 2014; NAEPP 2007). Many of the included studies stated that participants must have a positive skin prick test or serum‐specific IgE to the allergen in question, but investigators did not necessarily specify that asthma symptoms must be linked to allergen exposure, again raising doubts about the appropriateness of the study populations. In addition, patterns of allergen sensitisation and association with asthma may vary geographically, limiting the generalisability of the findings of this review, given that most of the included studies were conducted in Europe (ISAAC 1998).
Quality of the evidence
We assessed the quality of the evidence presented in this review using GRADEpro software and present this information in summary of findings Table for the main comparison. Overall, evidence was assessed to be of moderate, low or very low quality, and evidence was downgraded for several reasons. Heterogeneity varied across individual outcomes, ranging from I2 = 0% for decrease in ICS use and serious adverse events to I2 = 76% for bronchial provocation.
We assessed evidence on exacerbations requiring ED visit or hospital admission and exacerbations requiring OCS to be of low quality. Neither outcome had any contributory events, and these outcomes were reported by very few studies (n = 1, Calderon 2006 for exacerbations requiring ED/hospital admission; n = 2 for exacerbations requiring OCS, Calderon 2006 and Pajno 2003). In addition, Calderon 2006 was a short study of just four weeks' duration, during which differences in rare events, such as exacerbations, might not be detected. The small number of studies reporting this outcome might also represent a publication bias.
We did not assess the quality of evidence for the quality of life outcome, as no study contributed numerical data to this outcome.
We assessed evidence for serious adverse events and all adverse events to be of moderate quality. We downgraded quality to reflect risk of performance and detection bias in contributing studies and mixed study populations including patients with asthma, rhinitis or both. Recruiting a 'mixed' population may have resulted in a population of patients with very mild and intermittent asthma symptoms, leading to concerns that adverse events might be under‐represented compared with those expected in a study population including participants with a diagnosis of more severe asthma.
We also assessed evidence for reduction in use of ICS to be of low quality. Only two studies contributed to the meta‐analysis (Bousquet 1999; Pham‐Thi 2007). We downgraded the evidence for imprecision and possible publication bias. Both studies reported a statistically significant decrease from baseline ICS use in both treatment and control groups with no intergroup differences at the end of the treatment period.
Finally, we assessed evidence for bronchial provocation to be very low in quality. Only four studies (Keles 2011; Marogna 2005; Pajno 2003; Stelmach 2009) contributed to this analysis. We were required to use standardised mean difference (SMD) analyses to combine PD20 and PC20 data, and we found that levels of heterogeneity and imprecision were high, as was risk of performance and detection bias, in two of the contributing studies.
Potential biases in the review process
We conducted this review in accordance with established Cochrane standards. Two review authors independently screened search results and resolved discrepancies by discussion and, if necessary, consultation with a third person. We did not restrict the search by language and as a result included four studies published in languages other than English (three in Spanish and one in Chinese). We attempted to contact study authors when it was not clear whether a study met our inclusion criteria. We may have missed some unpublished data as, owing to the large number of manufacturers, we did not search individual manufacturers' trial registers for possible included studies.
At least two review authors extracted all study characteristics and numerical data and resolved discrepancies through discussion. The same was true for risk of bias ratings. In a change to our protocol, and as a result of the large number of included studies (14 of which were abstracts), we did not attempt to contact study authors to clarify methodological and outcome information, relying instead on what was presented in the report.
We adapted the protocol in two other ways that may have introduced bias. First, we had not anticipated how many studies had recruited mixed populations of patients with rhinitis 'and/or' asthma. As outcomes for participants with asthma were rarely presented separately, we had to make a pragmatic decision as to whether or not to include these studies. We decided, after consultation with a third person, to include studies in which at least 80% of participants had received a diagnosis of asthma. If this was not clear from the report, we attempted to contact study authors to confirm this. We excluded these 'mixed population' studies from the sensitivity analysis for adverse events, although this exclusion did not substantially alter the outcome.
Second, we had not planned to extract outcome data for all adverse events, instead opting to include the more clinically important serious adverse events as a primary outcome. So few serious adverse events were reported that we decided to extract all adverse events additionally. Analysis of this additional post hoc outcome was the only analysis with enough data to allow exploratory analyses with subgroups, and so these results should be interpreted with caution.
None of the review authors have reported conflicting interests.
Agreements and disagreements with other studies or reviews
Several published systematic reviews have addressed the question of whether SLIT is effective and safe in asthma (Calamita 2006; Compalati 2009; Lin 2013; Penagos 2008; Tao 2014) and have reached somewhat conflicting conclusions. Tao 2014 reported findings from 16 double‐blind placebo‐controlled trials that randomly assigned 794 participants with asthma. Lin 2013 is a systematic review that reported on SLIT for allergic rhinoconjunctivitis and asthma but without a formal meta‐analysis. This review synthesised findings from 63 studies, including 5131 participants. Calamita 2006 included 25 studies that randomly assigned 1706 participants. Penagos 2008, a systematic review of SLIT for allergic asthma in children three to 18 years of age, included nine studies assessing 441 participants. Compalati 2009 reported the findings of nine double‐blind placebo‐controlled studies of SLIT for allergic asthma that assessed 452 study participants.
All four meta‐analyses used SMD to meta‐analyse composite asthma symptom scores. Compalati 2009, Penagos 2008 and Tao 2014 reported a statistically significant reduction in asthma symptoms favouring SLIT, but with a high level of heterogeneity (I2 ≥ 90%). Calamita 2006 reported a statistically significant 'general improvement' in asthma, but this conclusion appears to have been reached from a combined analysis of asthma symptoms, need for reliever medication, lung function tests and lung hyper‐reactivity. Study authors reported improvement in asthma symptoms alone when data were analysed using SMDs, but this finding did not reach statistical significance. Lin 2013 narratively reported improvement in asthma symptoms favouring SLIT in all placebo‐controlled studies included in the review and rated the strength of this evidence as 'high'.
Similarly, all four meta‐analyses reported composite asthma medication use scores. Compalati 2009, Penagos 2008 and Tao 2014 reported a statistically significant reduction in medication scores favouring SLIT, but again using an SMD analysis and with high heterogeneity (I2 ≥ 90%). Calamita 2006 found no statistically significant differences between groups in asthma symptom scores. Lin 2013 narratively reported benefit of SLIT over control in 40 out of 41 studies that reported medication use scores but did not present findings for asthma separately from those for rhinoconjunctivitis.
Calamita 2006, Tao 2014 and Penagos 2008 reported adverse events and, consistent with our findings, observed very few serious events. Tao 2014 concluded that participants receiving SLIT experience more adverse events overall than those receiving placebo, but that most of these events were considered to be mild in severity, again in keeping with our findings. Lin 2013 concluded that adverse events were insufficiently reported to allow further comment on the safety of SLIT.
None of the four meta‐analyses included quality of life or exacerbations as a prespecified primary or secondary outcome. Lin 2013 found that validated disease‐specific quality of life was reported in only eight of the 63 studies included in the review; half reported a statistically significant benefit of SLIT over control, but none of these eight studies met our inclusion criteria.
In contrast to the meta‐analyses described above, and as per our protocol, we chose not to combine different, unvalidated symptom and medication scores in a meta‐analysis. We believe that heterogeneity across measurements would lead to a potentially misleading outcome. As in Lin 2013, we reported these findings only narratively.
In Nieto 2009, review authors evaluated five meta‐analyses of SLIT for respiratory disease and recommended that as a result of discrepancies, inconsistencies and lack of robustness in the included meta‐analyses, evidence at that time did not support its use. Similarly, Incorvaia 2010 presented an overview on the position of SLIT for treatment of allergic asthma and called for additional research to resolve conflicting results.
Risk of bias summary: review authors' judgements about each risk of bias item for each included study.
Forest plot of comparison: 1 SLIT vs control, outcome: 1.3 Serious adverse events.
Forest plot of comparison: 1 SLIT vs control, outcome: 1.5 All adverse events.
Cates plot illustrating all adverse events (created at: www.nntonline.net)
Comparison 1 SLIT vs control, Outcome 1 Exacerbation requiring ED or hospital visit.
| Study | Outcome name | Scoring | Data type | SLIT | Control |
| Bousquet 1999 | Short‐Form Health Status Survey; physical pain | 22 items divided into 7 scales measuring physical functioning, limitations in role functioning due to physical health problems,social functioning, general mental health, general health perception, physical pain and vitality. Each scale is 0 to 100 with lower score for poorer health. Measured at 25 months. | Means, no variance | 86.2 (n=18) | 68.3 (n=20) |
| Bousquet 1999 | Short‐Form Health Status Survey; general mental health domain | 22 items divided into 7 scales measuring physical functioning, limitations in role functioning due to physical health problems,social functioning, general mental health, general health perception, physical pain and vitality. Each scale is 0 to 100 with lower score for poorer health. Measured at 25 months. | Means, no variance | 79.7 (n=18) | 60.7 (n=20) |
| Bousquet 1999 | Short‐Form Health Status Survey; general perception of health domain | 22 items divided into 7 scales measuring physical functioning, limitations in role functioning due to physical health problems,social functioning, general mental health, general health perception, physical pain and vitality. Each scale is 0 to 100 with lower score for poorer health. Measured at 25 months. | Means, no variance | 76.5 (n=18) | 56.8 (n=20) |
| Lewith 2002 | Diary quality of life assessment | Proportion of days in each of the assessment periods | Means (SD) | 0.090 (‐0.096 to 0.150) | 0.117 (‐0.096 to 0.150) |
Comparison 1 SLIT vs control, Outcome 2 Quality of life.
Comparison 1 SLIT vs control, Outcome 3 Serious adverse events.
Comparison 1 SLIT vs control, Outcome 4 Exacerbation requiring OCS.
Comparison 1 SLIT vs control, Outcome 5 All adverse events.
Comparison 1 SLIT vs control, Outcome 6 Bronchial provocation.
Comparison 1 SLIT vs control, Outcome 7 ICS use.
| Study | Outcome name | Scoring | Data type | SLIT | Control |
| Alvarez‐Cuesta 2007 | Bronchial symptom scores during cat exposure | 0 (absent) to 3 (severe), multiple measurements | Mean area under the curve (CI) | 45.74 (10.8 to 80.67) n=17 | 143.44 (61.98 to 224.9) n=16 |
| Bousquet 1999 | Daytime asthma score | 0 (no symptoms) to 3 (severe symptoms) | Mean change (SD) | 0.17 (0.5) n=32 | 0.19 (0.44) n=33 |
| Bousquet 1999 | Nighttime asthma score | 0 (no symptoms) to 3 (severe symptoms) | Mean change (SD) | 0.17 (0.51) n=32 | 0.11 (0.35) n=33 |
| Caffarelli 2000 | Bronchial symptom score | 0 (no symptoms) to 3 (severe symptoms), weekly mean of daily ratings during pollen season | Weekly mean (SD) | 2.4 (2.7) n=24 | 4.6 (3.5) n=20 |
| Cooper 1984 | Days with asthma symptoms | Number of days during pollen season (max 70) | Means, no variance | 34.3, n=11 | 40.3, n=8 |
| Cooper 1984 | Asthma symptom severity score | 0 (none) to 3 (severe) | Means, no variance | 40.5, n=11 | 58.2, n=8 |
| Dahl 2006 | Asthma symptom score (before pollen season) | 0 (no symptoms) to 3 (severe symptoms), rated daily | Mean (SD) | 0.23 (0.34) n=73 | 0.33 (0.33) n=40 |
| Dahl 2006 | Asthma symptom score (during pollen season) | 0 (no symptoms) to 3 (severe symptoms), rated daily | Mean (SD) | 0.44 (0.68) n=68 | 0.74 (0.92) n=39 |
| Dahl 2006 | Percentage well days | Defined post hoc as a day during the pollen season with a symptom score 2 or less and no rescue medication required | Mean (SD) | 58.9 (27.6) n=61 | 38.2 (32.9) n=32 |
| Eifan 2009 | Visual analogue score for asthma/rhinitis symptoms | 0 cm (no symptoms) to 10 cm (highest level of symptoms) | Mean (SD) | 2.7 (2.1) n=15 | 4.6 (1.5) n=14 |
| Eifan 2009 | Asthma symptom score | 0 (no symptoms) to 3 (severe symptoms), rated daily | Mean (SD) | 0.2 (0.4) n=15 | 2.5 (1.6) n=14 |
| Ippoliti 2003 | Asthma symptom score | 0 (no symptoms) to 3 (severe symptoms), mean of daily ratings throughout 6 months of therapy | Means, no variance | 1.28, n=47 | 3.15, n=39 |
| Lewith 2002 | Visual analogue scale, asthma severity | Higher scores indicate more severe asthma | Mean (SE), read from graph | 2.44 (0.32) n=101 | 2.62 (0.31) n=101 |
| Lewith 2002 | Number of asthma symptoms | Unclear | Mean (SE), read from graph | 0.99 (0.14) n=101 | 1.14 (0.15) n=101 |
| Lue 2006 | Daytime asthma symptom score | 0 (no symptoms) to 3 (severe symptoms), rated daily | Mean (SD) | 0.13 (0.19) n=10 | 0.49 (0.38) n=10 |
| Lue 2006 | Nighttime asthma symptom score | 0 (no symptoms) to 3 (severe symptoms), rated daily | Mean (SD) | 0.16 (0.15) n=10 | 0.50 (0.47) n=10 |
| Marogna 2005 | Composite asthma symptom score | Monthly individual symptom ratings 0 (absent) to 3 (severe) combined | Mean (SEM), read from graph | 50 (15) n=29 | 150 (25) n=23 |
| Mungan 1999 | Asthma symptom score | 0 (no symptoms) to 3 (severe symptoms), rated daily during second 6 months of treatment | Means, no variance | 0.41, n=15 | 0.88, n=11 |
| Niu 2006 | Daily asthma symptom score | Combined daytime and nighttime score, each rated 0 (no symptoms) to 3 (severe symptoms) | Means and p‐values for within group change | ‐0.07 (p=0.108) n=49 | 0.01 (p=0.998) n=48 |
| Pajno 2000 | Nighttime asthma symptom score | Number per month during last year of treatment | Means (p<0.0001 for difference between groups) | 6, n=12 | 13.2, n=9 |
| Pham‐Thi 2007 | % asthma‐free days | Number of days when day and nighttime score was 0 (no symptoms) | Mean (SD) | 85.8 (23.8) n=54 | 91.1 (15.4) n=55 |
| Pham‐Thi 2007 | Nighttime asthma score | 0 (no symptoms) to 3 (severe symptoms) | Mean (SD) | 0.10 (0.19) n=54 | 0.07 (0.16) n=55 |
| Pham‐Thi 2007 | Daytime asthma score | 0 (no symptoms) to 3 (severe symptoms), mean of daily scores from past 3 weeks | Mean (SD) | 0.15 (0.26) n=54 | 0.08 (0.17) n=55 |
| Reilly 1994 | Visual analogue scale for asthma symptoms | Minimum=fine, maximum=terrible (measured in mm) | Mean change (SEM) | ‐7.2 (3.2) n=11 | 7.8 (3.0) n=13 |
| Stelmach 2009 | Asthma symptom score (second pollen season) | As for first pollen season | Mean weekly score (SD) | 7.15 (5.43) n=20 | 11.99 (7.32) n=15 |
| Stelmach 2009 | Asthma symptom score (first pollen season) | Day, night and beta‐agonist use rated 0 to 3 and combined 0 (no symptoms and no use of b‐agonists use) to 9 (severe symptoms during day and night, and > 3 | Mean weekly score (SD) | 18.07 (11.58) n=20 | 16.13 (9.34) n=15 |
| Zheng 2012 | Cough/asthma symptom score | 0 (no symptoms) to 3 (severe symptoms); assessed for both night and day | Mean decrease in score after 25 weeks treatment | 3.3 (2.1) n=53 | 1.3 (2.1) n=53 |
Comparison 1 SLIT vs control, Outcome 8 Unvalidated asthma symptom scores.
| Study | Outcome name | Scoring | Data type | SLIT | Control |
| Bousquet 1999 | Inhaled corticosteroid use | mcg beclomethasone/day | Mean (SD) | 348 (410) n=32 | 308 (408) n=33 |
| Dahl 2006 | Asthma medication score (during season) | Average daily composite score of beta2‐agnoist, ICS use and OCS use; maximum daily score 16 | Daily mean (SD) | 0.71 (1.28) n=68 | 0.66 (1.08) n=39 |
| Dahl 2006 | Asthma medication score (before season) | Average daily composite score of beta2‐agnoist, ICS use and OCS use; maximum daily score 16 | Daily mean (SD) | 0.09 (0.23) n=73 | 0.09 (0.14) n=40 |
| Eifan 2009 | Total medication score | 1 point: beta2‐agnoists and antihistamines; 2 points: inhaled/intranasal steroids 3 points: one tablet of corticosteroid | Mean (SD) | 1.2 (0.9) n=15 | 2.8 (1.1) n=14 |
| Lewith 2002 | Short acting bronchodilator use | Puffs/week | Mean (SD), read from graph | 3.35 (0.48) n=101 | 3.4 (0.5) n=101 |
| Lue 2006 | Medication score | Mean daily use of corticosteroids, beta2‐agnoist, antihistamines ‐ scoring unclear | Mean (SD) | 1.0 (0.94) n=10 | 1.1 (1.15) n=10 |
| Marogna 2005 | Salbutamol use | Puffs/month at end of treatment | Mean (SD), read from graph | 2 (0.5) n=29 | 11.5 (1) n=23 |
| Mungan 1999 | Medication scores (second 6 months of treatment) | ICS, beta2‐agnoists and antihistamines scored 1 to 4 depending on dose and/or frequency (maximum score 12) | Means, no variance | 1.97, n=15 | 5.24, n=11 |
| NCT00633919 | Average Daily Asthma Medication Score During a 2‐months Evaluation Period Autumn 2008 (later time point) | 1 to 2 inhalations twice daily of salbutamol (200 mcg per inhalation) = 2 scores; 1 to 2 inhalation twice daily of budesonide/formoterol 80 (4.5 mcg per inhalation) = 4 scores; 1 inhalation twice daily of budesonide/formoterol 160 (4.5 mcg per inhalation) = 8 scores; up to 10 tablets once daily of prednisone (5 mg) = 1.6 scores. The total maximum daily scores were 40 | Mean (SD) | 4.4 (5.9) n=36 | 4.7 (5.4) (n=)39 |
| Niu 2006 | Short acting broncodilator use | Puffs/day | Mean change (SD) | ‐0.04 (0.32) n=49 | 0.02 (0.27) n=48 |
| Niu 2006 | Oral corticosteroid use | Tablets/day | Mean change (SD) | ‐0.08 (0.42) n=49 | 0 (0.27) n=48 |
| Niu 2006 | Inhaled corticosteroid use | Puffs/day | Mean change (SD) | ‐0.23 (0.67) n=49 | ‐0.1 (1.08) n=48 |
| Pajno 2000 | Total medication score (end of treatment) | 1: bronchodilators; 2: ICS; 4: | Means (SD imputed from p‐value) | 82.68 (55) n=12 | 205.2 (55) n=9 |
| Pham‐Thi 2007 | Inhaled corticosteroid use | mcg budesonide/day | Mean (SD) | 257 (232) n=54 | 223 (270) n=55 |
| Pham‐Thi 2007 | Short acting bronchodilator use | Puffs/day | Mean (SD) | 0.55 (0.6) n=54 | 0.47 (0.5) n=55 |
| Stelmach 2009 | Medication score (second pollen season) | Mean weekly medication score during second pollen season, adjusted for pollen concentration | Mean (SD) | 6.22 (2.45) n=20 | 7.37 (2.7) n=15 |
| Stelmach 2009 | Medication score (first pollen season) | Mean weekly medication score during first pollen season, adjusted for pollen concentration | Mean (SD) | 5.1 (1.77) n=20 | 7.48 (2.78) n=15 |
Comparison 1 SLIT vs control, Outcome 9 Unvalidated medication use scores.
Comparison 2 Subgroup and sensitivity analyses, Outcome 1 Adverse events by age.
Comparison 2 Subgroup and sensitivity analyses, Outcome 2 Adverse events by allergen.
Comparison 2 Subgroup and sensitivity analyses, Outcome 3 Adverse events by study duration.
Comparison 2 Subgroup and sensitivity analyses, Outcome 4 Adverse events (sensitivity for risk of bias: blinded studies only).
Comparison 2 Subgroup and sensitivity analyses, Outcome 5 Adverse events (sensitivity analysis removing studies with mixed population of asthma and rhinitis).
| Subgroup and sensitivity analyses compared with placebo for asthma | ||||||
| Patient or population: adults and children with asthma Weight mean duration of all included studies: 54 weeks (Fadel 2010 and Rodriguez 2012 excluded, as duration not reported) | ||||||
| Outcomes | Illustrative comparative risks* (95% CI) | Relative effect | Number of participants | Quality of the evidence | Comments | |
| Assumed risk | Corresponding risk | |||||
| Control | SLIT | |||||
| Exacerbation requiring ED or hospital visit Study duration: | No events | No events | Not estimable | 47 | ⊕⊕⊝⊝ | |
| Quality of life | No meta‐analysis possible | Not applicable | ‐ | (0 RCTs) | Not applicable | 5 studies reported quality of life outcomes but we were not able to perform a meta‐analysis |
| Serious adverse events Weighted mean duration of studies: 49 weeks | 14 per 1000 | 12 per 1000 (0 to 24) | RD 0.0012, (‐0.0077 to 0.0102) | 2560 (22 RCTs) | ⊕⊕⊕⊝ | |
| Exacerbation requiring OCS Weighted mean duration of studies: 25 weeks | No events | No events | Not estimable | 77 | ⊕⊕⊝⊝ | |
| All adverse events Weighted mean duration of studies: 60 weeks** | 222 per 1000 | 327 per 1000 | OR 1.70 (1.21 to 2.38) | 1755 | ⊕⊕⊝⊝ | |
| Bronchial provocation | Mean bronchial provocation in control group was 1020 mcg (PD20) and 5.45 mg/mL (PC20) | Mean bronchial provocation in intervention group was 0.69 standard deviations higher (0.04 lower to 1.43 higher) | ‐ | 139 | ⊕⊝⊝⊝ | 3 studies reported outcome as PC20 and 1 as PD20. We combined the different scales using standardised mean differences |
| ICS use | Mean ICS use in control group was 255 mcgl | Mean ICS use in intervention group was 35.1 higher (‐50.21 to 120.42) | ‐ | 174 | ⊕⊕⊝⊝ | Both treatment and control groups in both studies included in this analysis showed significantly decreased ICS use at end of study compared with baseline but no intergroup difference was detected |
| *The basis for the assumed risk (e.g. median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). **All adverse events was not a prespecified outcome, but we have included it in the 'Summary of findings' table, as substantial data were contributed to this outcome. We have left out the asthma symptom scores outcome, as we were able to perform only a limited narrative analysis. | ||||||
| GRADE Working Group grades of evidence. | ||||||
| aOnly a small number of included studies reported this outcome, suggesting lack of relevance in this study population. Treatment period in Calderon 2006 was just 4 weeks and exacerbations requiring ED/hospital admission/OCS are rare events. Downgrade once. bNo events but could be a product of the asthma severity of the recruited population. No downgrade. cFunnel plot not possible as no one outcome shows > 10 studies contributing events. Many reports are conference abstracts without associated peer‐reviewed full publication. Downgrade once. d5/21 studies included in this analysis were assessed as having high risk of performance and detection bias, but none of the 5 contributing events. No other serious issues. e5/21 studies included a mixed population of participants with asthma and rhinitis (but all > 80% with asthma), but the 5 studies contributing events to this analysis recruited exclusively participants with asthma. fEvents rare. Participants had largely mild to moderate asthma and may have been less at risk of serious adverse events. Downgraded once for indirectness. gTwo studies contributing events assessed as having high risk of performance and detection bias, with 3 others at high risk but not contributing events. Study contributing greatest weight (41%) to the analysis reported only as a conference abstract with uncertainty about attrition bias. Downgrade once. hSix out of 19 studies included mixed rhinitis and asthma populations, and of those contributing events made up approximately 25% of the analysis weight. Most of these events were mild and transient and did not lead to participant withdrawal. Downgrade once. iTwo out of four (contributing > 50% of analysis weight) studies assessed at high risk of performance and detection bias. Downgrade once. jHigh level of heterogeneity (I2 = 76%) and combines PC20 with PD20 scores using SMDs. Downgrade once. kPossibility of benefit in control group not excluded by confidence intervals. Downgrade once. lCalculated as the weighted mean of control group scores of the included studies. mImprecise estimate with confidence intervals including the possibility of a clinically important harm or benefit from SLIT. Downgrade once. nMany participants in included studies had mild asthma and so would be less likely to be using ICS. This was a predefined outcome, which may have less relevance to the study population.. Downgrade once. | ||||||
| Study ID | Total N | Allergen | Comparator | Age range | Country | Duration | % with asthma |
| 120 | HDM | Placebo | Not reported | Cuba | 52 weeks | 100 | |
| 50 | Cat dander | Placebo | 14‐55 | Spain | 52 weeks | 81.8 | |
| 15 | HDM | Placebo | 7‐18 | Turkey | 26 weeks* | 100 | |
| 85 | HDM | Placebo | 7‐42 | France | 108 weeks* | 100 | |
| 48 | Grass pollen | Placebo | 4‐14 | Italy | 13 weeks* | 89.6 | |
| 43 | Grass pollen | Placebo | 18‐65 | Unclear | 4 weeks* | 100 | |
| 71 | HDM | Placebo | 18‐65 | UK and Denmark | 4 weeks | 100 | |
| 72 | HDM | Placebo | 5‐14 | Spain | 4 weeks | 100 | |
| 19 | Grass pollen | Placebo | 5‐15 | UK | > 8 but < 16 weeks* | 100 | |
| 44 | Alternaria | Pharmacotherapy | 18‐65 | Spain | 52 weeks | 100 | |
| 114 | Timothy grass | Placebo | 18‐65 | Denmark and Sweden | 19.5 weeks | 100 | |
| 48 | HDM | Pharmacotherapy | 5‐10 | Turkey | 52 weeks | 85 | |
| 55 | Grass pollen | Placebo | 18‐50 | Syria | Not reported | 100 | |
| 60 | HDM | Placebo | 13‐45 | Mexico | 26 weeks | 100 | |
| 60 | HDM | Placebo | Not reported | Not reported | 13 weeks | 100 | |
| 32 | HDM | Placebo | Not reported | Turkey | 52 weeks | 100 | |
| 86 | HDM | Placebo | 5‐12 | Italy | 26 weeks* | 100 | |
| 31 | HDM | Pharmacotherapy | 'Children' | Unclear | 156 weeks | 100 | |
| 53 | HDM | Pharmacotherapy | Not reported | Unclear | 17.3 weeks | 100 | |
| 58 | HDM | Pharmacotherapy | 5‐12 | Turkey | 52 weeks* | 100 | |
| 56 | HDM/Parietaria | Pharmacotherapy | 7‐68 | Italy | 52 weeks | 100 | |
| 18 | Artemisia pollen | Placebo | 15‐56 | Unclear | 7.14 weeks* | 100 | |
| 242 | Homeopathic HDM | Placebo | 18‐55 | UK | 16 weeks | 100 | |
| 20 | HDM | Placebo | 6‐12 | Taiwan | 24 weeks* | 100 | |
| 24 | HDM | Placebo | 4‐16 | Italy | 52 weeks | 84.6 | |
| 79 | Birch pollen | Pharmacotherapy | 18‐65 | Italy | 156 weeks* | 100 | |
| 604 | HDM | Placebo | 14+ | Denmark, Germany, Italy, Spain, United Kingdom, Sweden, France, Poland | 52 weeks | 100 | |
| 116 | Ultra‐rush birch pollen | Placebo | 6‐14 | Germany | 0.015 weeks | 100 | |
| 36 | HDM | Placebo | 18‐46 | Turkey | 52 weeks | 88 | |
| 28 | HDM | Placebo | 4‐9 | Italy | 52 weeks | 100 | |
| 124 | HDM | Placebo | 18‐65 | Spain | 104 weeks | 100 | |
| 110 | HDM | Placebo | 6‐12 | Taiwan | 24 weeks* | 100 | |
| 47 | HDM | Pharmacotherapy | Not reported | Italy | 156 weeks | 100 | |
| 24 | HDM | Placebo | 8‐15 | Italy | 104 weeks | 100 | |
| 30 | Parietaria | Placebo | 8‐14 | Italy | 56 weeks* | 100 | |
| 111 | HDM | Placebo | 5‐16 | France | 78 weeks | 100 | |
| 106 | HDM | Placebo | 5‐13 | Romania | 26 weeks | 100 | |
| 28 | Homeopathic HDM/feathers/ mixed moulds | Placebo | 16+ | Scotland | 4 weeks* | 100 | |
| 40 | HDM | Placebo | 'Adults' | Cuba | Not reported | 100 | |
| 50 | HDM | Pharmacotherapy | 6‐15 | Cuba | 104 weeks | 100 | |
| 264 | HDM | Pharmacotherapy | 3‐13 | China | 52 weeks | 82 | |
| 50 | Grass pollen | Placebo | 6‐17 | Poland | 104 weeks | 100 | |
| 60 | HDM | Placebo | 4‐18 | China | 48 weeks | 100 | |
| 24 | Birch pollen | Placebo | Not reported | Unclear | 104 weeks | 100 | |
| 834 | HDM | Plaecbo | Not reported | Austria, Croatia, Denmark, France, Germany, Lithuaina, Netherlands, Poland, Serbia, Slovakia, Spain, United Kingdom | 78 weeks | 100 | |
| 66 | Olive pollen | Placebo | 7‐17 | Greece | 104 weeks | 90.6 | |
| 484 | HDM | Placebo | 16‐50 | China | 52 weeks* | 100 | |
| 89 | Greer German cockroach | Placebo | 5‐17 | USA and UK | 13 weeks | 80 | |
| 32 | HDM | Placebo | Not reported | Turkey | 52 weeks | 100 | |
| 63 | HDM | Placebo | 'Adults' | France | 1.4 weeks | 100 | |
| 128 | HDM | Pharmacotherapy | 4‐14 | Taiwan | 104 weeks | 100 | |
| 106 | HDM | Pharmacotherapy | 4‐14 | China | Outcomes reported at 25 weeks | 100 | |
| *Studies that included post‐treatment follow‐up periods. | |||||||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Exacerbation requiring ED or hospital visit Show forest plot | 1 | 47 | Odds Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 2 Quality of life Show forest plot | Other data | No numeric data | ||
| 3 Serious adverse events Show forest plot | 22 | 2560 | Risk Difference (M‐H, Random, 95% CI) | 0.00 [‐0.01, 0.01] |
| 4 Exacerbation requiring OCS Show forest plot | 2 | 77 | Odds Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 5 All adverse events Show forest plot | 19 | 1755 | Odds Ratio (M‐H, Random, 95% CI) | 1.70 [1.21, 2.38] |
| 6 Bronchial provocation Show forest plot | 4 | 139 | Std. Mean Difference (IV, Random, 95% CI) | 0.69 [‐0.04, 1.43] |
| 6.1 PD20 | 1 | 52 | Std. Mean Difference (IV, Random, 95% CI) | 1.46 [0.84, 2.08] |
| 6.2 PC20 | 3 | 87 | Std. Mean Difference (IV, Random, 95% CI) | 0.40 [‐0.26, 1.05] |
| 7 ICS use Show forest plot | 2 | 174 | Mean Difference (IV, Random, 95% CI) | 35.10 [‐50.21, 120.42] |
| 8 Unvalidated asthma symptom scores Show forest plot | Other data | No numeric data | ||
| 9 Unvalidated medication use scores Show forest plot | Other data | No numeric data | ||
| Outcome or subgroup title | No. of studies | No. of participants | Statistical method | Effect size |
| 1 Adverse events by age Show forest plot | 19 | 1755 | Odds Ratio (M‐H, Random, 95% CI) | 1.70 [1.21, 2.38] |
| 1.1 Children (mean age < 18 years) | 8 | 626 | Odds Ratio (M‐H, Random, 95% CI) | 2.13 [0.83, 5.47] |
| 1.2 Teenagers and adults (mean age > 18 years) | 8 | 964 | Odds Ratio (M‐H, Random, 95% CI) | 1.48 [1.06, 2.06] |
| 1.3 Mixed age study population or age range not specified | 3 | 165 | Odds Ratio (M‐H, Random, 95% CI) | 2.06 [0.47, 9.05] |
| 2 Adverse events by allergen Show forest plot | 18 | 1726 | Odds Ratio (M‐H, Random, 95% CI) | 1.70 [1.21, 2.38] |
| 2.1 HDM SLIT | 10 | 1386 | Odds Ratio (M‐H, Random, 95% CI) | 1.47 [1.10, 1.97] |
| 2.2 Pollen SLIT | 6 | 251 | Odds Ratio (M‐H, Random, 95% CI) | 5.48 [1.99, 15.05] |
| 2.3 Other/mixed allergens | 2 | 89 | Odds Ratio (M‐H, Random, 95% CI) | 0.0 [0.0, 0.0] |
| 3 Adverse events by study duration Show forest plot | 19 | 1815 | Odds Ratio (M‐H, Random, 95% CI) | 1.70 [1.21, 2.38] |
| 3.1 Duration less than 52 weeks | 7 | 427 | Odds Ratio (M‐H, Random, 95% CI) | 1.53 [0.38, 6.19] |
| 3.2 Duration 52 weeks and longer | 12 | 1388 | Odds Ratio (M‐H, Random, 95% CI) | 1.77 [1.22, 2.58] |
| 4 Adverse events (sensitivity for risk of bias: blinded studies only) Show forest plot | 14 | 1329 | Odds Ratio (M‐H, Random, 95% CI) | 1.47 [1.10, 1.96] |
| 5 Adverse events (sensitivity analysis removing studies with mixed population of asthma and rhinitis) Show forest plot | 13 | 1293 | Odds Ratio (M‐H, Random, 95% CI) | 1.42 [1.06, 1.91] |
