House dust mite control measures for asthma: systematic review

Authors


Dr Peter C Gøtzsche, MD, DrMedSci, MSc
The Nordic Cochrane Centre
Rigshospitalet, Department 3343
Blegdamsvej 9
DK-2100 Copenhagen Ø
Denmark

Abstract

The major allergen in house dust comes from mites. We performed a systematic review of the randomized trials that had assessed the effects of reducing exposure to house dust mite antigens in the homes of people with mite-sensitive asthma, and had compared active interventions with placebo or no treatment. Fifty-four trials (3002 patients) were included. Thirty-six trials assessed physical methods (26 mattress covers), 10 chemical methods and eight a combination of chemical and physical methods. Despite the fact that many trials were of poor quality and would be expected to exaggerate the reported effect, we did not find an effect of the interventions. For the most frequently reported outcome, peak flow in the morning (1565 patients), the standardized mean difference was 0.00 (95% confidence interval (CI) −0.10 to 0.10). There were no statistically significant differences in number of patients improved (relative risk 1.01, 95% CI 0.80–1.27), asthma symptom scores (standardized mean difference −0.04, 95% CI −0.15 to 0.07) or in medication usage (standardized mean difference −0.06, 95% CI −0.18 to 0.07). Chemical and physical methods aimed at reducing exposure to house dust mite allergens cannot be recommended.

Exposure to different allergens can trigger asthma attacks in sensitized individuals. House dust is a mixture containing many different allergens, but the major allergen is derived from mites, especially the species Dermatophagoides pteronyssinus and Dermatophagoides farinae. A common site for house dust mites is the bed, where pillows, quilts and mattresses often serve as reservoirs for the allergen. Carpets and upholstered furniture may also contain high mite levels (1, 2). It appears very reasonable, and is usually recommended, that environmental control of allergens, although difficult, should be an integral part of the overall management of sensitized patients. However, some of the evidence behind these recommendations is derived from observational studies, including some in which patients were moved to high altitudes or hospitals, whereupon their symptoms improved (3). These measures are not feasible for most patients, and it is not clear whether the allergen levels that can be obtained in the patients’ homes are large enough to lead to improvements in the asthma.

Different methods for reducing mite exposure have been tried, for example chemical methods, physical methods and combinations of these (1). We published a systematic review of these methods in 1998 (4, 5) and the current review is the most recent update, which has also been published in The Cochrane Library (6).

Methods

We studied whether patients with asthma who are sensitized to house dust mites benefit from measures designed to reduce their exposure to mite antigen in the home. We included randomized trials and accepted reports in any language. As some mite control measures are impossible to blind, we included nonblinded trials. The participants were all diagnosed as having bronchial asthma by a physician and their mite sensitization was assessed by skin testing, bronchial provocation tests or serum assays for specific immunoglobulin E (IgE) antibodies.

The active interventions were (i) chemical (acaricides), (ii) physical (e.g. mattress covers, vacuum-cleaning, heating, ventilation, freezing, washing, air-filtration and ionizers) and (iii) combinations of these. The control group received placebo or no treatment.

If available, we extracted data for the following outcome measures: subjective well-being, asthma symptom scores, medication usage, days of sick-leave from school or work, number of unscheduled visits to a physician or a hospital, forced expiratory volume in 1 s (FEV1), peak expiratory flow rate (PEFR) and provocative concentration that causes a 20% fall in FEV1 (PC20).

We searched PubMed from 1966 with the terms mite* AND asthma*, combined with one or more of the following: random* OR control* OR blind* (last search November 2007). The Cochrane Library was searched with the terms mite* AND asthma* (last search November 2007). We also checked the bibliography of each trial report for additional references.

We selected the trials for inclusion independently. Assessment of the risk of bias and extraction of data was primarily performed by one author (PCG) and checked by the other (HKJ) for the current version of the review. All assessments were open. When it was not stated at what time of the day the peak flow had been recorded, we assumed it was in the morning. Any ambiguities were resolved by discussion. When necessary, we contacted the trial authors for clarification.

Statistical methods

We calculated 95% confidence intervals (CI) with a fixed effect model using Review Manager (7). Heterogeneity was tested with the chi-squared test and its magnitude was assessed as I2, which gives the amount of between-trial variation in relation to the total variation (8). If there was heterogeneity (< 0.10), the reasons were explored.

As the results from crossover trials were usually reported by the authors in summary form, as if they had come from a group comparative trial, we analysed these data accordingly, assuming that no important carry-over effects had occurred (paired data were only available for some of the crossover trials and not for all the recorded variables).

When continuous data were presented on different scales, e.g. PEFR as absolute values or as per cent of predicted values, we used the standardized mean difference (difference in effect divided by the pooled standard deviation of the measurements). By that transformation, the effect measures become dimensionless and results from different scales may therefore often be combined. Data on well-being and asthma symptom scores were reported in different ways, but as these two outcome measures were closely related or even equivalent, we summarized categorical data in the well-being category (number of patients who improved) and continuous data (which mostly concerned asthma symptoms) in the asthma symptoms score category. Data on PC20 were usually analysed after logarithmic transformation, as they were highly skew. We analysed the data accordingly, and when the authors had converted their means and standard deviations from the logarithmic scale to the arithmetic scale, we converted them back again (9). We excluded PC20 data that had not been analysed after logarithmic transformation.

When several options were available for use of medication, we chose bronchodilators. When data were recorded at several points in time, we used the longest observation period during which the patients were still on randomized treatment, unless performance bias occurred, for example by a planned reduction in dose of inhaled steroids.

We did not adjust for baseline differences, as imbalances occurring despite the randomization would be expected to equal each other out in a large sample of trials. Second, baseline recordings were not always available. Third, if we had made adjustments when possible, we would have risked biasing the review, as investigators are inclined to show baseline differences and adjust for them when this procedure favours the experimental treatment (10). It has also been shown that bias occurring during data analysis is very common and almost without exception favours the new treatment over the control treatment (11).

To avoid double-counting of the control group when there was more than one active group in a trial, or vice versa, we pooled the groups when feasible. In one case where this was not possible, we split the number of patients instead.

Description of studies

We included 54 trials (3002 patients) in the review [70 publications (12–81), as some trials were published more than once]. The main characteristics of the trials are shown in Table 1. Additional details on methods, interventions and data extraction appear in a more extensive table in the Cochrane review (6), which also lists 47 excluded studies and the reasons for exclusion (most were excluded because they were not randomized or did not contain any clinical data).

Table 1.   Characteristics of the included studies
Study (reference)MethodsParticipants (mean age or range shown)Interventions (in the active groups, only the most important ones mentioned)Outcomes
  1. Additional details on methods, interventions and data extraction appear in the Cochrane review (6).

  2. FEV1, forced expiratory volume in 1 s; IgE, immunoglobulin E; NA, not available; PC20, provocative concentration that causes a 20% fall in FEV1; PEFR, peak expiratory flow rate.

Antonicelli 1991 (12)Crossover
Randomisation method: NA
Not blind (apart from PD20)
n = 9 (9 in analyses)
16 years
Test: HEPA-filter
Control: none
8 weeks
Symptom score, medication score, FEV1, PEFR, PD20
Additional data from author
Bahir 1997 (13)Randomisation method: NA
Double-blind
n = 40 (30 in analyses)
6–17 years
Test: acaricide
Control: placebo
6 months
Symptom score, use of beta-2 agonists, FEV1, PEFR, Acarex test
Fig. 3 indicates SEM which must be an error
Burr 1976 (14)Crossover
Randomisation method: NA
Not blind
n = 32, also in analyses
33 years
Test: vacuum cleaning, laundering, mattress cover
Control: no intervention
6 weeks
Medication use past 24 hours, PEFR
Burr 1980 (15)Randomisation method: NA
Blind assessment
n = 55 (53 in analyses)
5–14 years
Test: nurse visit, extensive scheme with vacuum cleaning, laundering
Control: nurse visit, placebo, mainly removal of dust
8 weeks
Numbers improved, PEFR
Burr 1980 (16)Crossover
Randomisation method: NA
Not blind
n = 21 from trial Burr 1980 (15) who still had symptomsTest: mattress cover and other measures
Control: as in Burr 1980 (15)
1 month
Preference for period, PEFR
Carswell 1996 (17–20)Randomisation method: NA
Double-blind
n = 70 (49 in analyses)
9.9 years
Test: acaricide, vacuum cleaner, allergen exclusion covers
Control: cotton placebo covers, etc
24 weeks
Numbers improved, symptoms, medication, PEFR, FEV1, PC20
Chang 1996 (21)Randomisation method: NA
Not blind
n = 26 (11 children and 15 adults, 26 in analyses)Test: acaricide
Control: no acaricide
3 months
Symptoms, medication, FEV1, PEFR, PC20
Charpin 1990 (22)Randomisation method: table provided by laboratory
Double-blind
n = 42 (11 had rhinitis)
Numbers in analyses not clear
27 years
Test: acaricide
Control: no acaricide
3 months
Global assessment by patient and doctor, PEFR, number of attacks
Chen 1996 (23)Randomisation method: NA
Double-blind
n = 56 (35 in analyses)
5–14 years
Test: mattress cover
Control: placebo mattresses cover
12 months
Symptoms, PEFR
Cinti 1996 (24)Randomisation method: sealed opaque envelopes
Double-blind
n = 20 (20 in analyses)
30 years
Test: mattress and pillow covers
Placebo: cotton covers
12 weeks
Symptom scores, number of acute episodes, medication, eosinophil cationic protein, PEFR
Additional data supplied by author
Cloosterman 1999 (25, 26)Randomisation method: open list of random numbers
Blind to patient and technician
n = 204 (157 in analyses)
33 years
Test: acaricide and mattress cover
Control: water and cotton cover
20 weeks
Symptoms, medication, FEV1, PEFR, PC20
de Vries 2007 (27)Randomisation method: open list
Double-blind
n = 143 (105 completed 2 years)
42 years
Test: mattress, duvet and pillow covers
Placebo: permeable covers
2 years
Symptoms, medication, PEFR
Dharmage 2006 (28)Randomisation method: toss of a coin
Double-blind
n = 32 (30 in analyses)
32 years
Test: mattress, doone and pillow covers
Placebo: permeable covers
6 months
Asthma symptoms, medication use, FEV1, PEFR, quality of life, time spent home, PD20
Dietemann 1993 (29–31)Randomisation method: NA
Double-blind
n = 26 (23 in analyses)
35 years
Test: acaricide
Control: placebo
1 year
Symptoms, medication score, FEV1, FVC, FEF25-75, PEFR, clinical score
Dorward 1988 (32)Randomisation method: NA
Blinded assessment
n = 21 (18 in analyses)
13–53 years
Test: extensive cleaning and removal of objects
Control: normal cleaning
8 weeks
Symptom score, salbutamol, PEFR, PC20, S-IgE
Ehnert 1992 (33)Randomisation method: NA
Test 1: double-blind
Test 2: not blind
n = 24, 8 in each group (21 in analyses)
7–15 years
Test 1: acaricide and vacuum cleaning
Test 2: mattress covers and tannic acid
Control: placebo foam
1 year
PC20
To avoid double-counting, we split the 7 patients in the control group into 4 patients for one comparison, and 3 for the other
Fang 2001 (34)Randomisation method: NA
Not blind
n = 43 (number randomised not clear)
37 years
Test: washing, sun and ventilation
Control: untreated
2 years
Symptoms, medication, PEFR
Frederick 1997 (35)Crossover
Randomisation method: NA
Single-blind
n = 31
5–15 years
Test: mattress, duvet and pillow covers
Control: placebo covers
3 months
Symptoms, medication, FEV1, PEFR, PC20
Geller-Bernst 1995 (36)Randomisation method: NA
Double blind
n = 32 (14 in most analyses)
4–12 years
Test: acaricide, change of bedsheet and blanket, vacuum cleaning
Control: placebo spray
6 months
Symptoms, medication, FEV1, PEFR, doctor’s and patient’s opinion, serum IgE
Ghazala 2004 (37)Crossover
Randomisation method: NA
Double blind
n = 17 (12 in analyses)
Age: NA.
Test: mattress covers
Control: cotton covers
9–11 weeks
Symptoms, medication
Gillies 1987 (38)Randomisation method: NA
Not blind
n = 26 (25 in analyses)
6–16 years
Test: mattress and pillow covers, synthetic bedding, vacuum cleaning
Control: no such measures
6 weeks
Symptoms, medication, PEFR PC20, serum IgE
Halken 2003 (39–41)Randomisation method: computer program
Described as double-blind but the covers were different
n = 60 (47 in analyses)
5–15 years
Test: mattress and pillow covers
Control: placebo covers
12 months
Medication, FEV1, PEFR, symptoms, PC20
Howarth 1992 (42)Randomisation method: NA
Double-blind
n = 35 (number in analyses not reported, some had rhinitis)
13–23 years
Test: covers of mattress, duvet and pillow
Control: placebo covers
6 weeks
Symptoms
Htut 2001 (43–46)Randomisation method: open table
Double-blind
n = 30 in trial report, n = 33 in previous abstract (23 in analyses)
18–45 years
Test 1: steam cleaning
Test 2: same, plus ventilation system installed
Control: sham steam cleaning
1 year
PD20
Huss 1992 (47)Randomisation method: NA
Not blind
n = 52 (52 in analyses)
18–75 years
Test: computer assisted instruction in physical interventions
Control: verbal and written guidance
12 weeks
Symptoms, medication, FEV1
Jooma 1995 (48)Randomisation method: open table
Not blind
n = 60 (numbers in analyses not stated)
6–14 years
Test 1: mattress and pillow covers and tannic acid
Test 2: acaricide
Control: none
6 months
PC20
Korsgaard 1983 (49)Randomisation method: NA
Not blind
n = 51 (46 in analyses)
30 years
Test: vacuum cleaning and washing
Control: none
12 weeks
PEFR, bronchodilator, symptoms
Kroidl 1998 (50)Randomisation method: sealed envelopes
Double-blind
n = 118 (78 in analyses)
8–50 years
Test: acaricide
Control: no acaricide
1 year
Well-being, PC20, RAST, changes in skin prick test
Drop-out data provided by author
Lee 2003 (51, 52)Randomisation method: coin tossing
Not blind
Conflicting information in two reports
n = 42 in analyses
Most above 30 years
Test: cotton covers, heating, sunlight
Control: no intervention
4 weeks
PEFR, symptoms
Luczynska 2003 (53)Randomisation method: open
Double-blind
n = 58, 45 started, 31 in analyses
18–54 years
Test: bed, blanket and pillow covers
Control: sham covers
1 year
PEFR, days with chest tightness, quality of life, asthma attacks and medication
Maesen 1977 (54)Crossover
Randomisation method: unclear
Double-blind
n = 30 (28 in analyses)
25 adults and 5 children
Test: air-filtration
Control: placebo
1 month
Subjective improvement, medication, PEFR
Manjra 1994 (55)Randomisation method: unclear
Not blind
n = 60 (59 in analyses)
5–12 years
Test 1: detergent
Test 2: detergent and acaricide
Control: none
3 months
PC20
Marks 1994 (56, 57)Randomisation method: NA
Blinded participants
n = 39 (35 in analyses)
13–60 years
Test: mattress, pillow and duvet covers, tannic acid and acaricide
Control: inactive placebo spray
6 months
Symptom score, FEV1, PEFR, PD20
Matthys 1996 (58)Crossover
Randomisation method: NA
Single-blind
n = 14 (10–14 in analyses)Test: air-dryer and water filter
Control: air-dryer without water filter
4 weeks
Medication, PEFR, symptoms
Mitchell 1980 (59)Crossover
Randomisation method: NA
Not blind
n = 10 (10 in analyses)
7–14 years
Test: electrostatic precipitator plus standard avoidance measures
Control: standard avoidance measures
2 weeks
Medication, PEFR
Popplewell 2000 (60)Randomisation method: NA
Not blind
n = 60 (51 in analyses)
21 children and 39 adults
Test: super vaccum cleaner
Control: standard vaccum cleaner
1 year
Medication, FEV1, PEFR, PC20
Reiser 1990 (61)Randomisation method: NA
Double-blind
n = 51 (46 in analyses)
5–16 years
Test: natamycin spray on mattresses
Control: placebo spray
3 months
Symptoms, medication, FEV1, PEFR, bronchial provocation test
Rijssenbeek 2002 (62, 63)Randomisation method: Zelen design
Double-blind
Unclear, two papers, both published in 2002, describe 38 and 27 patients, respectively
11–44 years
Test: mattress, pillow and bedding covers
Control: placebo covers
1 year
PEFR, FEV1, symptoms, medication, PC20, quality of life
Sette 1994 (64)Randomisation method: NA
Double-blind
n = 24 (24 in analyses)
13 years
Test: acaricide
Control: placebo
ca. 2 weeks
PC20, serum IgE
Shapiro 1999 (65)Randomisation method: sealed and opaque envelopes
Double-blind but intervention frequency is different
n = 44 (36 in analyses)
6–16 years
Test: mattress covers and tannic acid
Control: placebo tannic acid
1 year
FEV1, PEFR, symptoms, PD20, visits to emergency department and hospital, steroid courses
Author provided data on FEV1
Sheikh 2002 (66)Randomisation method: central
Double blind
n = 47 (43 in analyses)
5–14 years
Test: mattress covers
Control: placebo covers
6 months
PEFR, symptoms, night-time waking, medication, visits to doctor, emergency department and hospital, steroid courses
Sooltangos 1992 (67)Randomisation method: unclear
Not blind
n = 33 (no information on missing values)
34 years
Test: acaricide
Control: none
8 months
Symptoms, PEFR, FEV1, medication
Thiam 1999 (68)Randomisation method: NA
Not blind
n = 24 (24 in analyses)
6–14 years
Test 1: matress covers
Test 2: HEPA filters
Control: none
4 months
FEV1, PEFR, symptoms, bronchial provocation test
van den Bemt 2004 (69, 70)Randomisation method: NA
Double-blind
n = 52 (51 in some analyses)Test: mattress, duvet and pillow covers
Placebo: permeable covers
9 weeks
Symptoms, PEFR, medication
Data obtained from author on PEFR
van der Heide 1997 (71)Randomisation method: not clear whether randomised
Double-blind
n = 59 (40 in analyses)
31 years
Test: acaricide
Control: detergent
1 year
FEV1, PC20, serum total IgE
van der Heide 1997 (72)Randomisation method: computer
Double-blind
n = 30 (for relevant comparison; no information on missing values)
18–45 years
Test 1: air-cleaners
Test 2: air-cleaners and mattress and pillow covers
Control: placebo for both
6 months
FEV1, PEFR, PC20
van der Heide 1999 (73)Crossover
Randomisation method: computer
Double-blind
n = 22 (20 in analyses)
12 years
Test: air-cleaners
Control: placebo
3 months
Symptoms, FEV1, PEFR, PC20
Verrall 1988 (74)Crossover
Randomisation method: NA
Double-blind
n = 16 (13 in analyses)
14 years
Test: HEPA-filter
Control: placebo
3 weeks
Symptoms, medication, PEFR
Walshaw 1986 (75)Randomisation method: NA
Not blind
n = 50 (42 in analyses)
34 years
Test: mattress and pillow covers, vacuum cleaning
Control: no such measures
1 year
Symptom score, medication, FEV1, PEFR, PC20, serum immunoglobulins, RAST to D pter
Warburton 1994 (76)Crossover
Randomisation method: NA
Double-blind
n = 13 (12 in analyses)
46 years
Test: air filtration unit
Control: placebo
4 weeks
Symptom score, medication, frequency of nocturnal wakening, FEV1, PEFR, PD20
Warner 1993 (77)Crossover
Randomisation method: NA
Double-blind
n = 20 (14 in analyses)
3–11 years
Test: ioniser
Control: placebo
6 weeks
Symptom score, medication, PEFR
Warner 2000 (78)Randomisation method: NA
Double-blind
n = 40
27 children and 13 adults
Test 1: mechanical ventilation system and super vacuum cleaner
Test 2: mechanical ventilation system
Test 3: super vacuum cleaner
Control: no intervention
12 months
PEFR, FEV1, PC20, symptoms, medication
Woodcock 2003 (79, 80)Randomisation method: central
Double-blind
n = 1122, n = 732 were mite sensitive (628 of these in analyses)
36 years
Test: mattress, pillow and quilt covers
Control: placebo covers
1 year
PEFR, beta-agonists, symptoms, exacerbations and hospital visits, days of work missed, quality of life
Zwemer 1973 (81)Crossover
Randomisation method: NA
Double-blind
n = 18 (12 in analyses)
6–16 years
Test: laminar air flow system
Control: dummy filter
4 weeks
Symptoms
Three patients had sick days in the control group, none in the test group

Twelve trials had a crossover design (12, 14, 16, 35, 54, 58, 59, 73, 74, 76, 78, 81). The remaining were group comparative trials. All trials but six had used skin prick testing for diagnosis of mite sensitivity. Extracts used were D. pteronyssinus or D. farinae apart from two trials where subjects were tested with unspecified house dust extract (54, 81). In three trials, sensitivity was established by specific serum IgE (53, 73, 80), in two trials by either skin prick testing or IgE (62, 68), and in one trial published only as an abstract, the means of diagnosis was not given (42).

Thirty-six trials used physical methods to reduce exposure to mites, ten used chemical methods and eight used a combination of chemical and physical methods (Table 1). Twenty-six of the trials had used mattress covers (14, 16, 17, 23–25, 27, 28, 33, 35, 36, 38, 39, 42, 48, 52, 53, 57, 62, 65, 66, 68, 70, 72, 75, 80). Mite reduction occurred in 17 trials (17, 22, 25, 27, 28, 32, 34, 35, 39, 45, 47, 62, 65, 70, 75, 77, 80), reduction was unsuccessful in 24 and was not measured or reported in the remaining 13 trials.

Risk of bias in the included studies

The randomization method was rarely described, and even using rather broad criteria, only eight trials reported adequate concealment of allocation: sealed, opaque envelopes (24, 65), computer program (39), sealed envelopes with consecutive numbers (50), centralized, using numbers generated from a random numbers table (66), computer using minimization (72, 73), coordination centre, using minimization (80). All eight trials with adequate concealment of allocation were also reported to have been blinded, although in at least one trial, the attempted blinding was not perfect (39, 82), and in another, the intervention frequency differed between the groups (65). One trial maintained the blinding during data analysis (66).

The potential for outcome reporting bias, i.e. the omission or incomplete reporting of outcomes that were not statistically significant (83), was high. Eleven trial reports did not contain any usable data for meta-analysis (22, 35, 37, 42, 48, 49, 55, 65, 67, 72, 73). In the remaining trials, many outcomes were reported in a way that did not allow us to use them in a meta-analysis, and it was often unclear how many patients contributed values to the various analyses. The most frequently reported outcome was PEFR in the morning (1565 patients in our meta-analysis). Length of the intervention and follow-up varied from 2 weeks to 2 years.

Results

We did not find an effect of control measures to reduce the exposure to mites or their products in the 54 trials we reviewed. The total number of patients who improved after the experimental interventions was very similar to the corresponding number in the control groups, relative risk 1.01 (95% CI 0.80–1.27) (Fig. 1, data available for seven trials).

Figure 1.

 Number of patients improved.

Asthma symptom scores were very heterogeneous (P = 0.0002 for test of heterogeneity, I2 62%) (Fig. 2, 19 trials). The heterogeneity was caused by two small trials of poor quality that were the only ones that reported a significantly positive effect (68, 81). The standardized mean difference for all trials was −0.04 (95% CI −0.15 to 0.07). After exclusion of the two trials of poor quality, the standardized mean difference was −0.00 (95% CI −0.11 to 0.11).

Figure 2.

 Asthma symptom scores.

Medication usage was very similar in the experimental and control groups (Fig. 3, 10 trials). The standardized mean difference was −0.06 (95% CI −0.18 to 0.07). Data for chemical methods were given in only one trial (29) in which medication usage was significantly larger in the experimental group than in the control group (0.89, 95% CI 0.02–1.75). This finding is of doubtful value, however, as the standard deviation was unusually low and may have been erroneous. If this trial is excluded, the standardized mean difference is −0.08 (95% CI −0.20 to 0.05).

Figure 3.

 Medication usage.

For FEV1, the standardized mean difference was 0.11 (95% CI −0.05 to 0.28) (Fig. 4, 14 trials). In one trial, unusually large variations in FEV1 from visit to visit were reported which indicates that the data may not have been reliable (68). If this trial is excluded, the standardized mean difference is 0.08 (95% CI −0.08 to 0.25).

Figure 4.

 Forced expiratory volume in one second.

For peak flow in the morning, the standardized mean difference was 0.00 (95% CI −0.10 to 0.10) (Fig. 5, 23 trials). For peak flow in the evening, the standardized mean difference was also 0.00 (95% CI −0.21 to 0.21) (12 trials, data not shown).

Figure 5.

 Peak expiratory flow rate in the morning.

For PC 20, the standardized mean difference was 0.05 (95% CI −0.13 to 0.22) (Fig. 6, 14 trials).

Figure 6.

 Provocative concentration that causes a 20% fall in FEV1.

Only two trials reported on unscheduled visits to a physician or hospital, or on missed work or school days. In the largest trial included in our review, 38 patients required a hospital visit or a course of oral steroids in the intervention group and 27 in the control group; number of days of work missed was 0.10 vs 0.23 (95% CI for difference −0.28 to 0.01) (80). A crossover trial of poor quality reported that none of 12 patients missed school during the treatment period, as opposed to three during the control period; however, only 18 patients were randomized and there was no mention of reasons for missing school, and no data for six of the randomized patients (81).

Discussion

We were unable to demonstrate any clinical benefit to mite-sensitive patients with asthma of measures designed to reduce mite exposure. It is not likely that we missed a clinically relevant effect because the total number of patients in the trials was quite large. The most commonly used outcome, morning peak flow, is related to the severity of the asthma, and peak flow measurements did not suggest any worthwhile effect. This can be seen more clearly if the difference in morning peak flow is translated into the most commonly used unit, l/min. With a standard deviation of 100 l/min (in accordance with the data in Fig. 5) and a control group peak flow of 300 l/min, the experimental group peak flow would also be 300 l/min, with a 95% confidence interval that ranges from 290 to 310 l/min. A similarly narrow confidence interval around no effect was seen for asthma symptoms. When there is no indication of an effect of an intervention, subgroup analyses should not be performed, because they would be expected to be seriously misleading.

Adherence to the applied measures was rarely evaluated, but successful mite reduction was obtained in several trials, including the biggest one (80) that contributed 628 patients of a total of 1565 to the measurements of morning peak flow. It should be noted, however, that mite reduction was determined in different ways in the various studies. Some recorded mite counts and some measured antigen levels, using dust samples from different sources, and the reductions reported do not necessarily correspond to a similar reduction in the patients’ exposure. For example, removing mites from the surface of mattresses and pillows does not affect the mite content of blankets or duvets, and merely killing the mites does not necessarily reduce airborne mite antigen, if nothing is performed to remove the faecal particles that contain it. A potential reservoir for mites is the scalp, and it has been suggested that neglect of this source may explain the failure of many trials of mite eradication (84).

It seems unlikely that the initial mite levels were already too low for any reduction to be effective. It has been shown that quite low allergen concentrations can affect bronchial responsiveness (85, 86) and the concentrations were such as would usually be considered to represent a risk to mite-sensitive asthmatics. Allergen levels varied between the studies, and there was a wide range of concentrations in each study, so that some subjects’ exposure may have been very low, but this was uncommon.

Potential sources of bias should be considered. The randomization methods were rarely described. It is likely that some studies were not truly randomized or that the allocation was not adequately concealed, which are defects that would be expected to lead to bias in favour of a treatment effect (87). Most trials were very small and our sample of trials may therefore have been influenced by publication bias, which also tends to exaggerate the effect of treatment. The reporting of the data was often poor, e.g. many authors only reported that there were no significant differences between the intervention and the control groups. This lack of proper reporting would also be expected to lead to bias in favour of a treatment effect. In a comparison of 102 trial protocols with subsequent publications, it was shown that the chance that an outcome was fully reported was twice as high if the result was statistically significant (83).

We tried carefully to avoid bias during data extraction, for example by making blinded decisions when several options were available. On a few occasions, however, we could not select the data in a neutral fashion but had to choose data which favoured the hypothesis that interventions were effective. For the biggest trial (80), we selected data after 6 months rather than 12 months, in accordance with the authors’ power calculation, as this part investigated the effects of allergen reduction on asthma symptoms and was not biased by the planned reduction of steroids (there was also significant allergen reduction after 6 months, but not after 12 months). Further, there was no indication that we had excluded trials with positive results (6). We therefore believe that we have not favoured the null hypothesis of no treatment effect in our meta-analysis; if anything, we have favoured the alternative hypothesis.

Physical interventions may need to be applied repeatedly before the reduction in allergen levels is sufficient to be effective. However, the lack of effect was also apparent in the subgroup of trials with long treatment duration or follow-up. Furthermore, if the interventions were effective, one would expect to see at least some effect also in short-term trials as mite allergen causes a type 1 hypersensitivity reaction.

The house dust mite is the allergen to which asthmatics are most frequently sensitive and the acute effects of exposure on the symptoms of asthma are well established. The explanation that we find most plausible for the lack of effect of the interventions is therefore that the methods we have reviewed do not adequately reduce mite antigen levels, as it seems inherently implausible to suggest that complete removal of a major provoking agent would be ineffective. It is important to remember, however, that mite-sensitive asthmatic patients are usually sensitive to other allergens, so that successful elimination of only one allergen may have limited benefit, whatever its success. We excluded a large trial of multiple interventions in 937 patients with multiple allergies that is interesting in this respect (88). This trial reported positive effects on clinically relevant outcomes, such as number of days with symptoms, night awakenings and missed school days. However, the study was not blinded and the positive results for these subjective outcomes were obtained through telephone interviews. Furthermore, the intervention group received more home visits than the control group, results for objective outcomes such as FEV1 and PEFR were very similar for the two groups, and the allergen levels decreased by less than 50%, compared with the control group, which is far too little to be expected to have any effect. A meta-analysis that compared multifaceted with monofaceted interventions for preventing the development of asthma in newborns suggested that multifaceted interventions might be more effective, but as the comparisons were indirect, the authors also recommended to compare these modalities directly in randomized trials (89).

Reviews and guidelines do not reflect the fact that measures designed to reduce the patients’ exposure to mite antigen in the home are ineffective. In fact, they usually recommend several measures as being effective, and provide a highly selected and biased sample of references in support of such claims. The most quoted trial in 70 reviews had only seven patients per group (33), its claimed significant result was probably erroneous and it did not report a clinical outcome (90). Furthermore, recommendations were often based on nonrandomized studies (90) and the most quoted nonrandomized study (91) had included only 10 patients per group but claimed very positive results.

The recently published, very extensive US guidelines for asthma control (92) were also misleading (93). On p. 171, the expert panel recommends various interventions, including encasing the mattress in an allergen-impermeable cover. The panel quotes 10 papers in support of this, but one is an editorial, one is a review, one is a before-after study, one is about rhinitis, one was excluded from our review as only some of the patients were allergic to mites and as no outcome data were provided for this group, and one is not relevant as it involved multiple interventions and allergens. What remains are only five trials and these did not show an effect of mattress covers.

The 2008 PRACTALL consensus report (94), which was written by nominated expert teams from the European Academy of Allergy and Clinical Immunology and the American Academy of Allergy, Asthma and Immunology, is also unhelpful. It lists various measures that can reduce exposure to mites, including impermeable mattress, pillow and quilt covers. This is not wrong per se, but it is misleading, as the text says nothing about the lack of clinical effect of such measures.

We conclude that chemical and physical methods aimed at reducing exposure to house dust mite allergens cannot be recommended, as they are ineffective. Reviews and guidelines should reflect this fact. It is doubtful whether further studies, similar to the ones in our meta-analysis, are worthwhile. In particular, it should be noted that several of the trials had used very extensive mite eradication and avoidance schemes, involving many different measures applied simultaneously (6). If other types of studies are considered, we suggest that they should be methodologically rigorous and use other methods than those used so far, with careful monitoring of mite exposure and relevant clinical outcomes.

Acknowledgments

We would like to thank Leonardo Antonicelli, Cristina Cinti, S Cloosterman, Susanne Halken, Tim Higenbottam, OF Jooma, Rolf Kroidl, Christina Luczynska, Heinrich Matthys, Lucia Rijssenbeek-Nouwens, Gail Shapiro, Onno van Schayck and Lisette van den Bemt for providing additional information on their trials. We thank Cecilia Hammarquist, Michael Burr and Lasse Schmidt who were authors on previous versions of this review and Xiaohui Chen Nielsen for translation of a paper in Chinese.

Funding

Funding for previous versions of this review was obtained from The Swedish Heart Lung Foundation (grant 54506), Nordic Council of Ministers and Sygekassernes Helsefond, Denmark.

Potential conflicts of Interest

None.

Ancillary