Health‐related quality of life, using disease‐specific health‐related quality of life scores

One study used the Rhinosinusitis Outcome Measure (RSOM‐31) (Hansen 2010). Stjarne 2006a reported that scores for quality of life were recorded at every study visit using an investigator‐administered scale with the following items: "nose breathing", "experience of smell and taste", "interference with daily activities caused by nasal symptoms" and "sleep disturbance". It is unlikely that this is a validated quality of life instrument.

Disease severity, as measured by patient‐reported symptom score

All studies but one (Lang 1983) measured disease severity as reported by patients, but only Lund 2004 used a validated scale (Chronic Sinusitis Survey ‐ CSS). However, Lund 2004 reported that at the time of the study, a validated version of the Hungarian questionnaire was not available (the study was conducted in 19 centres, six of which were in Hungary).

Most studies presented the results (partially) in graphs as median or mean values. The method of measurement was reported in several ways and differed in the choice of scale, type/combination of symptoms measured, timing of measures and scoring/analysis method (see Appendix 3).

Due to these variations, it was unclear whether the total scores between studies were comparable with each other and we had to make modifications in order to standardise scores before meta‐analysis could be conducted (see Methods and Potential biases in the review process).

Significant adverse effect: epistaxis

All but five studies reported the number of patients with epistaxis and these could be meta‐analysed (Holmberg 1997; Johansen 1993; Lang 1983; Lund 1998; Parikh 2001). Holmberg 1997 and Johansen 1993 did not provide enough information about the types of adverse effects experienced and how many patients experienced them, whereas the other three studies were very small (fewer than 15 patients per treatment arm) and it was unclear whether epistaxis was not experienced by any patients or not reported.

The studies included nosebleeds of all severities, including mild.

Health‐related quality of life, using generic quality of life scores, such as the SF‐36, EQ‐5D and other well‐validated instruments

Lund 2004 used the SF‐36, but only reported for which subscale a statistically significant difference was found.

Other local adverse effects: local irritation (including oral thrush, sore throat and other local nasal irritation such as dryness, itchiness etc.)

A variety of types of possible local irritation symptoms were reported in the studies, ranging from "itchiness" to "nasal burns" and "nasopharyngeal pain/discomfort". Some studies used broader classifications whereas others used more specific description. For example, Vlckova 2009 reported an unusual combination of "adverse events", listing "rhinalgia" and "nasal discomfort" separately alongside headache; they also reported "rhinitis" as an adverse event alongside "sneezing" and "rhinorrhoea".

Other systemic adverse effects (in children ‐ stunted growth, in adults ‐ osteoporosis)

Only one study followed up patients in the longer term (Lang 1983), and this adverse effect was not reported or relevant in the other studies.

Nasal endoscopic score (either nasal polyps size score or endoscopy score, e.g. Lund‐Kennedy)

Lund 1998 and Stjarne 2006a did not report any endoscopy results.

Studies that recruited patients with nasal polyps mostly used a 0‐ to 3‐point scale. In most studies, this was the Lildholdt scale, but the definitions were not reported clearly in the other studies. Whenever studies reported these as mean values at endpoint or change from baseline, they summed the scores from both sides of the nose (0 to 6 range). However, no definitions were given in the studies that reported the results as proportions of people who had improved.

One study used a 0 to 4 scale (Holmberg 1997), after some modification to the Lildholdt scale. The investigators in Aukema 2005 scored the volume of polyps on a 0 to 10 cm visual analogue scale after conducting endoscopy.

Studies that included patients without polyps used the Lund‐Kennedy scale (Hansen 2010; Mosges 2011; Parikh 2001). It was unclear whether Lund 2004 used endoscopy scores as an outcome ‐ this was not reported.

Chronic Sinusitis Survey

Lund 2004 (chronic rhinosinusitis without nasal polyps) reported that at the time of the study, a validated version of the Hungarian questionnaire was not available (the study was conducted in 19 centres, six of which were in Hungary). Only the mean change and 95% confidence interval (CI) for each group was reported, without a sample size. We imputed the sample size based on the number of patients available at the end of the study and it seems that all available patients had filled in the CSS. The mean difference (MD) was 2.84 points (95% CI ‐5.02 to 10.70; 134 participants) between groups (range 0 to 100, higher values = better) (Analysis 1.1). The magnitude of the difference is insignificant.

Global rating scale

Vlckova 2009 used a five‐point global rating scale (very much improved, improved, same, worse or very much worse). At 12 weeks, 76% of patients in the treatment group were "improved" or "very much improved" compared to 27% of patients in the placebo group (risk ratio (RR) 2.78, 95% CI 1.76 to 4.40; 109 participants; one study) (Analysis 1.2).

Combined symptom scores

Six studies provided information on mean and standard deviation values, or had enough information in their graphs for us to estimate these values for various symptom scores (Aukema 2005; Lund 2004; Small 2005; Stjarne 2006; Vlckova 2009; Zhou 2015). Of these, only Vlckova 2009 and Lund 2004 (chronic rhinosinusitis without nasal polyps) reported an overall combined score for all four groups of symptoms covered by EPOS 2012 (nasal blockage, nasal discomfort (facial and sinus pain/pressure) and rhinitis symptoms (nasal secretion, itching, irritation and sneezing), and loss of sense of smell). Parikh 2001 also included these symptoms in their diary card and symptom scores, but it is less clear how these were calculated. Although the means and standard deviations were reported for this study, we did not include it in the meta‐analysis because the values were obviously skewed (mean of 29.4 (standard deviation (SD) 37) for intranasal corticosteroids group (n = 9), mean of 16.9 (SD 48.5) (n = 13) for placebo group, P value = 0.39 using Mann Whitney U test).

Therefore we only included six studies in our analysis. We report the average values when all four types of symptoms mentioned in EPOS were measured versus when only three types (loss of sense of smell, nasal blockage and nasal discharge) and two types of symptoms (nasal blockage and nasal discharge) were measured. All studies used 0‐ to 3‐point scales in their diaries, except for Aukema 2005, which used a 0 to 100 visual analogue scale (VAS) measured during follow‐up. All studies reported change from baseline, except for Aukema 2005, which reported the mean difference at the end of the study. To allow for ease of interpretation, we converted these 0 to 100 VAS scores into 0 to 3 by a division of 33.333. All studies used the 'usual dose' of intranasal steroids, except for Vlckova 2009, which only used a higher dose of fluticasone propionate (800 µg/day), delivered in two divided doses. Two studies, Small 2005 and Stjarne 2006, had two treatment arms using mometasone furoate nasal spray with low (200 µg/day) and high (400 µg/day) doses. Only one study included patients without nasal polyps (Lund 2004). All studies were conducted in adults.

The pooled results are as follows:

• Combined symptom score for four EPOS domains, average score: MD ‐0.26 (95% CI ‐0.37 to ‐0.15; 243 participants; two studies; I² = 46%), scale range: 0 to 3, lower = better, indicating less severe symptoms in the intranasal corticosteroids group (Analysis 1.3).

• Combined symptom score for three EPOS domains (nasal blockage, rhinorrhoea and loss of sense of smell), average score: MD ‐0.31 (95% CI ‐0.38 to ‐0.23; 1345 participants; four studies; I² = 0%), scale range: 0 to 3, lower = better (Analysis 1.3).

• Combined symptom score for two EPOS domains (only nasal blockage and rhinorrhoea), average score: MD ‐0.31 (95% CI ‐0.38 to ‐0.24; 1702 participants; six studies; I² = 0%), scale range: 0 to 3, lower = better (Analysis 1.3).

The observed mean differences correspond to a moderate effect size (the SMD was about 0.4 for the two and three domain average symptom scores and 0.6 for the four domain average symptom score). The quality of the evidence is low for the four domain scores and moderate for the three and two domains scores (facial pain/pressure not considered), with the main concerns being the use of non‐validated symptom scores and a high risk of reporting bias (many studies did not publish the results in detail and this could be linked to a lack of observed efficacy).

Individual symptom scores

• Nasal blockage: MD ‐0.40 (95% CI ‐0.52 to ‐0.29; 1702 participants; six studies; I² = 47%) (Analysis 1.4).

• Rhinorrhoea: MD ‐0.25 (95% CI ‐0.33 to ‐0.17; 1702 participants; six studies; I² = 6%) (Analysis 1.4).

• Loss of sense of smell: MD ‐0.19 (95% CI ‐0.28 to ‐0.11; 1345 participants; four studies; I² = 0%) (Analysis 1.4).

• Facial pain/pressure: MD ‐0.27 (95% CI ‐0.56 to 0.02; 243 participants; two studies; I² = 78%). Of these two studies, Lund 2004 included patients with chronic rhinosinusitis without nasal polyps, whereas Vlckova 2009 included patients with chronic rhinosinusitis with nasal polyps. Due to the differences in type, dose and delivery method (fluticasone propionate 800 µg per day using breadth actuated inhaler versus budesonide 128 µg per day as a nasal spray), the source of heterogeneity was unclear.

We used a random‐effects model to conduct the analysis; if a fixed‐effect model is used the statistical significance of the pooled MD for facial pain is ‐0.24 (95% CI ‐0.37 to ‐0.11). The quality of the evidence is moderate for nasal blockage, rhinorrhoea and loss of sense of smell, but low for facial pain/pressure.

Chronic rhinosinusitis with nasal polyps ‐ reduction in polyps size

The MD in polyps score was ‐0.58 (95% CI ‐0.90 to ‐0.26; 1417 participants; four studies; I² = 83%) indicating less severity for the intranasal corticosteroids group (Analysis 1.7). All reported the sum of polyps score from both sides of the nose (range 0 to 6). One study, Vlckova 2009, had an effect that was larger than the other studies, with a mean difference of ‐1.21 (95% CI ‐1.56 to ‐0.86) points between treatment arms in the reduction of polyps size score. When this study is removed, the heterogeneity is resolved and the observed effect size is smaller (MD ‐0.35, 95% CI ‐0.47 to ‐0.24; 1308 participants; three studies; I² = 0%).

Five studies reported the proportion of participants who had an improvement in their polyps scores. The chance of an improvement was higher in patients on intranasal corticosteroids (RR 1.77, 95% CI 1.06 to 2.95; 676 participants; five studies; I² = 66%) (Analysis 1.8). The observed heterogeneity was resolved when we removed a study with an outlier effect (Vlckova 2009), but the RR became slightly smaller (RR 1.46, 95% CI 1.12 to 1.90; 567 participants; four studies; I² = 0%).

Using a method recommended in the Cochrane Handbook for Systematic Reviews of Interventions, we converted the continuous outcome data into proportions and analysed them together using the generic inverse variance method. For Vlckova 2009, which reported both the mean difference and proportions, we used the values reported for proportions to avoid double counting and to minimise imputations. The overall pooled odds ratio (OR) was 2.07 (95% CI 1.48 to 2.91; 1984 participants; eight studies) (Analysis 1.9). The observed heterogeneity was quite substantial (> 50%). We observed that Vlckova 2009 seemed to have a larger effect than the other studies. If this study is removed from the analysis, the OR is slightly lower (OR 1.71, 95% CI 1.43 to 2.04; 1875 participants; seven studies) and heterogeneity is no longer observed.

Chronic rhinosinusitis without nasal polyps ‐ endoscopy score

The studies that included patients without polyps, such as Hansen 2010, Mosges 2011 and Parikh 2001, used a modified Lund‐Kennedy score. However, the results were either only partially reported or could not be meta‐analysed due to highly skewed distribution (very small sample sizes). Lund 2004 did not report the endoscopy score as an outcome and it was unclear whether this was measured.

Parikh 2001 reported the percentage of change and standard deviation. This was ‐22.3% (SD 61.8) (n = 9) for the intranasal corticosteroids group and +19.9% (SD 58.3) (n = 13). This was not statistically significant (Mann Whitney U test). Mosges 2011 showed the total endoscopy score for redness, oedema and discharge on a chart. There were no statistically significant differences between groups. Similarly, Hansen 2010 reported no statistically significant differences between groups for different aspects of the assessment, except for oedema.