Mortality associated with tiotropium mist inhaler? A critical appraisal of the authors' selection and use of previously communicated tiotropium Respimat data
19 July 2011
Singh et al. recently reported the results of a meta-analysis of five previously published randomized controlled trials communicating a statistically significant increase in mortality risk associated with tiotropium mist inhaler in patients with COPD (1). The authors conclude that this meta-analysis "explains safety concerns by regulatory agencies".
Boehringer Ingelheim (BI) and Pfizer disagree with the authors' interpretation of the data. The companies believe that the benefits and risks associated with tiotropium Respimat have been appropriately communicated through labelling to patients and physicians. All trials considered for the meta-analysis have been made public either as a full publication by the study investigators or as a website posting by BI/Pfizer. Pooled-analyses based on the same dataset, yet evaluating patient-level data, had been conducted by BI/Pfizer and were submitted to regulatory authorities worldwide by 2009 with a request to modify labelling, and were made public (2, 3, 4). The primary analysis was based on the marketed dose in 4 major studies. The reported numerical, non- statistically significant increase in all cause mortality was accepted as an adequate representation of the data and included in the updated Product Information in 2010, e.g. in the European Union stating: "In a retrospective pooled analysis of the three 1-year and one 6-month placebo- controlled trials with Spiriva Respimat including 6,096 patients a numerical increase in all-cause mortality was seen in patients treated with Spiriva Respimat (68; incidence rate (IR) 2.64 cases per 100 patient- years) compared with placebo (51, IR 1.98) showing a rate ratio (95% confidence interval) of 1.33 (0.93, 1.92) for the planned treatment period; the excess in mortality was observed in patients with known rhythm disorders." (5) Therefore, the meta-analysis recently reported by Singh et al does not convey new information as the numerical mortality imbalance has been included in the Product Information, which also includes updated beneficial efficacy information for tiotropium Respimat, e.g. a significant reduction of hospitalized exacerbations of COPD in the major 1 -year study (5, 6).
We noticed inconsistencies in Singh et al's selection and use of available data as well as in the analytical approach and derived conclusions, which merit consideration and can explain the differences between the Singh analysis and the BI/Pfizer pooled-analysis.
First, Singh et al. have incorrectly assigned fatal cases to treatment groups, one excess case in the tiotropium 5 mcg group (12-week study BI code 205.251/2 (3, 7)), one excess case in the tiotropium 10 mcg group and one missing case in the placebo group (1 year study BI code 205.255 (8)). After correcting the fatal event counts, non-significant differences between the treatment groups result from a range of sensitivity analyses except one (see Table 1).
Second, the primary analysis of Singh et al. included only 5 of 6 studies of more than 4 weeks duration reportedly due to missing detail on trial characteristics. The excluded 6 months study was a randomised, double blind, placebo controlled study including the approved dose (5 mcg) of tiotropium Respimat and placebo as comparators (BI study code 1205.14 (3, 9)). The study was comparable in design and patient selection to the tiotropium programme and included pre-planned follow-up of patients who were discontinued early.
Third, the primary analysis of Singh et al. combined two doses, the marketed 5 mcg dose and a 10 mcg dose. Singh et al. refer to the Cochrane Handbook (10), which recommends "amalgamation of relevant treatment arms into one group". However, one may question why a primary analysis would include a non-marketed higher dose for the safety assessment of the approved and marketed dose. Notably, the Product Information of tiotropium Respimat advises that the recommended dose of 5 mcg/day should not be exceeded (5). We recognize, though, that inclusion of the higher than marketed dose may be appropriate to serve as a sensitivity analysis.
Fourth, the authors combined studies with different standards of vital status follow-up of early discontinued patients, in which different censoring rules were applied. In many COPD studies, active treatment arms, e.g. tiotropium, showed a lower rate of premature discontinuation than placebo, and patients who were discontinued early typically had more serious COPD than completers. Such differential discontinuation bias can induce a healthy survivor effect in control arms. Therefore, vital status follow-up of all patients and the use of patient level rather than study level data (e.g. rate-ratios adjusted for different exposure in treatment arms) are of great importance for the validity of pooled data.
When following the analytical approach of Singh et al., but using the corrected counts for fatal events, all analyses in Table 1, except one, show a numerical difference in mortality with a confidence interval including 1, so descriptive statistical significance was not observed. The one exception, with a confidence interval not including 1, is the approach published by Singh et al. as the primary analysis, which includes the higher not marketed dose plus exclusion of the 6-month study plus using a fixed effects approach.
Table 1: Sensitivity analyses within 6-study dataset, risk ratios (RR) with 95% confidence intervals (CI) based on Mantel-Haenszel analysis.
* 1-year studies censored day 369 as reported by Bateman (8), 1-year study censored day 337 as
reported by Bateman (6), 3-months studies on treatment as reported by Voshaar (7), 6-month study
B.I. code 1205.14, on treatment (3, 9).
# Primary analysis following Singh et al. using corrected numbers of fatal events. Singh et al. had published RR 1.50, CI 1.05, 2.15 (1)
BI/Pfizer had based their primary pooled analysis on patient level data of the marketed dose of 5 mcg in the four studies, which include vital status assessment of prematurely discontinued patients (risk calculated as rate ratios, cases censored for time of planned exposure + 1 day, i.e. 337 days in 1-year studies and 169 days in the 6-months study). The numerical difference of RR 1.33, CI 0.93, 1.92 is within the range of results in Table 1, which means the approach by Singh et al. when using correct data can be regarded a sensitivity analysis to the earlier BI/Pfizer approach.
Lastly, Singh et al. hypothesize that the mortality imbalance may be attributed to increased exposure to tiotropium administered through Respimat Soft Mist Inhaler compared to tiotropium Handihaler (1). However, a 16% greater exposure at steady state with widely overlapping ranges of plasma levels describes no meaningful difference in exposure. Furthermore, peak plasma levels, which rapidly decline, have no impact, because tiotropium is a slowly equilibrating compound at muscarinic receptors (11). Notably, another pharmacokinetic investigation showed comparable systemic exposure with the two formulations (12).
The unexplained numerical increase in all-cause mortality compared to placebo was concentrated in patients with cardiac arrhythmia at entry into the study. However, the causes of death were diverse and a causal relationship between the use of tiotropium Respimat and mortality has not been established. Important differences between the two formulations, tiotropium Respimat and Handihaler, could not be identified. Variability of outcomes and trial related factors may also explain the finding.
A rigorous investigation between the tiotropium Respimat and Handihaler formulations in a well controlled large outcome study, including all-cause mortality, is warranted to investigate any pharmacological differences between the formulations and for a reliable determination of benefit and risk. Such a study must include an active comparator with established safety to avoid ethical concerns with prolonged placebo treatment and bias by trial related factors. TIOSPIR a large-scale, prospective, randomized outcome study comparing tiotropium Respimat 5 mcg, 2.5 mcg and Handihaler 18 mcg was designed to assess these factors. The protocol was approved by health authorities in 50 countries and the study is supervised by an independent Data Safety Monitoring Board with access to fully unblinded data, which recommended continued study conduct, as planned, in June 2011. To date the exposure in this trial of more than 10,000 patient-years already exceeds the exposure reported in both, the BI/Pfizer and Singh et al. pooled analyses for tiotropium Respimat.
The selected active comparator, tiotropium Handihaler, has been extensively characterised in several prospective, large, well controlled (placebo and active) trials and pooled-analyses including 26 studies, one of which was a 4-year study. This analysis of patient level data from more than 17,000 patients showed a rate ratio for all-cause mortality of 0.85, CI 0.75-0.97 (tiotropium HandiHaler vs. placebo, rate ratio calculated from rate differences as published (13)). Potential cardiovascular and mortality risks of tiotropium Handihaler were recently assessed by a US Food and Drug Administration Advisory Committee (14), who confirmed that the safety of the product is adequately documented. This conclusion is also reflected in a publication by FDA staff members (Michele et al. (15)), so does not confirm an earlier meta-analysis by Singh et al., that had identified an increased cardiovascular risk with tiotropium Handihaler (16).
In conclusion, the present knowledge base of tiotropium Respimat Soft Mist Inhaler has been in the public domain since 2009 and is derived from rigorous scientific evaluation of all available patient level data. The Product Information of tiotropium Respimat as updated in 2010 constitutes an accurate reflection of benefit and risk. The recent meta-analyis by Singh et al. does not add new clinical information. The suggestion of confirmatory evidence and biological plausibility is not supported by the data. Potential differences in efficacy and safety between the two formulations of tiotropium are being explored in the large outcome study TIOSPIR.
Bernd Disse, MD PhD, VP Therapeutic Area Respiratory*, Norbert Metzdorf, PhD, Global Clinical Programme Leader Tiotropium*, Antonio Martin MD PhD, Medical Affairs Product Lead, SPIRIVA, Allergy and Respiratory Medical Affairs#, Franklin Cerasoli PhD, Global Therapeutic Area Lead, Cardiovascular and Metabolic Disease Medical Affairs# Inge Leimer, PhD, Statistician*, *Boehringer Ingelheim Pharma GmbH&Co KG, Ingelheim, Germany, # Pfizer Inc, New York, NY, USA
(1) Singh S, Loke YK, Enright PL, Furberg CD. Mortality associated with tiotropium mist inhaler in patients with chronic obstructive pulmonary disease: systematic review and meta-analysis of randomised controlled trials. Br Med J 2011; 342: d3215
(2) Division of Pulmonary-Allergy Drugs Advisory Committee and Office of Surveillance and Epidemiology, US Food and Drug Administration.FDA briefing document. 2009. www.fda.gov/downloads/AdvisoryCommittees/CommitteesMeetingMaterials/Drug... -AllergyDrugsAdvisoryCommittee/UCM190463.pdf
(3) Boehringer Ingelheim, Briefing Document: Tiotropium (SPIRIVA): Pulmonary Allergy Drug Advisory Meeting - November 2009 www.fda.gov/downloads/AdvisoryCommittees/CommitteesMeetingMaterials/Drug... -AllergyDrugsAdvisoryCommittee/UCM190466.pdf
(4) Boehringer Ingelheim International. Tiotropium (Spiriva) Respimat: evaluation of fatal events--February 2010. 2010. http:// trials.boehringer-ingelheim.com/res/trial/data/pdf/Pooled_Analysis_U10- 3255-01.pdf
(5) Summary of Product Characteristics (SPC): Spiriva Respimat 2.5 micrograms solution for inhalation (United Kingdom): www.medicines.org.uk/EMC/medicine/20547/PIL/
(6) Bateman ED, Tashkin D, Siafakas N, Dahl R, Towse L, Massey D, Pavia D, Zhong NS. A one-year trial of tiotropium Respimat plus usual therapy in COPD patients. Respir Med 2010;104: 1460-1472
(7) Voshaar T, Lapidus R, Maleki-Yazdi R, Timmer W, Rubin E, Lowe L, Bateman E. A randomized study of tiotropium Respimat Soft Mist inhaler vs. ipratropium pMDI in COPD. Respir Med 2008;102(1):32-41
(8) Bateman E, Singh D, Smith D, Disse B, Towse L, Massey D, Blatchford J, Pavia D, Hodder R. Efficacy and safety of tiotropium Respimat SMI in COPD in two 1-year randomized studies. Int J Chronic Obstruct Pulm Dis 2010;5:197-208
(9) ClincalTrials.gov: An Efficacy and Safety Study to Compare Three Doses of BEA 2180 BR to Tiotropium and Placebo in the Respimat Inhaler www.clinicaltrials.gov/ct2/show/NCT00528996
(10) Higgins, JPT, Green S (eds.) and Sally Green Cochrane Handbook for Systematic Reviews of Interventions, Version 5.1.0 www.cochrane- handbook.org/
(11) Disse B, Speck GA, Rominger KL, Witek TJ, Hammer R. Tiotropium (Spiriva): Mechanistical considerations and clinical profile in obstructive lung disease. Life Sci 1999; 64(6/7): 457-464
(12) Ichinose M, Fujimoto T, Fukuchi Y. Tiotropium 5 ?g via Respimat and 18 ?g via HandiHaler; efficacy and safety in Japanese COPD patients. Respir Med 2010;104(2):228-236
(13) Kesten S, Celli B, Decramer M, Leimer I, Tashkin D. Tiotropium HandiHaler in the treatment of COPD: a safety review. Int J Chronic Obstruct Pulm Dis 2009;4:397-409
(14) Summary Minutes of the Pulmonary-Allergy Drugs Advisory Committee (PADAC), November 19, 2009: www.fda.gov/downloads/AdvisoryCommittees/CommitteesMeetingMaterials/Drug... -AllergyDrugsAdvisoryCommittee/UCM196995.pdf
(15) Michele TM, Pinheiro S, Iyasu S. The safety of tiotropium - the FDA's conclusions. N Engl J Med 2010;363(12):1097-1099
(16) Singh S, Loke YK, Furberg CD. Inhaled anticholinergics and risk of major adverse cardiovascular events in patients with chronic obstructive pulmonary disease: a systematic review and metaanalysis. JAMA 2008;300:1439-50. (Erratum in JAMA 2009;301:1227-30.)
Competing interests: Bernd Disse, Norbert Metzdorf and Inge Leimer are full-time employees of B.I., Antonio Martin and Frank Cerasoli are full time employees of Pfizer
Boehringer Ingelheim Pharma GmbH&Co KG, Germany* and Pfizer Inc, New York, NY, USA#
Click to like: