Real World Evidence: From Efficacy to Effectiveness Professor Kamlesh Khunti University of Leicester, UK Leicester Diabetes Centre at University Hospitals of Leicester NHS Trust, 2015. Not to be reproduced in whole or in part without the permission of the copyright owner.
Leicester
Leicester Diabetes Centre
Overview What is RWD Comparison of RWD vs RCT Examples of RW studies Strengths of RWE RCT, randomized controlled trial; RWD, real-world data
Levels of evidence for therapy 1A 1B 2A 2B Systematic review (with homogeneity) of RCTs Individual RCT with narrow confidence interval Systematic review (with homogeneity) of cohort studies Individual cohort study 3 Individual case control study Present this in a pyramid diagram 4 Case series (and poor-quality cohort and case control studies) 5 Expert opinion without explicit critical appraisal RCT, randomized controlled trial
Real-world data Data that are collected outside the controlled constraints of conventional randomised clinical trials to evaluate what is happening in normal clinical practice 1 Ever-increasing role in decisions that affect patients access to therapies 2 Randomised controlled trials Real-world data Can it work? 3 Does it work? 3 1. ABPI. At: www.abpi.org.uk/our-work/library/industry/documents/vision-for-real-world-data.pdf. Accessed August 2013; 2. Peperell et al. Value Health 2012;15(7): Abstract PRM6. 3. Luce et al. Milbank Q 2010:88(2):256 276
RCT data vs. Real World data RCTs Real World Studies Type of Trial Experimental / interventional Observational / non-interventional Primary focus Efficacy, safety, quality and costeffectiveness Effectiveness, safety, quality, cost-effectiveness, natural history, compliance and adherence, service models, patient preference Patient population Narrow, restricted and motivated Diverse, large and unrestricted Monitoring Intense (ICH-GCP compliant) Not required (?) Comparators Gold standard / placebo None / standard clinical practice / multiple iterations Outcomes Clear sequence Wide range Data collection Standardised, controlled Routine, recruitment bias (?) confounders Validity Internal External Randomisation & Yes No Blinding Follow-up Short (?) Long Cost $$$$ $
Diverse and complex patients are present in clinical practice
Multimorbidity: Comorbidity of top 10 common conditions Guthrie B et al. BMJ 2012;345:e6341
Sources of real-world data Patient registry 1,2 Observational Specified outcomes Defined population Practical clinical trial 1,3 Prospective, randomised Large, diverse population Long follow-up Randomised controlled trial supplement 1 Additional data collection Resource utilisation Patient-reported outcome Real-world Real-world world data data data Electronic Health Record 1 Disease-specific symptoms/treatments Patient-level outcomes Health survey 1 Health status Healthcare utilisation Treatment patterns Claims database 1 Administrative data Diagnoses Procedures Costs 1. ISPOR Using Real World Data Task Force. At: www.ispor.org/workpaper/rwd_tf/rwtfdraftreport.pdf. Accessed August 2013; 2. Gliklich & Dreyer (eds). Registries for evaluating patient outcomes: a user s guide. 2 nd ed. Rockville, MD: AHRQ, 2010; 3. Tunis et al. JAMA 2003;290:1624 1632.
Digitized Medical Information
Achievements of A1c in Europe: GUIDANCE Study GUIDANCE study 7,597 patients with type 2 diabetes Gap exists between checking HbA1c and achieving target HbA1c < 7.0% Percent, % 100 90 80 70 60 50 40 30 20 10 0 Belgium France Germany Ireland Italy Netherlands Sweden UK Total sample HbA1c checked Met HbA1c target GUIDANCE, Guideline Adherence to Enhance Care; HbA1c, glycated haemoglobin A1c. Stone MA, et al. Diabetes Care. 2013;36:2628-38.
136 Reports, 19 different treatments NEJM 2000; 342:1878-1886 99 Reports, 5 clinical topics NEJM 2000; 342:1887-1892 results of well-designed observational studies do not systematically overestimate the magnitude of the effects of treatment as compared with those in randomized, controlled trials on the same topic
Adverse effects f interventions: Real world vs RCT 19 studies, 58 meta-analysis Conclusions: no difference on average in the risk estimate of adverse effects of an intervention derived from meta-analyses of RCTs and metaanalyses of observational studies it may be preferable for systematic reviewers to evaluate a broad range of studies Golder S et al. PLOS Medicine 2011:8:e1001026
Real-world safety data RCTs may miss rare or slow-developing adverse events owing to limited patient numbers and duration Long-term surveillance in large numbers of patients, including groups unrepresented in RCTs is necessary to detect rare events Many types of real-world study provide safety data, including registries, claims databases and long-term observational studies RCT, randomised controlled trial 1. ISPOR Using Real World Data Task Force. At: www.ispor.org/workpaper/rwd_tf/rwtfdraftreport.pdf. Accessed September 2013
Metformin use in people with heart failure Uncertainty on best management approaches to HF management HF patients excluded from trials Metformin Contraindicated in HF due to potential risk of lactic acidosis Based on observational studies, FDA asked for warning to be removed in 2007 (Inzucchi et al Diabetes Care. 2007;30:e129) Systematic reviews of observational studies: Pooled adjusted risk ratios for metformin compared with other treatments for all-cause mortality Eurich DT et al. Circ Heart Fail. 2013;6:395-402
Influence of Severe Hypoglycemia on Events in ADVANCE Number (%) of patients with event Events Sv. Hypo: Yes (n=231) Sv. Hypo: No (n=10909) Hazard ratio (95% CI) Major macrovascular events Adjusted model Major microvascular events Adjusted model All-cause deaths Adjusted model CVD deaths Adjusted model Non-CVD deaths Adjusted model 33 (15.9%) 1114 (10.2%) 24 (11.5%) 1107 (10.1%) 45 (19.5%) 986 (9.0%) 22 (9.5%) 520 (4.8%) 23 (10.0%) 466 (4.3%) 3.53 (2.41 5.17) 2.19 (1.40 3.45) 3.27 (2.29 4.65) 3.79 (2.36 6.08) 2.80 (1.64 4.79) 0.1 1.0 10.0 Hazard ratio Zoungas S et al. N Eng J Med. 2010
Incidence of CVD and mortality in patients experiencing hypoglycaemia 100 T1DM 100 T2DM Incidence rate (per 1,000 person-years) 90 80 70 60 50 40 30 20 10 0 CV events All-cause mortality Data are unadjusted incidence rates CV, cardiovascular; CVD, cardiovascular disease Incidence rate (per 1,000 person-years) 90 80 70 60 50 40 30 20 10 0 CVD history No CVD history CV events All-cause mortality Khunti K, Davies M et al. Diabetes Care 2015;38:316 22
Time from hypoglycaemia to first CV event or death T1D T2D Patients with hypoglycaemia and 1 CV event (n) Time from first hypoglycaemic episode to first CV event (years) Patients with hypoglycaemia and death (n) Time from first hypoglycaemic episode to death (years) 38 97 1.5 (0.5; 3.5) 1.5 (0.5; 3.0) 169 493 1.1 (0.3; 2.3) 0.8 (0.3; 2.3) Data are median (interquartile range) from CPRD plus HES data. CPRD: Clinical Practice Research Datalink; HES: Hospital Episode Statistics. Khunti K, Davies M et al. Diabetes Care. 2015;38:316-322.
Prospective vs Retrospective data
HAT study design Baseline (R-Day) Retrospective Prospective Inclusion criteria T1DM and T2DM Treated with insulin for more than 12 months Adult > 18 years Informed consent to participate Part 1 6 months - SH 4 weeks - NSH Part 2 4 weeks SH & NSH Exclusion criteria Non-ambulatory patients Illiterate patients Patients unable to complete written survey One day retrospective crosssectional evaluation Day 1 Return SAQ Part 1 4 weeks prospective observational evaluation Day 28 Post SAQ Part 2 Patient Diary SH, severe hypoglycaemia; NSH, non-severe hypoglycaemia; SAQ, self-assessment questionnaire
HAT: retrospective and prospective hypoglycaemia rates in T2D T2D (n=19,563) Retrospective Prospective IRR 1.20 Hypoglycaemia prevalence, % of patients Retrospective Prospective Any (4 wks) 51% 47% IRR 0.69 Nocturnal (4 wks) 22% 16% Severe (6 mths/4 wks) 16% 9% Any and nocturnal based on 4 week period. Retrospective data based on 6 month period; Prospective data based on 4 week period; IRR, incidence rate ratio; IRR calculated from completers analysis set, incidence and prevalence calculated from full analysis set. Khunti et al. Diab Obes Met 2016
ABCD nationwide liraglutide and exenatide audits: UK Two audits of liraglutide and exenatide real clinical use in the UK HbA 1c Hypoglycaemia Exenatide audit 1 Launched December 2008 Prospective database Routine measurements (non-interventional) The audit collected anonymised data on 6717 patients (2007 to 2009) submitted by 315 contributors from 126 centres Side effects Insulin use Weight Combinations Liraglutide audit 2 Launched autumn 2009 Prospective database Routine measurements (non-interventional) March 2013 status: 5948 patients from 89 centres Professional drivers CV events Lipids/ALT Responders ABCD, Association of British Clinical Diabetologists; ALT, alanine aminotransferase; CV, cardiovascular; HbA 1c, glycosylated haemoglobin 1. https://www.diabetologists.org.uk/abcd_nationwide_exenatide_audit.htm; 2. http://www.diabetologists-abcd.org.uk/glp1_audits/liraglutide_audit.htm;
Real-world effects of liraglutide on HbA 1c and body weight Baseline BMI (kg/m 2 ): 25 29.9 30 34.9 35 39.9 40 44.9 45 49.9 50 54.9 Mean change from baseline HbA 1c (%) Body weight (kg) LEAD programme excluded patients with BMI >40 kg/m 2 BMI, body mass index Ryder et al. Diabetologia 2012;55(Suppl 1):S330. Abstract 801
Effect of background diabetes therapy on HbA 1c reduction 0.0 1 OAD 2 OADs 3 OADs Insulin ± OAD Adjusted mean change in HbA 1c (%) -0.5-1.0-1.5-2.0 N=119 N=209-1.7-1.7 N=67-1.9 N=243-0.9-2.5 ED: -1.0 (95% CI: -0.6, -1.5) p<0.001 ED: -0.8 (95% CI: -0.5, -1.1) p<0.001 ED: -0.8 (95% CI: -0.4, -1.1) p<0.001 Data on adjusted mean (SE) and ED were analysed by ANCOVA with baseline HbA 1c as a co-variate. ANCOVA, analysis of covariance; CI, confidence interval; ED, estimated differences; HbA 1c, glycosylated haemoglobin; OADs, oral antidiabetic drugs; SE, standard error. Thong et al. Diabetes 2012;61(Suppl 1):A266.
Effect of diabetes duration on HbA 1c reduction 0.0 0 5 years 6 10 years >10 years Adjusted mean change in HbA 1c (%) 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 N=181-1.7 N=195-1.4 N=210-1.2 ED: -0.2 (95% CI: -0.5, 0.1) p=0.140 ED: -0.5 (95% CI: -0.2, -0.8) p<0.001 Data on adjusted mean (SE) and ED were analysed by ANCOVA with baseline HbA 1c as a co-variate. ANCOVA, analysis of covariance; CI, confidence interval; ED, estimated differences; HbA 1c, glycosylated haemoglobin; SE, standard error. Thong et al. Diabetes 2012;61(Suppl 1):A266.
ADA/EASD: position statement for managing hyperglycaemia Healthy eating, weight control, increased physical activity Initial monotherapy Metformin Two-drug combinations SU TZD DPP-4i SGLT2i GLP-1 RA Insulin Three-drug combinations TZD DPP-4i SGLT2i GLP-1 RA Insulin SU DPP-4i SGLT2i GLP-1 RA Insulin SU TZD SGLT2i Insulin SU TZD DPP-4i Insulin SU TZD Insulin TZD DPP-4i SGLT2i GLP-1 RA More complex strategies Insulin (MDI) Escalate therapy at 3 months if target not achieved. DPP-4i, dipeptidyl peptidase-4 inhibitor; SGLT2i, sodium-glucose cotransporter 2 inhibitor; GLP-1 RA, glucagon-like peptide-1 receptor agonist; MDI, multiple daily injection; SU, sulfonylurea; TZD, thiazolidinedione. Inzucchi SE, et al. Diabetes Care. 2015;38:140-9.
Longitudinal HbA1c changes in patients on GLP1-RA -1.15% 66,583 patients Mean age 56 years 87% Obese added Insulin within 24 months no Insulin within 24 months Earlier intensification by 6 months is associated with 18% higher odds of lowering HbA1c below 7% at 24 months Montvida O et al. Diab Obes Met 2016 (online)
Longitudinal HbA1c changes switched to Insulin within 24 months added Insulin within 24 months no Insulin within 24 months Delays in recognising glycaemic failure (therapeutic inertia) Montvida O et al. Diab Obes Met 2016
Early Glycemic Control May Predict Persistence of Glycemic Control In the ACCORD trial, strict glycemic control (target HbA 1c <6.0%) was initially targeted Patients were then transitioned to a target HbA 1c of 7.0 7.9%, and followed for an additional 1.1 ± 0.2 years Patients with lower pre-transition HbA 1c were more likely to maintain an HbA 1c <6.5% over time RR (95% CI) of achieving a final HbA 1c of <6.5% 7 6 5 4 3 2 1 Increased likelihood (RR) of achieving a final HbA 1c <6.5% Decreased likelihood (RR) of achieving a final HbA 1c <6.5% Unadjusted Adjusted a 0 5.5 6 6.5 7 7.5 Pre-transition HbA 1c (%) a Adjusted for demographics, co-interventions, baseline pre-randomisation HbA 1c, BMI, and medications, and post-transition change in BMI and medications BMI, body mass index; CI, confidence interval; HbA 1c, glycated hemoglobin; RR, relative risk Punthakee Z, et al. Diabetologia 2014;57:2030 2037
Percentage of patients at HbA 1c target ( 7.0%) at 3 and 24 months post index 40,627 patients, 5 European countries and US Patients initiating Basal Insulin Mauricio D et al. Diab Obes Metabol 2017 (online)
Failure to achieve target at 3 months associated with increased odds of not achieving target at 24 months Mauricio D et al. Diab Obes Metabol 2017 (online)
Consequences of delayed intervention HbA1c, % 8.5 8.0 7.5 Patients with HbA1c 7% not receiving IT within 1 year Patients with HbA1c <7% who received IT before 1 year of diagnosis At 5.3 years, significantly increased risk of: MI 67% (CI 39 101%) Stroke 51% (CI 25 83%) HF 64% (CI 40 91%) Composite CVE 62% (CI 46 80%) 7.0 6.5 Bad glycaemic legacy Drive risk for complications 6 12 48 54 60 Months CVE, cardiovascular endpoint; HF, heart failure; IT, treatment intensification; MI, myocardial infarction Paul S, et al. Cardiovasc Diabetol 2015;14:100 doi:10.1186/s12933-015-0260-x
Summary and conclusions Well-designed real-world studies complement RCTs Need to be aware of types of real world data Choose appropriate data and methodology to answer the question Need to be aware of strengths and limitations
Thank you www.leicesterdiabetescentre.org.uk www.facebook.com/leicesterdiabetescentre @kamleshkhunti @LDC_Tweets