Results of the search

Based on our searches, we screened 12,230 references from electronic databases, and one study from other sources. The last search was run in December 2020, which updated previous searches in June 2016 and 2017 and August 2019. Based on title and abstract and, if unavailable, referring to the full text, a total of 230 references entered the full‐text screening. This resulted in 13 studies to be included in the data extraction process and 192 studies were excluded after removing of duplicates (see Excluded studies). Handsearching of the reference lists did not yield additional studies.

Included studies

Study types

Occupational setting

We initially found eight studies conducted in the occupational setting and comprising office (three studies), factory (one study) and hospital workers (four studies). The data for one study (factory study) could not be extracted, as the data reported did not match our data extraction requirements (Enomoto‐Koshimizu 2002).

Findings of seven studies were therefore reported, with three controlled before‐after studies (Hashiguchi 2008; Norbäck 2000; Nordström 1994) and three controlled cross‐over studies (Gavhed 2005; Reinikainen 2003; with one cluster‐randomised cross‐over design (Reinikainen 1992)). Here, the unit of randomisation was the office section (in two wings) of the building, randomly assigned at the beginning of the study (Reinikainen 1992). One study used a non‐randomised controlled parallel‐group design (Green 1981).

Educational setting

Five studies targeted an educational setting and were designed as non‐randomised controlled parallel‐group studies (Green 1975; Reiman 2018; Ritzel 1966; Sale 1972; Sataloff 1963).

Participants

Occupational setting

A total of 3422 office and hospital workers were reported as participants, of whom 3160 (92%) were included in the data analyses. No age‐related information was given in the studies of Green 1981 and Reinikainen 2003. The ages of the study populations of Hashiguchi 2008; Norbäck 2000 and Nordström 1994 were as follows: Hashiguchi 2008: mean 40.1 years (± 10 standard deviation (SD)) in the intervention and mean 38.7 (± 11.5 SD) in the control group; Norbäck 2000: mean 39 (± 10 SD) in the intervention group and mean 44 (± 12 SD) in the control group; Nordström 1994: mean 40 years (± 13 SD) in the intervention and mean 38 (± 9 SD) in the control group (at the end of the study). Regarding Gavhed 2005 and Reinikainen 1992, a wider age range was described (see Characteristics of included studies). The population in the studies of Hashiguchi 2008; Norbäck 2000; and Nordström 1994 included predominantly women, whereas in Reinikainen 1992 there were almost half women and men. No gender‐related information was presented in Gavhed 2005; Green 1981, and Reinikainen 2003.

Educational setting

Studies conducted in the educational setting included kindergarten and school children. No study mentioned the number of participants in the intervention and the control groups. Three studies did not report the number of individuals that entered the analyses (Green 1975; Ritzel 1966; Reiman 2018), but indicated the total school days per group (Ritzel 1966; Reiman 2018). The two remaining studies included 222 children in the intervention and 203 in the control group in the analyses (Sataloff 1963; Sale 1972). Sataloff 1963 provided no age information for the participants. Reiman 2018; Ritzel 1966 and Sale 1972 included children at preschool age (2 ‐ 5, 4 ‐ 6, 2½ ‐ 6 years of age, respectively) and the pupils in the study of Green 1975 were aged between six and 14 years. None of the studies reported the gender distribution.

Sample size

Occupational setting

Three studies (Gavhed 2005; Norbäck 2000; Hashiguchi 2008) conducted in the occupational setting had very small sample sizes, below 50 participants (39, 32, 45) and a further three studies (Nordström 1994; Reinikainen 1992; Reinikainen 2003) reported fewer than 550 participants (104 (dynamic population), 290, 517). In Green 1981 the number of participants exceeded 2000 (2395), but with only 185 participants in the intervention group. The data for participants entering the data analysis are provided in Characteristics of included studies and the summary of findings tables. No sample size or power calculations were reported.

Educational setting

The number of study participants ranged between 116 and 515 (116, 232, 515, 162) (Reiman 2018; Ritzel 1966; Sale 1972; Sataloff 1963) and one study only reported that 12 schools with children of grades 1 to 8 were included (Green 1975). The number of participants entering the data analysis is provided in Characteristics of included studies and the summary of findings tables. No sample size or power calculations were reported.

Year of publication and geographical location

Occupational setting:

These studies were published between 1981 and 2008, with three articles before 2000 and four between 2000 and 2008. They were conducted in Canada (1), Finland (2), Japan (1), and Sweden (3).

Educational setting:

These articles were published between 1963 and 1975 and in 2018. The studies were performed in Canada (1), Switzerland (1), and the USA (3). The Swiss study was published in German.

Exposure and co‐exposure

All the included studies assessed relative indoor air humidity as an indicator for dry air. In most studies, humidity was recorded continuously during the whole or parts of the study period. In one study (Gavhed 2005), the method of exposure assessment was not reported, and values of absolute humidity were mentioned in three studies (Reiman 2018; Reinikainen 2003; Hashiguchi 2008). The levels of relative humidity in the included studies varied considerably and the measured values were stated as a range, a mean value with or without standard deviation or as an unspecified value. The differences between humidity levels in the control and intervention groups were also heterogeneous. In most of the studies there were only small differences (less than 10% in humidity levels between humidified and un‐humidified conditions) or the relative humidity in both conditions overlapped (Nordström 1994; Reinikainen 1992; Reinikainen 2003; Reiman 2018; Sale 1972; Sataloff 1963). The differences between humidity levels were analysed in one study and there were important differences in humidity levels between groups (Nordström 1994). An obvious difference (more than 10% difference in humidity levels) was reported in Gavhed 2005 and Reiman 2018. In all studies temperatures for both conditions were stated, and in 10 of the 12 studies the temperature levels were almost the same in the intervention and control setting (Gavhed 2005; Green 1975; Hashiguchi 2008; Norbäck 2000; Nordström 1994; Reiman 2018; Reinikainen 1992; Ritzel 1966; Sale 1972; Sataloff 1963). All studies were conducted during the heating period.

Occupational setting

In some occupational studies co‐exposure was assessed. In Norbäck 2000 and Nordström 1994 volatile organic compounds were measured, and Reinikainen 2003 assessed the formaldehyde concentration. Nordström 1994 and Reinikainen 2003 additionally assessed particle concentration. Biological exposure to bacteria and fungal spores was measured in Reinikainen 2003 and Reinikainen 1992, and they also mentioned the ventilation rates. In all these studies the co‐exposure seemed to be at low levels or within reference ranges and therefore an influence on developing or increasing dryness symptoms of the skin, eye and upper respiratory tract was not considered.

Educational setting

The most recent study conducted in the educational setting (Reiman 2018) assessed particles as co‐exposure. The average concentration and size of particles in both conditions were measured, with 96% of the particles provided by humidification less than 1 μm. Humidified rooms showed a near doubling of both 1 ‐ 4 μm, and > 4 μm air particles. There was a significant increase in the population of larger‐sized particles (1 – 4 μm and > 4 μm) in the humidified versus control rooms. The other studies in the educational setting did not assess further co‐exposure factors, except for temperature (Green 1975; Ritzel 1966; Sale 1972; Sataloff 1963).

Interventions and comparisons

Each study evaluated one intervention: the effect of indoor air humidification.

Intervention in the occupational setting

In four studies (Gavhed 2005; Green 1981; Norbäck 2000; Nordström 1994) central humidification was conducted, and in three trials (Hashiguchi 2008; Reinikainen 1992; Reinikainen 2003) local humidification. Mostly, steam humidifiers were used. No study was dealing with other interventions (such as putting plants around the workplace or placing a container of water or wet cloths in proximity to a radiator or a heating system).

Intervention in the educational setting

Most studies used only local humidification (Reiman 2018; Ritzel 1966; Sataloff 1963), and one study reported central humidification (Sale 1972). Various types, such as steam humidifiers, water atomizers, spray cold humidifiers, air washers and boilers were used in one study involving 12 different schools (Green 1975). No study was using other interventions.

Study duration

Occupational setting

In the occupational setting, the study duration ranged from six weeks (Norbäck 2000; Reinikainen 1992; Reinikainen 2003) to seven months (Green 1981), including eight weeks (Gavhed 2005), 12 weeks (Hashiguchi 2008) and five months (Nordström 1994). Green 1981 was conducted during three seasons. In order to investigate effects of indoor air humidity on symptoms and infections after very short (days) and longer time periods, we initially intended to establish three groups using three time scales (up to one month, between one and three months, longer than three months). Our data did not allow us to undertake such subgroup analyses.

Educational setting

In the educational setting, the study duration ranged from seven weeks (Reiman 2018) to six months (Green 1975). One study (Green 1975) presented additional data over a time period of 10 years.

Control

Occupational setting

Three studies (Gavhed 2005; Reinikainen 1992; Reinikainen 2003), all comprising office workers, described a cross‐over design with a control group.

In the remaining four studies (Green 1981; Hashiguchi 2008; Norbäck 2000; Nordström 1994), the control group consisted of hospital employees. Green 1981 used two control groups, which worked in two different non‐humidified hospitals. In this trial, the number of participants in the control site was disproportionate higher than in the intervention site (n = 650 in one control group, n = 1560 in the second control group versus 185 in the intervention group). The intervention group was stationed in the third hospital building, which was humidified. In Hashiguchi 2008 and Norbäck 2000, the control and intervention group participants worked in the same building, but in different stations and units, respectively. Nordström 1994 was conducted in four units in two hospitals. Two randomly‐selected units, one per hospital, served as intervention and control sites, respectively.

Educational setting

Green 1975 was conducted in 12 primary schools. The control group included four non‐humidified buildings and also non‐humidified classes in another school. The control group of Ritzel 1966 were located in five non‐humidified kindergartens, and the intervention group in five nearby kindergartens. In Sale 1972, there was no clear control group. For the purposes of the analysis, we combined a group consisting of children who were not exposed to humidified air with the group of children exposed to humidified air at home only. Reiman 2018 and Sataloff 1963 were conducted in one building, where some classes were humidified and others were not.

Outcomes and outcome assessment

Occupational setting

Primary outcomes

Primary outcomes were investigated in only six of the seven studies in the occupational setting (Gavhed 2005; Hashiguchi 2008; Norbäck 2000; Nordström 1994; Reinikainen 1992; Reinikainen 2003). Different outcome definitions were used for the assessment of symptoms related to the location. We established groups of similar symptoms according to the targeted location.

Almost all studies used self‐administered questionnaires to survey symptoms. In Hashiguchi 2008 interviews with staff members about their symptoms were carried out once a week. In all studies, the symptoms and conditions were not assessed or diagnosed by a physician.

Two studies, (Norbäck 2000; Nordström 1994) referred to a standardised questionnaire for assessing symptoms, the MM‐040‐NA questionnaire. This validated instrument was developed at the Department of Occupational and Environmental Medicine, Örebro University Hospital in Sweden. As it was established as an exploratory questionnaire aiming to assess the sick building syndrome and no test statistics were mentioned, we left it unconsidered as a reference questionnaire for assessing work‐related dryness symptoms. Furthermore, there were indications that Nordström 1994 and Norbäck 2000 used modified versions of MM‐040‐NA (different number of questions, different scoring).

The questionnaire used by Norbäck 2000 comprised seven questions about eye irritation (one question), airway symptoms (three questions) and dermal symptoms (three questions). The prevalence of participants with at least one weekly symptom of the eyes, airway, and skin was determined. Raw ORs were adjusted for age, sex, atopy, smoking habits, employment time, type of occupation, and psychosocial work climate. We entered the number of events into RevMan and computed unadjusted ORs for comparability. The prevalence of participants with at least one weekly symptom of the eyes, airway, and skin was calculated in Nordström 1994.

Reinikainen 1992 used one weekly symptom diary, in order to record symptoms in the humidified and non‐humidified conditions. For data collection, Reinikainen 2003 used a structured diary which the participants filled in every afternoon. Strong symptoms were coded as 3 and no symptoms as 0. Gavhed 2005 used a self‐administered questionnaire with a five‐point Likert scale which was filled in once a week. For the analysis, we pooled the frequency responses 'seldom and never'; 'daily, many times a week', 'few times a week'. The data were transformed to ORs in RevMan.

In Norbäck 2000, there was an additional objective measurement of the outcomes (see below).

Eye symptoms

Five studies evaluated the effect of humidification on eye symptoms: three cross‐over studies (one RCT: Reinikainen 1992 and two non‐RCTs: Reinikainen 2003; Gavhed 2005) and two before‐after studies (Nordström 1994; Norbäck 2000). Reinikainen 1992 investigated eye symptoms described as 'dryness, irritation and itching'. Reinikainen 2003 used the term 'eye dryness', without listing more details, and Gavhed 2005 additionally used the terms 'itching' and 'burning' for describing dry eyes. Nordström 1994 defined eye symptoms as 'itching, burning, or irritation in the eyes' and Norbäck 2000 evaluated eye symptoms summarised as 'burning, dry, sore eyes, eye redness, swollen eyelids'.

Objective measurements were applied in one study (Norbäck 2000). Clinical signs of the eyes were rated by assessing tear‐film stability: “a standardised method, measuring the time (tear film breakup time) the subject could keep the eyes open without pain, when watching a fixed point at the wall”.

Skin symptoms

Two cluster non‐randomised cross‐over studies (Reinikainen 2003; Gavhed 2005) and two before‐after studies (Hashiguchi 2008; Nordström 1994) investigated 'dry skin' as an outcome (summary of findings Table 1). Norbäck 2000 evaluated the following skin symptoms: facial itching, facial rash, itching on the hands, rash on the hands, or eczema, whereas Reinikainen 1992 defined skin symptoms as dryness, irritation and itching.

Gavhed 2005 used a self‐administered questionnaire assessing the frequency of 'dry skin' symptoms once a week on a five‐point Likert scale. For the analysis, we pooled the answers 'seldom' and 'never'; 'daily', 'many times a week', 'few times a week'.

Hashiguchi 2008 used a four‐point scale for reporting the frequency of dry and itchy skin: 'none', 'rarely', 'sometimes' and 'frequently'. We pooled the responses for 'rarely', 'sometimes' and 'frequently' for the purposes of our review, and treated the group 'none' separately.

Upper respiratory tract (URT) symptoms

Five studies evaluated the effect of humidification on upper respiratory tract symptoms: three cross‐over studies (one RCT: Reinikainen 1992 and two non‐RCTs: Reinikainen 2003; Gavhed 2005) and two before‐after studies (Nordström 1994; Norbäck 2000). Different symptoms and location of the symptoms were investigated: Reinikainen 2003 and Reinikainen 1992 evaluated dry nose and pharyngeal dryness. Gavhed 2005 used 'dry mouth and throat' as an outcome with a five‐point Likert scale for the frequency of symptoms ('never', 'seldom', 'daily', 'few times a week', 'many times a week'). For this review, the points 'seldom' and 'never' as well as 'daily', 'many times a week', 'few times a week' were put together. In Norbäck 2000 the throat symptoms were defined as dryness in the throat, sore throat, irritative cough (not shown), and nose symptoms such as runny nose, nasal itching, sneezing, or nasal obstruction. Nordström 1994 summarised the airway symptoms as irritated, stuffy or running nose, hoarse or dry throat, or cough.
In addition, Norbäck 2000 measured nasal signs by acoustic rhinometry. As parameters, the minimum cross‐sectional areas (MCA) and the volumes of the nasal cavity (VOL) on each side of the nose were measured. The mean of three subsequent measurements was calculated. Furthermore, the concentrations of the following biomarkers: eosinophilic cationic protein (ECP), myeloperoxidase (MPO), albumin, and lysozyme were measured in the nasal lavage.

Upper respiratory tract (URT) infections

Symptoms of irritation, infection and allergy in the upper respiratory tract might overlap, which made their differentiation challenging. No studies investigated upper respiratory tract infections directly, which means that the diagnosis of this condition was not assessed as an outcome. Symptoms related to airway infection were investigated in Gavhed 2005; Norbäck 2000; Nordström 1994; Reinikainen 1992; Reinikainen 2003 (see above).

Of these studies, only Reinikainen 2003 mentioned an assessment of symptoms of upper respiratory infection. They stated that the applied diary comprised questions about symptoms of upper respiratory infection. These symptoms were analysed separately and no conclusion was drawn on the effect of indoor air humidification on upper airway infections.

Secondary outcomes

Perceived air quality

Perceived air quality was investigated in the following studies: Hashiguchi 2008; Norbäck 2000; Nordström 1994; Reinikainen 1992; Reinikainen 2003. Perception of dryness was assessed in five studies: one non‐RCT cross‐over study (Gavhed 2005), one RCT cross‐over study (Reinikainen 1992) and three before‐after studies (Hashiguchi 2008; Norbäck 2000; Nordström 1994). Perception of stuffiness was analysed in four studies: one non‐RCT cross‐over study (Reinikainen 2003), one RCT cross‐over study (Reinikainen 1992) and two before‐after studies (Norbäck 2000; Nordström 1994).

Nordström 1994 calculated changes in dryness perception and perception of stuffiness for each individual (range of the scale −2 to 2) at the beginning and end of the study period. The results were presented as incidence of decreased and increased perception. The changes after four months were presented, but the baseline data were unavailable. To achieve a comparison to Norbäck 2000, we recoded the reported levels of changes in participants' perceptions (i.e. increased, unchanged, decreased) into numerical variables (i.e. 0 = decreased, 1 = unchanged, 2 = increased). We transformed these data into means with standard deviations, and computed standardised mean differences in order to express the size of the intervention effect. The participants in Norbäck 2000 subjectively rated air quality (air dryness and stuffy air) on a scale between 0% to 100% at the beginning and the end of the study period in the humidified group and controls. We took the values at the study end in order to calculate the standardised mean difference. We pooled the computed values of Norbäck 2000 and Nordström 1994.

Gavhed 2005 used a self‐administered questionnaire with a five‐point frequency scale including 'too dry', 'slightly dry', 'neutral', 'slightly moist' and 'too moist'. For the analysis, we pooled the levels 'too dry, slightly dry', 'neutral' and 'slightly moist, 'too moist', respectively.

Hashiguchi 2008 used a seven‐point frequency scale for reporting the perception of dryness, with 'very dry', 'dry', 'slightly dry', 'neutral', 'slightly damp', 'damp' and 'very damp'. We pooled the results into two categories. The responses to 'very dry', 'dry' and 'slightly dry' were than compared with those of 'neutral', 'slightly damp', 'damp' and 'very damp'.

Absence from work

In the occupational setting only Green 1981 assessed absenteeism.

Educational setting

Primary outcomes

No studies conducted in educational settings investigated our primary outcomes.

Secondary outcomes

Absence from school

Five studies investigated absenteeism, but using different outcome definitions. Some studies assessed absenteeism related to sickness in general or to upper respiratory tract illnesses, whereas others presented results for absenteeism in general without presenting reasons: Ritzel 1966 assessed absenteeism due to cold symptoms (total number of absence days). Reiman 2018 evaluated absenteeism due to sickness and influenza‐like illness (with symptoms: fever + cough or fever + sore throat) and for all reasons (sickness, influenza‐like illness and vacation). Green 1975 investigated total absenteeism and absenteeism due to illness, e.g. due to cold, in 12 public schools. As the data for absenteeism due to cold were only presented for a subsample, total absenteeism during the heating periods between 1960 and 1970 were presented here.

In Sale 1972, four participant groups were established: group I (humidification at school and at home), group II (humidification at school only), group III (humidification at home only) and group IV (no humidification). In order to assign an intervention group, we pooled groups I and II, and groups III and IV were considered as the control group. Retrospectively, paediatricians of the children in groups I to III reviewed their records about the frequency and severity of illnesses. The results for group IV remain unreported, and we therefore present here the average weekly absence due to all causes.

Sataloff 1963 reported the average days of absence and the average number of illnesses of included children. As absences due to respiratory infections were not specifically assessed but only suggested, we refer to the average days of absence (“average number of school days missed”) in this review.

Excluded studies

At the full‐text screening, we excluded 192 studies as outlined in Figure 3, as they did not meet our inclusion criteria for study design (n = 92), intervention (n = 34), setting (n= 3), population targeted (n = 2), or outcome (n = 1). In total, five excluded studies were performed in tropical or subtropical climate zones and 55 were excluded for other reasons (e.g. reviews, book chapters, letters, studies dealing with other topics). The excluded studies with the reasons for exclusion are presented in the Characteristics of excluded studies. We removed the duplicates from the list of excluded studies. For some studies, there were several reasons for exclusion. We therefore decided to prioritise the exclusion criteria according to the sequence in the PICOS‐scheme (participants, intervention, comparator, outcome, study design).

Allocation

As one study (Reinikainen 1992) was randomised at a group level, the bias criteria: random sequence generation and allocation concealment were assessed for this study only. However, the method of randomisation was not stated, and we therefore assessed it as having an unclear risk of bias. Allocation concealment was not reported and was therefore rated as unclear.

Blinding

We assessed blinding of participants or organisations and, if applicable, outcome assessors. In the case of self‐reported questionnaires, blinding is not applicable for the outcome assessment tool. Due to lack of information on blinding, most of the studies had unclear risk of bias. For one study (Sale 1972), we rated the risk of performance bias as high, because parents of each child were informed in advance about the purpose of the study and the harmful effects of dry heated air with a booklet, and were encouraged to co‐operate.

Incomplete outcome data

We judged the risk of attrition bias to be low in six studies (Gavhed 2005; Green 1981; Hashiguchi 2008; Norbäck 2000; Reiman 2018; Ritzel 1966) and unclear in four studies (Green 1975; Reinikainen 2003; Sale 1972; Sataloff 1963). As the percentage of withdrawals and dropouts exceeded 20% for short‐term follow‐up, Nordström 1994 was classified as having a high risk of bias. In Reinikainen 1992 missing outcome data were observed without presenting the reasons for it.

Selective reporting

Selective reporting was difficult to judge because none of the included studies had published a protocol. As we therefore had insufficient information to permit a judgement of low risk or high risk, we assigned all studies as having an unclear risk of bias. We did not find any indications in the included studies that prespecified outcomes were not reported.

Other potential sources of bias

As other potential sources of bias, we considered seasonality for the conduct of the study and exposure conditions before the start of the study. For five studies (Green 1975; Green 1981; Reinikainen 2003; Ritzel 1966; Sale 1972) we had insufficient information to assess this domain, and therefore classified these studies as having some concerns. Five studies appeared to be free of other sources of bias (Gavhed 2005; Hashiguchi 2008; Norbäck 2000; Reiman 2018; Reinikainen 1992). Furthermore, these trials were seasonally conducted during the heating period and in the study setting, the participants were exposed to the same indoor air condition before the start of the study. Because of the dynamic population in Nordström 1994, we judged it to be at high risk of bias. As due to technical problems there was virtually no difference between humidity levels in the control and the intervention groups, Sataloff 1963 was classified as having a high risk of bias.

Bias due to confounding

All studies were assessed for confounding. Age, gender, season, comorbidities, atopic conditions and co‐exposures (especially temperature) were considered as relevant confounders. If at least 60% or more of the relevant confounders were considered in the statistical analysis and studies also used either restriction, matching/pre‐stratification, and/or adjustment for confounding in the statistical models, we rated studies at low risk of bias. According to these criteria, Green 1975; Green 1981; Hashiguchi 2008; Reiman 2018; Ritzel 1966; Sale 1972; Sataloff 1963 were classified as high risk of bias. Due to a cross‐over study design where each participant served as its own control, all three cross‐over studies (Gavhed 2005; Reinikainen 1992; Reinikainen 2003) had low risk of bias. Norbäck 2000 and Nordström 1994 considered at least 60% of the relevant confounders in the statistical analysis and were therefore judged to have a low risk of bias.

Bias due to selection of participants into the study (non‐randomised studies)

In the non‐randomised studies we judged selection of participants due to the inclusion and exclusion criteria as reported in the articles. Most of the included studies had insufficient information to permit judgement of low risk or high risk of bias. Hence, the selection was classified as having some concerns. In Norbäck 2000; Nordström 1994; Reinikainen 2003; Ritzel 1966 there were no indications of bias resulting from inclusion and exclusion criteria.

Bias due to inappropriate use of the cross‐over design and carry‐over effects (cross‐over studies)

All three cross‐over studies (Gavhed 2005; Reinikainen 1992; Reinikainen 2003) were considered as having a low risk of bias.

Overall risk of bias by study

We judged studies to have a low overall risk of bias if we assessed them to have a low risk of bias in the following domains: confounding, blinding of participants, incomplete outcome data. This left us with no studies at low risk of bias, but all at either unclear (Gavhed 2005; Reinikainen 2003) or high risk of bias (Green 1975; Green 1981; Hashiguchi 2008; Norbäck 2000; Nordström 1994; Reiman 2018; Reinikainen 1992; Ritzel 1966; Sale 1972; Sataloff 1963).

Primary outcomes

Outcome 'dry eye'

summary of findings Table 1 outlines details about the effectiveness of humidification on dryness symptoms of the eye. This outcome was assessed in five studies (data analysed of 720 participants): three cross‐over studies (one RCT: (Reinikainen 1992) and two non‐RCTs: (Reinikainen 2003; Gavhed 2005)) and two before‐after studies (Nordström 1994; Norbäck 2000).

The cluster randomised cross‐over study Reinikainen 1992, conducted over six weeks, showed a significant reduction in eye dryness following indoor air humidification. The study reported a risk ratio for dry eye symptoms of 0.64 (95% CI 0.44 to 0.93), which indicated that people were 36% less likely to experience dryness symptoms during the humidified phase compared to the non‐humidified phase. After the transformation of the results according to Elbourne 2002, the calculated OR was 0.54 (95% CI 0.37 to 0.79) (low‐certainty evidence) (Analysis 1.1).

We combined the results of two cluster non‐randomised cross‐over trials (Reinikainen 2003; Gavhed 2005) and revealed non‐significant positive effects on eye dryness following indoor air humidification over a study duration of six and eight weeks, respectively (OR 0.58, 95% CI 0.27 to 1.25) (low‐certainty evidence) (Analysis 1.2). SImilarly, two before‐after studies (Norbäck 2000; Nordström 1994) showed a non‐significant positive effect of indoor air humidification on dry‐eye symptoms over a study period of six weeks and four months, respectively (OR 0.57, 95% CI 0.23 to 1.41) (very low‐certainty evidence) (Analysis 1.3).

In the before‐after study of Norbäck 2000, the tear breakup times were similar in both the intervention and control groups (Analysis 1.5).

Outcome 'dry skin'

This outcome was investigated in four trials (two cluster non‐RCT cross‐over and two before‐after studies) (data of 528 participants entered analysis). Gavhed 2005 (cluster non‐RCT cross‐over study) showed a significant effect of humidification on dryness of the skin: OR 0.43 (95% CI 0.19 to 0.96). In Reinikainen 2003, humidification decreased skin dryness, but the result was not statistical significant: OR 0.89 (95% CI 0.52 to 1.51). Both these cross‐over studies showed a non‐significant alleviation of skin dryness following indoor air humidification over a study period of one to three months: OR 0.66 (95% CI 0.33 to 1.32) (low‐certainty evidence) (Analysis 1.6). One before‐after study yielded a positive effect of indoor air humidification on skin dryness over a study period of 12 weeks (Hashiguchi 2008), whereas the other before‐after study showed no effect following indoor air humidification over a study period of four months. The pooled analysis of these two trials provided no statistically significant result (OR 0.69, 95% CI 0.33 to 1.47) (Analysis 1.9).

Outcome 'skin symptoms'

Office workers in Reinikainen 1992 reported fewer skin symptoms of dryness, irritation or itching during the humidification phase than under non‐humidified conditions: OR 0.59 (95% CI 0.39 to 0.89) (Analysis 1.7). This difference in dermal symptoms reached statistical significance.

Skin results for Norbäck 2000 are not included here, as the symptoms were not unequivocally considered as dryness symptoms (“facial itching, facial rash, itching on the hands, rash on the hands, or eczema”).

Outcome 'dry nose'

The results are summarised in summary of findings Table 1. This outcome was reported in two cross‐over studies (data analysed of 579 participants): one RCT: Reinikainen 1992 and one non‐RCT: Reinikainen 2003. The cluster non‐randomised cross‐over study (Reinikainen 2003) reported an alleviation of nose dryness following indoor air humidification over a study period of six weeks. Hence, the result did not reach statistical significance: OR 0.87 (95% CI 0.53 to 1.42), and was of low‐certainty evidence (Analysis 1.11). The cluster‐randomised cross‐over study (Reinikainen 1992) revealed no effect of indoor air humidification on nose dryness over a study period of six weeks: OR 1.08 (95% CI 0.73 to 1.60) (low‐certainty evidence) (Analysis 1.12).

Outcome 'nose symptoms'

The outcome 'nose symptoms' in Norbäck 2000 comprised runny nose, nasal itching, sneezing or nasal obstruction. No effect of air humidification on nose symptoms was found (Analysis 1.13). Furthermore, no significant effects of humidification on any of the measured physiological signs during the study period of six weeks in exposed participants and controls were observed (Analysis 1.14).

Outcome 'airway symptoms'

Nordström 1994 investigated airway symptoms such as 'irritated, stuffy or running nose, hoarse or dry throat, or cough'. No significant differences between intervention and control participants were found (Analysis 1.15).

Outcome 'dry mouth and throat'

The five categories of the symptoms' frequency were pooled to two categories, as described earlier (Gavhed 2005). The air humidification had a positive effect on 'dry mouth and throat': OR 0.25 (95% CI 0.11 to 0.57) (Analysis 1.16). More specifically, under humidified conditions dry‐mouth‐and‐throat‐symptoms were less frequently reported than under non‐humidified conditions.

Outcome 'pharyngeal dryness'

Humidification increased pharyngeal dryness: OR 1.15 (95% CI 0.63 to 2.11) in Reinikainen 2003 (Analysis 1.17). In Reinikainen 1992 the intervention led to an alleviation of pharyngeal irritation: OR 0.73 (95% CI 0.49 to 1.07) (Analysis 1.18). Neither result was statistically significant.

Secondary outcomes

Outcome 'perception of dryness'

Perception of dryness was assessed in five studies: one cluster non‐RCT cross‐over study (Gavhed 2005), one cluster RCT cross‐over study (Reinikainen 1992) and three before‐after studies (Hashiguchi 2008; Norbäck 2000; Nordström 1994).

Norbäck 2000 showed that air dryness was reduced after the six‐week period of humidification (Analysis 1.21) and Nordström 1994 presented the same effect after four months (Analysis 1.23 and Analysis 1.24). Following pooling of the data, the pooled standardised mean difference was −0.48 (CI 95% −3.49 to 2.54) (Analysis 1.25). The heterogeneity between the studies was considerable (I² value of 97%).

Hashiguchi 2008 found a statistically significant reduction in the perception of dryness following air humidification (Analysis 1.22). We pooled the two before‐after studies of Norbäck 2000 and Hashiguchi 2008, and the result showed a statistically significant effect of air humidification on the perception of dryness: OR 0.05, CI 95% 0.01 to 0.22 (Analysis 1.26). The heterogeneity between the studies was considerable (I² value of 99%).

The only cluster‐randomised cross‐over study (Reinikainen 1992) reported that the sensation of dryness was significantly increased during the non‐humidified phase compare with during humidification (Analysis 1.19). Similarly, Gavhed 2005 revealed a significant effect of humidification on the perception of dryness (Analysis 1.20).

Outcome 'absenteeism'

Green 1981 had investigated absenteeism in three hospitals during three winter seasons (October to April) in the years 1973 to 1974, 1974 to 1975, 1975 to 1976, and concluded that increasing the relative humidity had reduced absenteeism (pooled mean difference over three seasons −0.57 (CI 95% −0.61 to −0.53) (Analysis 1.33).

Five studies in the educational setting investigated effects of air humidification on absenteeism; see summary of findings Table 3.

Ritzel 1966 showed a reduction in absenteeism due to cold symptoms in the humidified versus non‐humidified kindergartens: OR 0.54, CI 95% 0.45 to 0.65 (Analysis 2.1). Sataloff 1963 revealed no reduction in the average days of absence under humidified conditions in a public school when compared to the non‐humidified condition (Analysis 2.2). According to Sale 1972, the average weekly absence due to all causes was reduced in the humidified schools versus the non‐humidified ones: OR 0.38, Cl 95% 0.15 to 0.96 (Analysis 2.3).

The average 10‐year absenteeism rate in the non‐humidified schools was found to be 5.08%, and 4.63% in the humidified schools, which was reported to be statistically significant at a 95% confidence level (Green 1975) (Analysis 2.4).

The number of absences due to illness was the same in the humidified and the control rooms (Reiman 2018) (Analysis 2.5). The percentage of students with influenza‐like illness absences was lower under the humidified versus the non‐humidified condition (no statistical testing) (Analysis 2.6). These results were supported by the distribution of the positive influenza virus samples investigated by PCR under both conditions: in the humidified rooms, the genome copies per cubic meter of the influenza A virus were lower than in the non‐humidified rooms.

Adverse events

Four studies investigated 'perception of stuffiness', which was considered to be an adverse effect of humidification (one cluster‐RCT cross‐over trial, one cluster non‐RCT cross‐over trial and two before‐after studies) (summary of findings Table 2). Both cross‐over studies (Reinikainen 1992; Reinikainen 2003) showed that the perception of stuffiness was more common during humidification than in the non‐humidified phase after six weeks (Analysis 1.27; Analysis 1.28) which was found to be statistically significant: OR 2.18, CI 95% 1.47 to 3.23 and OR 1.70, CI 95% 1.10 to 2.61. In both before‐after studies (Norbäck 2000; Nordström 1994) no statistically significant effects of air humidification were observed for the perception of stuffy air after one and four months (pooled standardised mean difference 0.24 (−0.30 to 0.78)) (Analysis 1.29; Analysis 1.30; Analysis 1.31; Analysis 1.32).