Combination inhaled steroid and long‐acting beta2‐agonist versus tiotropium for chronic obstructive pulmonary disease

Abstract

Background

Combination therapy (inhaled corticosteroids and long‐acting beta₂‐agonists) and tiotropium are both used in the treatment of chronic obstructive pulmonary disease (COPD). There is uncertainty about the relative benefits and harms of these treatments.

Objectives

To compare the relative effects of inhaled combination therapy and tiotropium on markers of exacerbations, symptoms, quality of life, lung function, pneumonia and serious adverse events in patients with chronic obstructive pulmonary disease.

Search methods

We searched the Cochrane Airways Group Specialised Register of trials (November 2012) and reference lists of articles. We also contacted authors of the studies.

Selection criteria

We included only parallel, randomised controlled trials comparing inhaled combination corticosteroid and long‐acting beta₂‐agonist against inhaled tiotropium bromide.

Data collection and analysis

Two authors independently assessed trials for inclusion and then extracted data on trial quality and outcome results. We contacted study authors for additional information. We resolved discrepancies through discussion.

Main results

One large, two‐year trial (INSPIRE) and two smaller, shorter trials on a total of 1528 participants were found. The results from these trials were not pooled. The number of withdrawals from each arm of the INSPIRE trial was large and imbalanced and outcome data were not collected for patients who withdrew, raising concerns about the reliability of data from this study.

In INSPIRE, there were more deaths on tiotropium than on fluticasone/salmeterol (Peto odds ratio (OR) 0.55; 95% confidence interval (CI) 0.33 to 0.93). This was a statistically significant difference, however the number of withdrawals from each of the arms was 11 times larger than the observed number of deaths for participants on fluticasone/salmeterol and seven times larger for participants on tiotropium. There were more all‐cause hospital admissions in patients on fluticasone/salmeterol than those on tiotropium in INSPIRE (Peto OR 1.32; 95% CI 1.04 to 1.67). There was no statistically significant difference in hospital admissions due to exacerbations, the primary outcome of INSPIRE. There was no significant difference in exacerbations in patients on fluticasone/salmeterol compared to tiotropium when compared as either an odds ratio or a rate ratio (mean number of exacerbations per patient per year). Exacerbations requiring treatment with oral corticosteroids were less frequent in patients on fluticasone/salmeterol (rate ratio 0.81; 95% CI 0.67 to 0.99). Conversely exacerbations requiring treatment with antibiotics were more frequent in patients treated with fluticasone/salmeterol (rate ratio 1.19; 95% CI 1.02 to 1.38). There were more cases of pneumonia in patients on fluticasone/salmeterol than in those on tiotropium (Peto OR 2.13; 95% CI 1.33 to 3.40). Confidence intervals for these outcomes do not reflect the additional uncertainty arising from unknown outcome data for patients who withdrew.

Authors' conclusions

Since the proportion of missing outcome data compared to the observed outcome data is enough to induce a clinically relevant bias in the intervention effect, the relative efficacy and safety of combined inhalers and tiotropium remains uncertain. Further large, long‐term randomised controlled trials comparing combination therapy to tiotropium are required, including adequate follow‐up of all participants randomised (similar to the procedures undertaken in TORCH and UPLIFT). Additional studies comparing alternative inhaled long‐acting beta₂‐agonist/steroid combination therapies with tiotropium are also required.

Plain language summary

Combined inhalers compared to tiotropium inhalers for the treatment of chronic obstructive pulmonary disease

Chronic obstructive pulmonary disease (COPD) is a general term referring to chronic bronchitis and emphysema, or both. COPD occurs when airflow to the lungs is restricted. Symptoms include cough and breathlessness and inhalers are commonly used to prevent and relieve these symptoms. COPD is usually caused by smoking and the best way to improve symptoms is to give up smoking.

COPD trials lasting longer than six months often have large numbers of people leaving the trial early. In INSPIRE, the largest trial in our review, comparing fluticasone/salmeterol to tiotropium, there were seven to 11 times more people leaving the trial early than the number who died; a number that swamps the death rate. Therefore we felt unable to draw a reliable conclusion as to which treatment has the lowest mortality rate. This uncertainty also left us unable to reliably say which drug was better in terms of reducing COPD exacerbations, hospitalisations and serious adverse events or improving quality of life and health status.

More information about COPD and explanations of terms used in this summary can be found here.

Background

Description of the condition

Chronic obstructive pulmonary disease (COPD) is characterised by chronic airflow limitation and limited response to short‐acting beta₂‐agonists, and is associated with a smoking history (GOLD). Symptoms of COPD include breathlessness and impaired exercise capacity. There are a number of commonly used pharmacological treatments in COPD management including inhaled short‐acting beta‐agonists (SABA), long‐acting beta₂‐agonists (LABA) (Appleton 2006a), inhaled corticosteroids (ICS) (Yang 2012) and anticholinergics such as tiotropium (Karner 2012a) and ipratropium bromide (Appleton 2006b; Appleton 2006c). Self management education and pulmonary rehabilitation should accompany these pharmacological interventions (Effing 2007; Lacasse 2006).

Description of the intervention

Combination therapy is a maintenance inhaler that delivers an inhaled corticosteroid and a long‐acting beta‐agonist concurrently at the same dose. Combination therapy is currently available as combination fluticasone and salmeterol (marketed as Seretide or Advair, GlaxoSmithKline) and budesonide and formoterol (marketed as Symbicort, AstraZeneca). Both combination products are licensed for use in COPD at the highest doses of ICS (daily dose fluticasone 1000 μg and budesonide 800 μg). Inhaled corticosteroids are anti‐inflammatories and long‐acting beta‐agonists cause smooth muscle relaxation resulting in bronchodilation.

Tiotropium bromide (marketed as 'Spiriva', Boehringer Ingelheim) is an inhaled long‐acting anticholinergic agent, and has gained widespread acceptance as a maintenance therapy in COPD (Barr 2005; GOLD; UPLIFT). Tiotropium is a long‐acting anticholinergic agent that targets bronchospasm in COPD by relaxing airway smooth muscle.

Why it is important to do this review

Both tiotropium and combination inhalers have been shown to improve key clinical indicators of disease in clinical trials against placebo.

Combination therapy has been shown to reduce exacerbations and mortality, and to improve health status compared to placebo (Nannini 2007a). Inhaled corticosteroids alone reduce the mean rate of COPD exacerbations and the decline in quality of life although do not have an effect on lung function (Yang 2012). Long‐acting beta‐agonists have also been shown to reduce exacerbations and improve lung function (Appleton 2006a). The effects of combination treatment against component monotherapy are more variable (Nannini 2012; Nannini 2007b). However, concerns have been raised about the risk of pneumonia associated with ICS in COPD patients (Singh 2009).

Benefits of tiotropium in comparison with placebo include reduced exacerbations and related hospital admissions, and improvements in quality of life and lung function (Barr 2005; UPLIFT). However, in the UPLIFT trial conducted over four years tiotropium did not slow the rate of decline in FEV₁ (forced expiratory volume in one second) compared with placebo (UPLIFT). Concerns that inhaled anticholinergics increase the long‐term risk of major cardiovascular events in COPD have been raised (Singh 2008b), but this was not found in UPLIFT.

This leaves clinicians and patients facing uncertainty as to the relative merits of these treatments, and how the side effect profiles of each compare.

Other relevant Cochrane reviews compare tiotropium with placebo (Karner 2012a), tiotropium with LABA (Chong 2012), tiotropium and LABA with either drug as monotherapy (Karner 2012), inhaled steroid and LABA and tiotropium compared to LABA and tiotropium (Karner 2011a), and inhaled steroids with LABA (Spencer 2011).

Objectives

To compare the relative effects of inhaled combination therapy and tiotropium on markers of exacerbations, symptoms, quality of life, lung function, pneumonia and serious adverse events in patients with chronic obstructive pulmonary disease.

Methods

Criteria for considering studies for this review

Types of studies

We included only randomised controlled trials with a parallel‐group design comparing the interventions. Studies were not excluded on the basis of blinding. We included studies of 12 weeks or more in duration.

Types of participants

Populations with a diagnosis of chronic obstructive pulmonary disease. We only included studies where an external set of criteria had been used to screen participants for this condition (e.g. GOLD, ATS, BTS, TSANZ).

Types of interventions

Inhaled combination corticosteroid and long‐acting beta₂‐agonist (such as fluticasone/salmeterol, budesonide/formoterol, beclomethasone/formoterol) versus inhaled tiotropium bromide.

Types of outcome measures

Primary outcomes

1. Mortality (all‐cause)

2. Hospital admission

3. Exacerbations; all‐cause, requiring short courses of oral corticosteroids or antibiotics as defined by agreed criteria

4. Pneumonia

Secondary outcomes

1. Quality of life (measured with a validated scale for COPD, e.g. St George's Respiratory Questionnaire, Chronic Respiratory Disease Questionnaire)

2. Symptoms

3. Forced expiratory volume in one second (FEV₁)

4. Non‐fatal serious adverse events

5. Adverse events

6. Withdrawals

Search methods for identification of studies

Electronic searches

We identified trials using the Cochrane Airways Group Specialised Register of trials, which is derived from systematic searches of bibliographic databases including the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, EMBASE, CINAHL, AMED and PsycINFO, and handsearching of respiratory journals and meeting abstracts (see Appendix 2 for further details). All records in the Specialised Register coded as 'COPD' were searched using the following terms:

(tiotropium or spiriva) AND (((budesonide or fluticasone or beclomethasone or mometasone or steroid* or corticosteroid*) and (formoterol or salmeterol or indacaterol or (beta* and agonist*))) or (symbicort or viani or seretide or advair or foster or fostair or inuvair or combination*))

We conducted the latest search in November 2012.

Searching other resources

We reviewed reference lists of all primary studies and review articles for additional references. We contacted authors of identified trials and manufacturers to ask if they knew of other published or unpublished studies.

Data collection and analysis

Selection of studies

Two review authors (EJW and CJC) screened the titles and abstracts of citations retrieved through literature searches and obtained full papers of references deemed to be potentially relevant. We assigned each reference to a study identifier and assessed them against the inclusion criteria of the review.

Data extraction and management

We extracted characteristics and data for studies that met the eligibility criteria of the review using a dedicated extraction sheet. The extraction of characteristics and data was performed in duplicate, and we identified and resolved discrepancies through discussion.

Assessment of risk of bias in included studies

We assessed the risk of bias according to recommendations outlined in the Cochrane Handbook for the following items:

1. Allocation sequence generation

2. Concealment of allocation

3. Blinding of participants and investigators

4. Handling of missing data

We graded each potential source of bias as high, low or unclear risk of bias.

Measures of treatment effect

We intended to combine dichotomous data variables (such as mortality, pneumonia and study withdrawal) as Peto odds ratios (OR) with 95% confidence intervals (CI) as this is more suitable than Mantel‐Haenszel for rare events. However, the events in INSPIRE were not rare. We cross‐checked all the Peto odds ratios with the Mantel‐Haenszel odds ratios and found no difference, so we reported the Peto odds ratio as per protocol. We planned to combine continuous outcome data (such as symptoms, quality of life and FEV₁) as fixed‐effect mean differences with 95% CI. In INSPIRE, exacerbations were reported as rate ratios (RR) and we entered these data into RevMan 5 using the GIV function (RevMan 2012).

Unit of analysis issues

Data on exacerbations were provided by INSPIRE trialists as the difference in exacerbation rates between patients on fluticasone/salmeterol and those on tiotropium, and were based on rate ratios using negative binomial model estimates and 95% confidence intervals. We entered the natural log of the rate ratio along with the standard error calculated from the confidence interval into RevMan 5 using the GIV function and cross‐checked the resulting confidence intervals with those provided by INSPIRE for exacerbations.

Data for quality of life and FEV₁ were calculated as a mean difference with 95% confidence intervals. We entered the mean difference and standard errors calculated from 95% CIs into RevMan 5 and analysed it using the GIV tool.

Dealing with missing data

We contacted investigators or study sponsors in order to verify key study characteristics and to obtain missing numerical data.

Assessment of heterogeneity

We planned to assess the amount of statistical variation between the study results with the I² statistic.

Data synthesis

We planned to calculate numbers needed to treat from the pooled odds ratio and its confidence interval, and apply them to appropriate levels of baseline risk.

We intended to present the findings of our four primary outcomes (mortality, hospital admission, exacerbations and pneumonia) in a 'Summary of findings' table generated using GradePro software.

Subgroup analysis and investigation of heterogeneity

We planned to subgroup studies according to:

1. type and dose of combination therapy;

2. severity of disease at baseline.

Sensitivity analysis

We intended to assess the sensitivity of our primary outcomes to degree of bias. We planned to compare the Peto odds ratio results for dichotomous outcomes with Mantel‐Haenszel fixed‐effect and random‐effects models and with the pooled risk differences.

Results

Description of studies

See Characteristics of included studies and Characteristics of excluded studies.

Results of the search

The initial search was carried out in October 2009 and identified 81 references. Of these 13 were potentially relevant and were obtained in full text for further assessment. Eight of these full‐text documents were eligible and belonged to three included studies (Dawber 2005, excluded in the 2012 update; INSPIRE; SCO40034). We asked five trialists if they knew of any other published and unpublished trials for the original version of this review; there were five responses but no relevant studies that had not already been located by the search were identified. Eight further references were returned by an updated search in March 2010, but none of these was eligible. A search in November 2012 identified 90 references of which one was eligible (Fang 2008).

Included studies

Participants

A total of 1528 participants were recruited into the three eligible studies (Fang 2008; INSPIRE; SCO40034). INSPIRE was by far the largest included trial, with 1323 participants, whilst SCO40034 had 125 participants and Fang 2008 had 80 patients in the relevant treatment arms. INSPIRE was a two‐year trial, Fang 2008 was a year long and SCO40034 was 12 weeks in duration. Owing to the disparity in the trial lengths and because the primary focus of our review was on long‐term outcomes, we did not pool the results of the trials. We could not find a contact email address for the author of Fang 2008, therefore we were unable to contact them to ask for clarification of the data. This was disappointing because if the trialists had been able to provide the raw data we could have used it. Due to our concerns about inconsistency and possible errors in the data we were unable to incorporate the data into meta‐analysis and therefore we added the data narratively where possible. This review therefore focusses primarily on the results of INSPIRE.

Patients in INSPIRE were classified as having GOLD stage III (FEV₁ ≥ 30% to < 50% predicted) or GOLD stage IV (FEV₁ < 30% predicted). There were N = 540 patients with stage III COPD on fluticasone/salmeterol with a mean FEV₁ of 1.09 L and N = 537 patients with a mean FEV₁ of 1.11 L on tiotropium. There were N = 100 patients with stage IV COPD on fluticasone/salmeterol with a mean FEV₁ of 0.73 L and N = 101 patient with a mean FEV₁ of 0.71 L on tiotropium. In INSPIRE, 48% of participants in the fluticasone/salmeterol arm and 51% in the tiotropium arm stopped taking inhaled corticosteroids at baseline. All participants had a smoking history of greater than 10 pack‐years.

Interventions

Patients recruited to INSPIRE received either fluticasone/salmeterol (500/50 µg twice a day) as a dry powder via a DISKUS or Accuhaler inhaler or tiotropium (18 μg once a day) delivered via a dry‐powder Handihaler.

Participants in INSPIRE were allowed to take short‐acting beta₂‐agonists and short courses of oral corticosteroids alongside their study medications. Prior to randomisation, participants in INSPIRE were given oral prednisolone (30 mg) once a day and inhaled fluticasone/salmeterol (500/50 μg) twice a day during a two‐week run‐in period.

Outcomes

The primary outcomes varied between the studies and were different from our primary outcomes. The primary outcome for INSPIRE was the rate of health care utilisation for COPD exacerbations, which we incorporated into our review. SCO40034 was an exploratory study to compare the clinical efficacy of fluticasone/salmeterol against tiotropium and therefore did not define a primary outcome.

Excluded studies

Five studies failed to meet the eligibility criteria for the review and reasons for exclusion are reported in the Characteristics of excluded studies table. Three trials (Golabi 2006; Hara 2007; Singh 2008) were excluded because they were cross‐over trials, which are not suitable for assessing long‐term outcomes.

One trial (Bateman 2008) was excluded because the intervention compared fluticasone (250 μg) and salmeterol (50 μg) administered via separate metered dose inhalers to tiotropium (18 μg). We felt that this could lead to discrepancies in the analysis if participants stopped taking one inhaler and continued taking the other. This randomised, double‐blind, triple‐dummy pilot study administered fluticasone and salmeterol to 51 participants and tiotropium to 56 participants. The participants had moderate or severe COPD and a smoking history of greater than 10 pack‐years although the baseline characteristics were not comparable across both arms. The primary outcome was FEV₁ and there was no significant difference in the lung function or occurrence of adverse effects between the two intervention arms after 43 days and there was a single dropout in the fluticasone/salmeterol arm. We excluded a trial that was included in the original version of this review (Dawber 2005) as we introduced a minimum trial duration of 12 weeks.

Risk of bias in included studies

An assessment of the risk of bias is presented in the Characteristics of included studies table and summarised in a 'Risk of bias' table (Figure 1).

1.

Methodological quality summary: review authors' judgements about each methodological quality item for each included study.

Allocation

INSPIRE reported adequate sequence generation and allocation concealment. SCO40034 did not report full details in the study report, but GSK supplied unpublished information describing adequate computerised randomisation and allocation on request. Fang 2008 stated that they used random number tables, but no information was provided on allocation concealment.

Blinding

INSPIRE and SCO40034 were blinded by employing a double‐dummy design; the two drugs were administered via different types of inhaler, each participant was given two inhalers to use each day, one containing the intervention medication and a second containing placebo. Neither the patient nor the investigator knew what medication a particular participant was randomised to. Blinding was not stated or described in Fang 2008, indeed the non‐intervention arm did not appear to have a placebo and we took this as confirmation that the relevant treatment arms were not blinded.

Incomplete outcome data

INSPIRE suffered from high withdrawal rates and data were not collected for patients who withdrew. For a discussion of the methodological issues affecting COPD trials see Discussion.

Selective reporting

Two trials adequately reported outcome data for the primary and secondary outcomes that they had pre‐specified in the study record (INSPIRE; SCO40034).

Effects of interventions

Primary outcome: Mortality (all‐cause)

In INSPIRE, there were more deaths on tiotropium (38/665 people) than on fluticasone/salmeterol (21/658) (Peto odds ratio (OR) 0.55; 95% confidence interval (CI) 0.33 to 0.93). Although this was a statistically significant difference, the number of withdrawals from each of the arms was 11 times larger than the number of deaths for participants on fluticasone/salmeterol and seven times larger for participants on tiotropium. This uncertainty about the results is not reflected in the confidence interval for the odds ratio.

Primary outcome: Hospital admission

There were more all‐cause hospital admissions in patients on fluticasone/salmeterol (215/658) than those on tiotropium (179/665) in INSPIRE (Peto OR 1.32; 95% CI 1.04 to 1.67). The primary outcome of INSPIRE was hospital admissions due to exacerbations. More patients on salmeterol/fluticasone were hospitalised due to exacerbations (105/658) compared to tiotropium (86/665) (Peto OR 1.28; 95% CI 0.94 to 1.74), but this was not statistically significant.

Primary outcome: Exacerbations

Data for all‐cause exacerbations were reported as both count data and as a rate (i.e. the mean number of exacerbations per year) in INSPIRE. The differences between these methods of analysis are discussed in the Summary of main results.

Firstly, there was no significant difference in the number of people experiencing one or more exacerbations (408/658 in the fluticasone/salmeterol arm versus 393/665 in the tiotropium arm; Analysis 1.3).

1.3. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 3 Exacerbations (all‐cause): number of patients experiencing one or more exacerbations over two years.

The number of exacerbations was also reported as the mean number of events per year. There was no significant difference in exacerbations of any type in patients on fluticasone/salmeterol compared to tiotropium (rate ratio (RR) 0.97; 95% CI 0.84 to 1.12; Analysis 1.4). Overall 62% of the fluticasone/salmeterol group and 59% of the tiotropium group had one or more exacerbation requiring therapeutic intervention and the trialists estimated this to equate to 1.28 and 1.32 exacerbation per year for patients on fluticasone/salmeterol and tiotropium respectively. Exacerbations requiring treatment with oral corticosteroids were more frequent in patients on tiotropium (RR 0.81; 95% CI 0.67 to 0.99) and conversely exacerbations requiring treatment with antibiotics were more frequent in patients treated with fluticasone/salmeterol (RR 1.19; 95% CI 1.02 to 1.38).

1.4. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 4 Exacerbations (mean number of exacerbations per patient per year).

Primary outcome: Pneumonia

There were more cases of pneumonia in patients on fluticasone/salmeterol (50/658) than in those on tiotropium (24/665) (Peto OR 2.13; 95% CI 1.33 to 3.40). There were several figures reported for pneumonia in the INSPIRE trial (see Table 1) all of which indicated that there were more cases of pneumonia in participants on fluticasone/salmeterol compared to those on tiotropium. While this is statistically significant, there is considerable uncertainty over the clinical interpretation of this result owing to the large dropout rates. Moreover, if this is a true finding, the higher pneumonia rate in the combination group was not associated with a significant increase in exacerbations, hospitalisations due to exacerbations or deaths.

1. Differences in the reported cases of pneumonia in the INSPIRE trial.

	Pneumonias recorded in patients treated with fluticasone/salmeterol Number (%)	Pneumonias recorded in patients treated with tiotropium Number (%)
Most frequent adverse effects on ITT population*	43 (7)	23 (3)
Serious adverse events*	37 (6)	22 (3)
Fatal serious adverse events*	3 (< 1)	0
Adverse events including pneumonia, lobar pneumonia and bronchopneumonia**	50 (8)	24 (4)

*Reported in SCO40035.

**Reported in Wedzicha 2008.

ITT: intention‐to‐treat

Secondary outcome: Quality of life

Patients on fluticasone/salmeterol reported better quality of life than those on tiotropium at two years from baseline (mean difference (MD) ‐2.07; 95% CI ‐4.02 to ‐0.12). In real terms, this meant that patients on salmeterol/fluticasone reported a mean improvement in quality of life of 1.7 units on the St. Georges Respiratory Questionnaire (SGRQ) whilst those on tiotropium reported a mean worsening in their quality of life of 0.4 units. The minimum clinically important difference in quality of life measured on the SGRQ is four units. INSPIRE reported that 35% of patients on fluticasone/salmeterol and 27% patients on tiotropium (OR 1.29; 95% CI 1.04 to 1.60) experienced an improvement in quality of life of ≥ four units. Data for the number of patients who had a deterioration in quality of life of ≥ four units was not reported (Jones 2009).

Secondary outcome: Symptoms

Outcome not reported.

Secondary outcome: Forced expiratory volume in one second (FEV₁)

FEV₁ data were available at a number of time points in INSPIRE and therefore we entered data at both eight weeks and two years. At eight weeks, FEV₁ had increased by 0.04 L (± 0.010) from baseline in patients on fluticasone/salmeterol, whilst those on tiotropium had an increase of 0.06 L (± 0.010). There was no statistically significant between‐group difference (MD ‐0.02; 95% CI ‐0.05 to 0.01). At two years, the FEV₁ of patients on fluticasone/salmeterol had decreased by ‐0.01 L (± 0.012) compared to baseline whilst those on tiotropium showed an improvement compared to baseline of FEV₁ 0.01 L (± 0.013), again not a statistically significant between‐group difference (MD ‐0.02; 95% CI ‐0.05 to 0.01).

Secondary outcome: Serious adverse events (non‐fatal)

There were more serious adverse events in patients on fluticasone/salmeterol (194/658) than in patients on tiotropium (141/665) (Peto OR 1.55; 95% CI 1.21 to 1.98) in INSPIRE.

Secondary outcome: Adverse events

In INSPIRE there were fewer adverse events among patients on fluticasone/salmeterol (435/658) than tiotropium (414/665) (Peto OR 1.18; 95% CI 0.94 to 1.48).

Secondary outcome: Withdrawal

There were large numbers of withdrawals from INSPIRE and the withdrawal rate was higher in the tiotropium group than in the fluticasone/salmeterol group. There were fewer withdrawals for any reason for patients on fluticasone/salmeterol (232/658, 35%) compared with tiotropium (279/665, 42%) (Peto OR 0.75; 95% CI 0.60 to 0.94). The difference between the number withdrawing due to adverse events (Peto OR 1.03; 95% CI 0.72 to 1.47) and lack of efficacy (Peto OR 0.84; 95% CI 0.52 to 1.37) was not statistically significant. The INSPIRE trialists reported reasons for withdrawal (Table 2) and this demonstrated that the greatest difference was higher rates of withdrawal due to COPD exacerbations in patients on tiotropium than in those on fluticasone/salmeterol.

2. Reasons for withdrawals from INSPIRE.

Reason for withdrawal	Fluticasone/salmeterol	Tiotropium
Adverse event	67	66
Withdrew consent	61	82
Lost to follow‐up	15	13
Protocol violation	7	8
Failed entry criteria	0	3
COPD exacerbation	37	51
Lack of efficacy	32	38
Other	13	17
Missing	0	1
Total	232 (35.3%)	279 (42.0%)

COPD: chronic obstructive pulmonary disease

There were no data reported for symptoms.

Summary of results for Fang 2008 and SCO40034

There were no deaths in SCO40034. There was no significant difference in hospital admissions, cases of pneumonia, FEV₁, serious adverse events or adverse events in SCO40034. There were more withdrawals in patients on tiotropium compared to fluticasone/salmeterol in patients in the SCO40034 trial.

Fang 2008 reported no significant difference in absolute number of adverse events between treatment groups. Pre‐ and post‐bronchodilator FEV₁ were reported at the end of study, but we were not confident that the number reported represented the standard error. SGRQ score was reported, but the numbers appeared to differ between what was in the text of the trial report and those in a table; we were therefore unable to use the data.

Discussion

Methodological issues in chronic obstructive pulmonary disease (COPD) trials

Withdrawal rates in COPD trials are commonly high, especially in studies longer than six months in duration. There is no consensus on how to handle participants for whom data are not available (Cochrane Handbook). There are two available options: intention‐to‐treat analysis or available case analysis.

The principles of intention‐to‐treat analyses are that participants are analysed according to the intervention group to which they were initially randomised regardless of the treatment received. Outcome data are recorded for all participants and all randomised participants are included in the analysis. Performing analyses on a true intention‐to‐treat basis when outcome data for all participants are missing is not possible. To perform an intention‐to‐treat analysis regardless of the missing data does not take into account participants who withdraw and gives an overly precise estimate of the treatment effect. When withdrawals are related to the treatment (i.e. not at random), the participants who withdraw are likely to have poorer outcomes than those who remain in the trial and provide data (Suissa 2008). This can lead to a 'healthy survivor' effect.

One way to deal with missing outcome data due to withdrawals is imputation and there are three ways of estimating missing values. One approach is to assume that the rate of events in the participants who withdraw occurs at the same rate as it does in those who remained in the study and provide data. This increases the precision of the estimate of the confidence intervals, but if the assumed rate of events is incorrect, the effect estimate for the study will be biased. A second option is to simulate best‐case worst‐case scenarios, where it is assumed that all the participants who withdrew from one intervention arm experienced the event and then repeat the analysis with the other intervention arm, comparing the results as a sensitivity analysis. This is problematic when the number of dropouts is significantly greater than the numbers of patients experiencing the event in the study. The third way is to assume that the rates of events in the withdrawals is similar to those observed in other similar trials.

However, full follow‐up of patients that withdraw can introduce a different bias. Patients that withdraw are likely to begin treatment with another medication, but their outcomes are still attributed to the treatment group to which they were randomised. This is particularly problematic in head‐to‐head studies because patients often go onto the other study medication, which could potentially provide an inaccurate estimate of the true difference between the treatments.

Available case analysis, where treatment effect is based on the number of participants who provided data, does not take into account the outcomes of those who withdrew.

COPD patients who withdraw early tend to be sicker at recruitment and deteriorate faster than those who remain in the study. Additionally, because COPD is a chronic condition, many patients are already taking medication when they start a clinical trial and withdrawal symptoms can occur (Suissa 2008). In effect a trial can be looking at the effect of withdrawing an existing treatment as well as introducing a new one.

Participants in COPD trials are usually already taking medication for their condition before entry. Run‐in periods, where participants take a standard treatment for a few weeks prior to randomisation were designed to account for the improvement in health status that comes at the start of a trial due to increased medical attention (Calverley 2003). However, these run‐in periods limit the overall applicability of the results if the outcomes were affected by the run‐in drugs, but attributed to the study medication (Suissa 2008a).

Summary of main results

The number of withdrawals from both arms of the INSPIRE trial was large and outcome data were not collected for patients who withdrew. It is feasible to record mortality data (vital status) for participants who withdrew, but not reasonably possible to obtain data for other outcomes. Because outcome data for those who withdrew were not available, the confidence intervals for the outcomes do not reflect this additional uncertainty. We cannot be sure that the mortality rate for patients who withdrew on either drug was higher, lower or the same as those who completed the study. It would be, in our opinion, inappropriate to apply simple imputation of data in this instance.

When patients withdraw from a study for reasons relating to outcomes, the perceived benefit (or lack thereof) of a study drug can have great influence over their decision to remain in the trial (Kesten 2007). Kesten 2007 reported higher incidence rates of death following premature discontinuation of study medication. INSPIRE trialists cited the differential withdrawal rate as an indirect marker of treatment efficacy. There was no significant difference in the withdrawals due to lack of efficacy or adverse events in INSPIRE although more people on tiotropium withdrew compared to those on fluticasone/salmeterol overall. The proportion of missing outcome data compared to the observed outcome data is enough to induce a clinically relevant bias in the intervention effect.

We presented exacerbations using two different units of analysis in this review. There are several ways to analyse exacerbation rates and each is associated with advantages or disadvantages (Keene 2008). Looking at the number of patients experiencing one or more exacerbations does not give any information about exacerbation frequency in the same patient and does not take into account duration of study (Karner 2011; Keene 2008). You therefore get different information from the number of people experiencing one or more exacerbations and the mean number of exacerbations per year. While neither of the methods of reporting all‐cause exacerbations resulted in a statistically significant difference, it is interesting to note that the direction of the treatment effect is different. Reported as number of people experiencing one or more exacerbations, there are fewer exacerbations in people on tiotropium. However looking at the relative rate, there are fewer exacerbations per year in patients on fluticasone/salmeterol. This may represent the play of chance, but we cannot rule out the possibility that the two treatments have different impacts on patients with occasional or frequent exacerbations.

Quality of the evidence

Two trials had adequate sequence generation, allocation concealment and blinding. One was unblinded and did not report how allocation concealment was attempted and we were not able to contact the trialist to find out. Additionally, data were provided for the outcomes outlined in the protocols of the trials. The principal concern with the largest study addressing the review question (INSPIRE) relates to the handling of data from participants who withdrew.

Potential biases in the review process

We minimised bias in our search process thorough using comprehensive search terms and asking authors to identify other published an non‐published studies. Studies were determined as included or excluded, data were extracted and risk of bias attributed in duplicate to minimise error.

Agreements and disagreements with other studies or reviews

Previous systematic reviews evaluating combined inhalers compared to placebo (Nannini 2007a) and combined inhalers compared to long‐acting beta₂‐agonists (Nannini 2012) have shown an elevated risk of pneumonia with combination therapy (see Table 3). There were relatively more cases of pneumonia reported in these systematic reviews (which included TORCH) and TORCH than were recorded in INSPIRE.

3. Proportion of participants developing pneumonia.

Study	Pneumonia on FPS	Comparison treatment	Pneumonia on comparison treatment	Trial duration, weeks	Pneumonia on FPS per 52 weeks	Pneumonia on comparison per 52 weeks
Naninni 2007a†	325/2673 (12%)	Placebo	194/2556 (8%)	105*	5.9%	4.0%
TORCH	303/1546 (20%)	Placebo	164/1544 (11%)	156	6.7%	3.7%
Naninni 2012†	447/5922 (7%)	LABA	280/5154 (5%)	92*	4.2%	3.1%
TORCH	303/1546 (20%)	LABA	205/1542 (13%)	156	6.7%	4.3%
INSPIRE	50/658 (8%)	Tiotropium	24/665 (4%)	104	3.8%	1.8%

^†Includes data from TORCH.

*Weighted mean trial duration.

FPS: fluticasone/salmeterol 
 LABA: long‐acting beta₂‐agonist

There were relatively fewer deaths in INSPIRE compared to TORCH that included complete follow‐up of vital status for participants who withdrew; 1.6% per year in INSPIRE compared to 4.2% per year in TORCH. There may have been at least as many deaths in patients who withdrew from INSPIRE as in those who completed the trial.

Indirect comparison

There are two Cochrane reviews that may serve as a useful indirect comparison of treatment efficacy: combination inhaled steroids and long‐acting beta‐agonists versus placebo (Nannini 2007a) and tiotropium versus placebo (Barr 2005). We include the following descriptions for comparison only and have not calculated an estimate of relative benefits and harms from these meta‐analyses.

Nannini 2007a reported outcomes for a total of 6427 participants in 11 studies. Two studies compared budesonide/formoterol, while the remaining studies compared fluticasone/salmeterol, albeit at different doses. The meta‐analysis was dominated by the largest trial (TORCH) which had complete follow‐up for vital status of all participants. All‐cause mortality was reduced in patients on combined inhalers compared to placebo (OR 0.79; 95% CI 0.65 to 0.96). Exacerbations were less frequent in participants on combined inhalers (rate ratio 0.74; 95% CI 0.7 to 0.8) and an increase in the risk of pneumonia was noted in the groups that received inhaled steroids either alone or in combination.

Barr 2005 reported outcomes for a total of 6584 participants across nine studies, comparing tiotropium to placebo, ipratropium or long‐acting beta₂‐agonist. All‐cause mortality was reduced in participants on tiotropium compared to placebo (OR 0.73; 95% CI 0.35 to 1.49) and COPD exacerbations were also reduced (OR 0.75; 95% CI 0.66 to 0.85). The only adverse event data available to pool was for dry mouth which was more frequently experienced by patients randomised to tiotropium than placebo.

Authors' conclusions

Implications for practice.

INSPIRE had a high and unbalanced withdrawal rate. The proportion of missing outcome data compared to the observed outcome data is enough to induce a clinically relevant bias in the intervention effect. The relative efficacy and safety of combined inhalers and tiotropium remains uncertain.

Implications for research.

Further large, long‐term randomised controlled trials comparing combination therapy to tiotropium are required, including follow‐up of all participants randomised (similar to TORCH and UPLIFT). Additional studies comparing alternative inhaled long‐acting beta₂‐agonist/steroid combination therapies with tiotropium are also required.

Feedback

Clarification on mortality data and analysis of exacerbation rates requested, 10 January 2011

Summary

We have read this important review with great interest. In our assessment of the review we have several questions/comments.

1. It is stated that the INSPIRE trial investigators did not collect outcome events for patients who withdrew during the course of the study. We looked at the original INSPIRE publication (Wedzicha 2008) to confirm that this was the case. The INSPIRE investigators state "Mortality data was not collected after patients withdrew from therapy as in the TORCH (Toward a Revolution in COPD Health) study." This does not specifically state that other outcome events such as exacerbations were not collected for patients who withdrew prematurely. We wanted to know if this point was clarified with Wedzicha et al. We had assumed exacerbations were collected if they occurred after withdrawal but now we are not sure. Your help would be greatly appreciated.

2. For the INSPIRE trial rate ratios (RR) are given for the exacerbation analyses. We are not sure that sufficient explanations were given as to the difference in interpretation between rate ratios and relative risk ratios. This would be useful for clinician readers to know.

3. We congratulate the authors of this review for stating that conclusions are difficult to draw when there is so much missing data.

Reply

1. We did not confirm whether or not the trialists collected data on outcomes other than mortality in patients who withdrew therapy with the trialists before publication of this systematic review. Since receiving this feedback, we have been in contact with the pharmaceutical company who confirmed that there was no follow‐up of patients if they withdrew form the study for other outcome data such as exacerbations. Because patients were followed up until the resolution of a serious adverse event, there is some limited mortality data for deaths which occurred after cessation of treatment which the trialist confirmed was reported in the original publication and therefore this review.

2. The main difference between Rate Ratios and Risk Ratios (or Odds Ratios), is that Rate Ratios include multiple exacerbations from individual patients, whereas the unit of analysis for Risk and Odds ratios is patients with one or more exacerbations. We have now included a forest plot to indicate the result achieved using Odds Ratios and have added a discussion around the differences in reporting is included in the Discussion section.

We thank Aaron and Elsa for their comments, in particular the second comment which allowed us to explain the issue of rate ratio versus relative risk ratios within our review

Contributors

Aaron M Tejani and Elsa Liu

Submitter agrees with default conflict of interest statement: I certify that I have no affiliations with or involvement in any organization or entity with a financial interest in the subject matter of my feedback.

What's new

Date	Event	Description
30 November 2012	New search has been performed	New literature search run
30 November 2012	New citation required and conclusions have changed	One new study added. We added a minimum study duration of 12 weeks, therefore we excluded Dawber 2005

History

Protocol first published: Issue 3, 2009
 Review first published: Issue 5, 2010

Date	Event	Description
1 June 2011	Amended	Feedback incorporated
1 June 2011	Feedback has been incorporated	We included an analysis of exacerbations using odds ratios and compared this to the rate ratio analysis we originally presented. Please see Discussion and Feedback.

Appendices

Appendix 1. Definition of serious adverse events

The Expert Working Group (Efficacy) of the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use (ICH) define serious adverse events as follows (ICHE2a 1995):

"A serious adverse event (experience) or reaction is any untoward medical occurrence that at any dose:

• Results in death,

• Is life‐threatening,

• Requires inpatient hospitalisation or prolongation of existing hospitalisation,

• Results in persistent or significant disability/incapacity, or

• Is a congenital anomaly/birth defect.

NOTE: The term "life‐threatening" in the definition of "serious" refers to an event in which the patient was at risk of death at the time of the event; it does not refer to an event which hypothetically might have caused death if it were more severe."

Appendix 2. Sources and search methods for the Cochrane Airways Group Specialised Register (CAGR)

Electronic searches: core databases

Database	Frequency of search
MEDLINE (Ovid)	Weekly
EMBASE (Ovid)	Weekly
CENTRAL (The Cochrane Library)	Quarterly
PsycINFO (Ovid)	Monthly
CINAHL (EBSCO)	Monthly
AMED (EBSCO)	Monthly

Handsearches: core respiratory conference abstracts

Conference	Years searched
American Academy of Allergy, Asthma and Immunology (AAAAI)	2001 onwards
American Thoracic Society (ATS)	2001 onwards
Asia Pacific Society of Respirology (APSR)	2004 onwards
British Thoracic Society Winter Meeting (BTS)	2000 onwards
Chest Meeting	2003 onwards
European Respiratory Society (ERS)	1992, 1994, 2000 onwards
International Primary Care Respiratory Group Congress (IPCRG)	2002 onwards
Thoracic Society of Australia and New Zealand (TSANZ)	1999 onwards

Condition search

1. exp Asthma/

2. asthma$.mp.

3. (antiasthma$ or anti‐asthma$).mp.

4. Respiratory Sounds/

5. wheez$.mp.

6. Bronchial Spasm/

7. bronchospas$.mp.

8. (bronch$ adj3 spasm$).mp.

9. bronchoconstrict$.mp.

10. exp Bronchoconstriction/

11. (bronch$ adj3 constrict$).mp.

12. Bronchial Hyperreactivity/

13. Respiratory Hypersensitivity/

14. ((bronchial$ or respiratory or airway$ or lung$) adj3 (hypersensitiv$ or hyperreactiv$ or allerg$ or insufficiency)).mp.

15. ((dust or mite$) adj3 (allerg$ or hypersensitiv$)).mp.

16. or/1‐15

17. exp Aspergillosis, Allergic Bronchopulmonary/

18. lung diseases, fungal/

19. aspergillosis/

20. 18 and 19

21. (bronchopulmonar$ adj3 aspergillosis).mp.

22. 17 or 20 or 21

23. 16 or 22

24. Lung Diseases, Obstructive/

25. exp Pulmonary Disease, Chronic Obstructive/

26. emphysema$.mp.

27. (chronic$ adj3 bronchiti$).mp.

28. (obstruct$ adj3 (pulmonary or lung$ or airway$ or airflow$ or bronch$ or respirat$)).mp.

29. COPD.mp.

30. COAD.mp.

31. COBD.mp.

32. AECB.mp.

33. or/24‐32

34. exp Bronchiectasis/

35. bronchiect$.mp.

36. bronchoect$.mp.

37. kartagener$.mp.

38. (ciliary adj3 dyskinesia).mp.

39. (bronchial$ adj3 dilat$).mp.

40. or/34‐39

41. exp Sleep Apnea Syndromes/

42. (sleep$ adj3 (apnea$ or apnoea$)).mp.

43. (hypopnoea$ or hypopnoea$).mp.

44. OSA.mp.

45. SHS.mp.

46. OSAHS.mp.

47. or/41‐46

48. Lung Diseases, Interstitial/

49. Pulmonary Fibrosis/

50. Sarcoidosis, Pulmonary/

51. (interstitial$ adj3 (lung$ or disease$ or pneumon$)).mp.

52. ((pulmonary$ or lung$ or alveoli$) adj3 (fibros$ or fibrot$)).mp.

53. ((pulmonary$ or lung$) adj3 (sarcoid$ or granulom$)).mp.

54. or/48‐53

55. 23 or 33 or 40 or 47 or 54

Filter to identify RCTs

1. exp "clinical trial [publication type]"/

2. (randomised or randomised).ab,ti.

3. placebo.ab,ti.

4. dt.fs.

5. randomly.ab,ti.

6. trial.ab,ti.

7. groups.ab,ti.

8. or/1‐7

9. Animals/

10. Humans/

11. 9 not (9 and 10)

12. 8 not 11

The MEDLINE strategy and RCT filter are adapted to identify trials in other electronic databases

Data and analyses

Comparison 1. Fluticasone/salmeterol (FPS) versus tiotropium (TIO).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1 Mortality (all‐cause)	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
2 Hospital admission	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
2.1 Hospital admissions all‐cause	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	0.0 [0.0, 0.0]
2.2 Hospital admissions resulting from exacerbations	1		Peto Odds Ratio (Peto, Fixed, 95% CI)	0.0 [0.0, 0.0]
3 Exacerbations (all‐cause): number of patients experiencing one or more exacerbations over two years	1		Odds Ratio (M‐H, Fixed, 95% CI)	Totals not selected
4 Exacerbations (mean number of exacerbations per patient per year)	1		Rate Ratio (Fixed, 95% CI)	Totals not selected
4.1 Exacerbations (all‐cause)	1		Rate Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
4.2 Exacerbations requiring oral corticosteroids	1		Rate Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
4.3 Exacerbations requiring antibiotics	1		Rate Ratio (Fixed, 95% CI)	0.0 [0.0, 0.0]
5 Pneumonia	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
6 Quality of life	1		Mean Difference (Fixed, 95% CI)	Totals not selected
6.1 32 weeks	1		Mean Difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
6.2 104 weeks	1		Mean Difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
7 FEV1 (litres)	2		Mean Difference (Fixed, 95% CI)	Totals not selected
7.1 FEV1 at short time frame (3 to 12 weeks)	2		Mean Difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
7.2 FEV1 at 2 years	1		Mean Difference (Fixed, 95% CI)	0.0 [0.0, 0.0]
8 Serious adverse events (non‐fatal)	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
9 Adverse events	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
10 Withdrawal	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	Totals not selected
10.1 Total number of subjects withdrawn	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	0.0 [0.0, 0.0]
10.2 Due to adverse events	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	0.0 [0.0, 0.0]
10.3 Due to lack of efficacy	2		Peto Odds Ratio (Peto, Fixed, 95% CI)	0.0 [0.0, 0.0]

1.1. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 1 Mortality (all‐cause).

1.2. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 2 Hospital admission.

1.5. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 5 Pneumonia.

1.6. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 6 Quality of life.

1.7. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 7 FEV₁ (litres).

1.8. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 8 Serious adverse events (non‐fatal).

1.9. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 9 Adverse events.

1.10. Analysis.

Comparison 1 Fluticasone/salmeterol (FPS) versus tiotropium (TIO), Outcome 10 Withdrawal.

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Fang 2008.

Methods	Design: randomised, parallel. 18 months duration. Per protocol analysis. Patients recruited from outpatients and inpatients of The First Affiliated Hospital of Kunming Medical Collage, China Run‐in: 2‐week washout prior to study starting
Participants	Population: diagnosis of COPD onset at least once a year in recent 3 years Age salmeterol/fluticasone group: 62 ± 7 years; tiotropium group: 63 ± 7 Salmeterol/fluticasone group: male 31, female 9; tiotropium group: male 28, female 12 Inclusion criteria:  1) FEV1 is 25% to 70 % predicted pre‐bronchodilator; 2) FEV1/FVC < 70%; 3) FEV1 increased by < 15% predicted post‐bronchodilator (400 µg salbutamol) Exclusion criteria: 1) diseases other than COPD; 2) need long‐term oxygen therapy; 3) currently receiving system hormone therapy
Interventions	Salmeterol 50 µg, fluticasone 250 µg Tiotropium dry powder 18 µg
Outcomes	FVC, FEV1, PEF, worsening (judged by counting the use of hormone or antibiotics amount; St. George’s Respiratory Questionnaire (SGRQ); adverse events (fungus infection, nausea, vomiting, headache, vertigo, dry mouth) Follow‐up at 0, 2, 4, 8, 16, 24, 32, 40 and 52 weeks Values reported before and after treatment
Notes
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Random number table was used
Allocation concealment (selection bias)	Unclear risk	Not described
Blinding (performance bias and detection bias) All outcomes	High risk	Blinding not mentioned, treatment group translated as 'blank' therefore assume not blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	20% withdrew on salmeterol/fluticasone, 20% withdrew on tiotropium

INSPIRE.

Methods	Design: a randomised, double‐blind, double‐dummy, multi‐centre, parallel‐group study over 104 weeks from June 2003 to February 2006 at 173 centres in 20 European countries (Austria, Belgium, Czech Republic, Denmark, Estonia, Germany, Greece, Italy, Latvia, Lithuania, Netherlands, Norway, Romania, Russia, Slovak Republic, Slovenia, Spain, Sweden, Ukraine and the UK) Run‐in: 2 weeks on oral prednisolone and salmeterol
Participants	Population: 1323 adults with a clinical history of severe and very severe COPD (GOLD stage III and IV). FEV1 less than 50% for inclusion into the trial Baseline characteristics: mean age 64 years. FEV1 39% predicted. Inhaled corticosteroids used previously by 50% of participants. Exacerbation in previous 12 months in 86% of participants. 48% of participants on FPS and 51% on tiotropium stopped taking inhaled corticosteroids at baseline Inclusion criteria: aged 40 to 80 years, with a smoking history of 10 or more pack‐years, a clinical history of COPD exacerbations, post‐bronchodilator FEV1 less than 50% of predicted, bronchodilator reversibility of less than 10% in FEV1  to 400 mg salbutamol, score of 2 or more on the Modified Medical Research Council dyspnoea scale Exclusion criteria: asthma or atopic disease, a lung disease likely to confound the drug response other than COPD, a recent exacerbation (within 6 weeks of screening or during run‐in); receiving long‐term oxygen therapy or pulmonary rehabilitation or had a known or suspected hypersensitivity to beta2‐agonists, inhaled corticosteroids, anticholinergic agents or any components of these formulations
Interventions	Combination of fluticasone 500 µg and salmeterol 50 µg twice a day via DISKUS/ACCUHALER inhaler plus placebo capsules to match TIO delivered once daily via the Handihaler inhaler Tiotropium 18 μg once a day via Handihaler plus placebo to match FPS DISKUS/ACCUHALER combination product delivered twice daily
Outcomes	Primary outcome: to compare the rate of health care utilisation (HCU) COPD exacerbations in those using fluticasone/salmeterol (500/50 μg) versus those using tiotropium 18 μg
Notes	Sponsored by GlaxoSmithKline Participants were allowed to use short‐acting inhaled beta2‐agonists and standardised short courses of oral corticosteroids
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated central randomisation list. Stratified allocation by centre and smoking. Block size of 4
Allocation concealment (selection bias)	Low risk	Telephone interactive voice response system
Blinding (performance bias and detection bias) All outcomes	Low risk	Double‐blind, double‐dummy design Due to a difference in appearance between tiotropium bromide and placebo inhalation capsules, study medication was dispensed by someone not directly involved in trial. Neither the investigator nor site personnel were present when the subject administered his/her study medication. Subjects were instructed not to show their study medication to other subjects. Decisions to withdraw subjects due to adverse events were made before unblinding. Subjects were unblinded only in emergencies where knowledge of the investigational product was essential for the clinical management or welfare of the subject. Emergency unblinding was done via an automated telephone system.
Incomplete outcome data (attrition bias) All outcomes	High risk	On FPS 35% withdrew from the study and on tiotropium 42% withdrew. Trialists provided an adequate breakdown of reasons for withdrawal. Differential withdrawal rates in the 2 arms may have introduced bias into all the outcome assessments.

SCO40034.

Methods	Design: a randomised, double‐blind, double‐dummy, multi‐centre, parallel‐group exploratory study over 12 weeks from March 2003 to October 2003 at 17 centres in the Netherlands Run‐in: all oral, slow‐release and long‐acting beta2‐agonists, inhaled corticosteroids, anticholinergics and short‐acting beta2‐agonists were discontinued. Subjects were issued with either a VENTOLIN DISKUS/ACCUHALER inhaler (200 μg/actuation) or VENTOLIN MDI (100 μg/actuation) plus VOLUMATIC spacer for use as relief medication
Participants	Population: 125 adults with a clinical history of moderate to severe COPD as defined by the Global Initiative for Obstructive Lung Disease 2001 guidelines Inclusion criteria:  aged 40 to 80 years inclusive. Post‐bronchodilator FEV1 less than 70% of predicted normal. Participants must have had a smoking history (current or former smokers) of more than 10 pack‐years. Mean FEV1 1.4 L Exclusion criteria: within 4 weeks prior to visit 1; COPD exacerbation; received oral, parenteral or depot corticosteroids for a COPD exacerbation; received antibiotic therapy and/or been hospitalised for either a lower respiratory tract infection or for COPD exacerbation, or had any changes in their COPD medication
Interventions	Combination of fluticasone 500 µg and salmeterol 50 µg twice a day via DISKUS inhaler plus placebo capsules to match TIO delivered once daily via the Handihaler inhaler Tiotropium 18 μg once a day via Handihaler plus placebo to match FPS DISKUS combination product delivered twice daily
Outcomes	No primary outcomes as this was an exploratory study
Notes	Sponsored by GlaxoSmithKline
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"At randomisation (Visit 2/2A) all eligible subjects were randomly assigned to treatment by use of a Registration and Material Ordering System (RAMOS) which utilized an IVRS developed by GSK.  Subjects were assigned to a unique treatment number." "Random allocation to study drug was stratified according to smoking status of subjects at entry (current smoker, former smoker) on a 1:1 basis."
Allocation concealment (selection bias)	Low risk	"The treatment number was an identification number for the blinded study medication and was assigned from a randomisation schedule provided by GSK. This schedule was generated by a GSK randomisation program, Patient Allocation for Clinical Trials (PACT) and then stored in another GSK program known as RandAll. Treatment numbers were not assigned to a subject without contacting RAMOS, and once a treatment number had been assigned to a subject, it was not  reassigned to another subject in this study."
Blinding (performance bias and detection bias) All outcomes	Low risk	"Double blind double dummy." There was a difference between the active tiotropium bromide and matching placebo capsules. Therefore someone who was not directly involved in the study received and documented all returned medication in a drug accountability log, a separate accountability log was maintained for each subject and subjects administered their own study medication without the investigator or site personnel being present Subjects were unblinded only when knowledge of the treatment was essential for the clinical management or welfare of the subject. Cases of unblinding were to be reported and documented immediately. Patients experiencing serious adverse events were unblinded. Decisions by the investigator to withdraw the subject due to an AE were made before unblinding. Blinded study drug was supplied in treatment packs at Visits 2/2A, 4 and 5. The content of each treatment pack was detailed on the outer packaging
Incomplete outcome data (attrition bias) All outcomes	High risk	117/125 (94%) completed the study, but withdrawals were imbalanced with 1 from the FPS arm and 7 from the tiotropium arm

AE: adverse event 
 COPD: chronic obstructive pulmonary disease 
 FEV₁: forced expiratory volume in one second 
 FPS: fluticasone/salmeterol 
 FVC: forced vital capacity 
 PEF: peak expiratory flow 
 TIO: tiotropium

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Bateman 2008	This pilot study compared separate fluticasone and salmeterol inhalers to tiotropium
Biscione 2009	Study 4 weeks duration
Dawber 2005	Study 3 weeks duration
Golabi 2006	Study of cross‐over design
GSK 2009b	Study 3 weeks duration
Hanania 2012	Triple therapy: benefits of adding fluticasone propionate/salmeterol to tiotropium in moderate to severe COPD
Hara 2007	Study of cross‐over design
Hoshino 2011	Triple therapy versus tiotropium
Jung 2012	Triple therapy versus tiotropium
Kurashima 2009	Study of cross‐over design
Magnussen 2010	Tiotropium + salmeterol versus fluticasone + salmeterol
Maltais 2012	Triple therapy
Mittmann 2011	Triple therapy
NCT00864812	Triple therapy
NCT01124422	Triple therapy
NCT01397890	Triple therapy
Peters 2010	Study of cross‐over design
Singh 2008	Study of cross‐over design
Tanabe 2012	Tiotropium versus ICS
Welte 2009c	Triple therapy
Williamson 2010	Study of cross‐over design

COPD: chronic obstructive pulmonary disease 
 ICS: inhaled corticosteroids

Differences between protocol and review

We changed some of the primary and secondary outcomes from those stated in the protocol:

• "Exacerbations requiring antibiotics or short burst oral corticosteroids as defined by agreed criteria" was changed to "Exacerbations; all‐cause, requiring short burst oral corticosteroids or antibiotics as defined by agreed criteria". We felt it was helpful to include data for the total number of exacerbations and this reflected the data reported in INSPIRE.

• In addition to all‐cause hospital admissions, we included data for exacerbations resulting in hospital admissions to reflect the primary outcome of INSPIRE.

• We added withdrawal as a secondary outcome because the high and differential withdrawal rate was important.

Exacerbations, quality of life and FEV₁ data were entered using generic inverse variance because the data presented in the study report required it.

For the 2012 update, we added a minimum duration of study of 12 weeks to match inclusion criteria with our other Cochrane reviews on tiotropium.

Contributions of authors

Studies were assessed by CJC and EJW. CJC and EJW extracted data, entered it into RevMan and conducted the analysis. EJW wrote the review with input from CJC and PP.

CJC and PP developed the protocol.

Toby Lasserson was an author on the protocol but is not an author on the review.

Sources of support

Internal sources

• St George's University of London, UK.

External sources

• Emma Welsh and Chris Cates, UK.

  NIHR Programme Grant funding

Declarations of interest

None known.

New search for studies and content updated (conclusions changed)