Exhibit 99.4

Poster No.1248

Population pharmacokinetics and pharmacodynamics of GSK961081 (MABA) in patients with moderate to severe COPD

Claire Ambery(1), Pascal Wielders(2), Andrea Ludwig-Sengpiel(3), Robert Chan(1), John Riley(1)

(1)GlaxoSmithKline, Stockley Park West, Uxbridge, UK; (2)Dept. Pulmonary Diseases, Catharina Hospital, Eindhoven, Netherlands; (3)KLB Healthresearch, Luebeck, Germany

INTRODUCTION

·      GSK961081 is a potent bi-functional molecule that has demonstrated both anti-muscarinic receptor activity (MA) and beta2-adrenergic agonist activity (BA) in pre-clinical and clinical investigation [1,2,3].

·      In COPD trials, GSK961081 has shown clinically meaningful bronchodilation with rapid onset of action with a good safety and tolerability profile [4,5].

OBJECTIVES

·      To characterise the population pharmacokinetics (PK) and pharmacodynamics (PD) of GSK961081 in moderate to severe COPD subjects.

METHODS

·      Data were obtained from a 4-week, multicentre, randomised, double-blind, double dummy, placebo and salmeterol controlled parallel group study [5].

·      Once (OD) and twice (BD) daily GSK961081 DISKUS™ dosing regimens were assessed;

·                  OD doses: 100, 400 & 800 mcg

·                  BD doses: 100, 200 & 400 mcg

·      Trough FEV1 at day 29 was the primary endpoint.

·      Plasma GSK961081 concentrations were an additional endpoint.

·                  Blood samples were collected on Day 28: Pre AM dose (-1h to 0min) and post AM dose between 0-30min, 30min-2h, 2-6h and 6-11h. Pre PM dose (-1h to 0min) and post PM dose between 0-30min, 30min-2h, 2-6h and 6-11h. PK subset subjects only.

·                  PK bioanalysis was performed using a validated high performance liquid chromatography/mass spectrometry (HPLC-MS/MS) method with a lower limit of quantification (LOQ) of 25 pg/mL.

FIGURE 1. PK and PD Models

DISKUS™ is a trade mark of the GlaxoSmithKline group of companies

FIGURE 2. Final PK Model

PK and PD Analyses

·      The software NONMEM 7 (ICON Development Solutions) was used.

·      M3 methodology was used to handle non-quantifiable (NQ) PK data [6,7].

·      Models were prioritised using objective function values, plausibility of parameters and graphical checking.

·      Stepwise covariate model building was used (PD model only).

·      A 2-compartment PK model with first-order absorption was selected and fitted to the PK data (Figure 1A).

·      To characterise the dose-response curve an Emax model was selected and fitted to the PD data (Figure 1B).

TABLE 1. PK Model Parameters

CI: Confidence interval; %RSE: Relative standard error.

RESULTS

PK

·      Greater than 50% of PK data per treatment group was NQ; except for at 800mcg QD (27% NQ) and 400mcg BD (30% NQ);

·      Therefore the PK model was selected based on these two treatment groups only (N=47).

·      The PK model well characterised the observed 800mcg OD and 400mcg BD data (Figure 2(C) and (F)).

·      PK model parameters were estimated with good precision (%RSE<20%) (Table 1).

·      Covariates were not included in the PK model.

·      Evaluation of the PK model using the data not used in model building (N=139) showed a good fit (Figure 2(A),(B),(D) and (E)).

FIGURE 3. Final PD Model (Trough FEV1, day 29)

PD

·      The PD model gave a good fit to the observed data (Figure 3(A)) (N=347).

·      PD model parameters were estimated with precision%RSE<50% (Table 2).

·      Dosing regimen, age, weight, sex, height and inhaled corticosteroid use were not identified as statistically significant covariate effects.

·      Baseline FEV1 was identified as a statistically significant covariate effect on E0.

·      The PD model concurred well with the primary endpoint analysis (Figure 3(B) and [5]) based on post-hoc derived placebo corrected change from baseline.

TABLE 2. PD Model Parameters

CI: Confidence interval; %RSE: Relative standard error.

CONCLUSIONS

·      The PK model described will be used as a tool for guiding GSK961081 clinical development.

·    The model performed well and in addition was able to account for Non Quantifiable data.

·    Covariate inclusion in the model will be re-visited once additional PK data is available.

·      The PD model described will be used as a tool for guiding dose selection for GSK961081 in Phase III trials.

·    There was no influence of dosing regimen on the PD model; indicating there was no apparent difference between OD and BD dosing for day 29 trough FEV1 (primary end point).

REFERENCES

(1) Aiyer J, et al. Am J Resp Crit Care Med 2009;179:A4552.

(2) Pulido-Rios MT, et al. Am J Resp Crit Care Med 2009;Volume:179:A6195.

(3) Norris V, et al. Pulm Pharm Ther 2013;(in press http://dx.doi.org/10.1016/j.pupt.2013.03.009).

(4) Bateman ED, et al. Pulm Pharm Ther 2013:(in press http://dx.doi.org/10.1016/j.pupt.2013.03.015).

(5) Wielder PLML, et al. ERJ 2013;(in press doi: 10.1183/09031936.00165712).

(6) Ahn EA, et al. JPKPD 2008;35:401-421.

(7) Ribbing J, et al. JPKPD 2004; 31(2):109-134.

ACKNOWLEDGMENTS

·               The presenting author, Claire Ambery, declares the following real or perceived conflicts of interest during the last 3 years in relation to this presentation: Employed by, and share holder in, GlaxoSmithKline.

·               This study was registered on the Clinical Trials Register NCT01319019, used the study code MAB115032, and was funded by GlaxoSmithKline.

Presented at the European Respiratory Congress, Barcelona, Spain, 08 September 2013