TREATING CORONARY HEART DISEASE IN DIABETIC PATIENTS:A SYSTEMATIC REVIEW OF SYSTEMATIC REVIEWS

Claudia Nisa PhD1, Luis Filipe Azevedo MD PhD1,2, Frederic Resnic MD PhD,3,4, Mariana F LoboMSc1, Alberto Freitas PhD1,4, Vanessa Azzone PhD5, Leonor Bacelar Nicolau MSc6, ArmandoTeixeira‐Pinto PhD7, Sharon‐Lise Normand PhD5,8, Altamiro Costa‐Pereira MD PhD1,4

1Center for Health Technology and Services Research, Faculty of Medicine, University of Porto, Portugal; 2Department of Health Information and DecisionSciences, Faculty of Medicine, University of Porto, Portugal ; 3Tufts University School of Medicine, Boston, Massachusetts, United States; 4Department ofCardiovascular Medicine and Comparative Research Institute, Lahey Hospital and Medical Center, Burlington, Massachusetts, United States; 5Department ofHealth Care Policy, Harvard Medical School, Boston, Massachusetts, United States; 6Institute of Preventive Medicine and Public Health and ISAMB – Instituteof Environmental Health, Faculty of Medicine, University of Lisbon; 8Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston,Massachusetts, United States; 7School of Public Health and Faculty of Medicine, the University of Sidney, Sidney, Australia

MOTIVATION

INITIAL OBJECTIVEReview of Systematic Reviews comparing revascularization strategies

REVIEW Included Studies Follow‐up

Mortality MI Repeat revascularization StrokePotential sources of bias

RE (95%CI) Favors RE (95%CI) Favors RE (95%CI) Favors RE (95%CI) Favors

Tu et a 2014

FREEDOM, CARDia, VA CARDS, ARTS I & ARTS II, ERACI II & 

III, SYNTAX, PRECOMBAT

Min 30 days max 5 years

OR 1.63 (1.31‐2.02) CABG OR 1.28 (0.68‐2.40) n.s. OR 2.53 

(1.75‐3.64) CABG NR 

ARTS II and ERACI III not RCTs Mixing pre‐specified with post‐hoc diabetic subgroup analysis

Fanari et al 2014

SYNTAX, FREEDOM, CARDia

Min 1y max 5 y

1y OR 0.97 (0.68‐1.38); 5y OR 1.36 

(1.11‐1.66)

1y n.s.; 5y CABG

1y OR 1.27 (0.75‐2.15); 5y OR 2.01 

(1.54‐2.62)

1y n.s.; 5y CABG NR 

1y OR 0.40 (0.19‐0.81); 5y OR 0.59 (0.39‐

0.89)

DES

Incomplete search; Mixing pre‐specified with 

post‐hoc diabetic subgroup analysis

Wu et al 2015

ARTS II, ASAN‐MV, Briguori et al, Dohi et al, CARDia,  Lee et al, Dominguez‐

Franco et al, FREEDOM,  Luo et al, MAIN‐COMPARE, Ohno et al , Qiao et al, SYNTAX, VA 

CARDS, Yamagata et al, Zhao et al

Min 1y max 5 y

30days OR 2.03 (1.11‐3.73); 1y OR 0.97 (0.70‐1.33); max HR 0.74 (0.59‐0.92)

30days CABG; 1y and 5y n.s.

30days OR 1.41 (0.77‐2.57); 1y OR 0.72 (0.43‐1.20); max Or 0.56 (0.43‐

0.74)

30 days and 1y n.s.; 5y CABG

30days OR 0.76 (0.19‐3.02); 1y OR 0.28 (0.19‐0.42); max OR 0.36 

(0.23‐0.55)

30 days n.s.; 1y and 5y CABG

30days OR 6.02 (2.43‐14.87); 1y OR 2.75 (1.48‐5.10); max OR 2.02 (1.33‐3.06)

DES

Estimates using data from RCTs and non‐RCTs; 

Mixing pre‐specified with post‐hoc diabetic subgroup analysis

REVISED OBJECTIVEMeta‐Analysis of primary trials

To reassess the clinical evidence based on primary RCTs comparing CABG with PCI controlling for:

Follow‐up differences

Pre‐defined versus post‐hoc diabetics comparison

Primary endpoint: all‐cause mortality;

Secondary endpoints: myocardial infarction, repeatedrevascularization, and stroke;

No specific follow‐up period defined to evaluate the rangeof follow‐up endpoints available

RESULTS

13 RCTs included (4372 patients)

Only 6 RCTs with pre‐specified subgroupanalysis (46%) for diabetics

Trials with post‐hoc analyses reporting mostlymortality

Substantial difference between trials with ex‐ante versus ex‐post analyses in thedistribution of risk factors namely proportionof patients with multi‐vessel disease andinsulin dependence

Up to 1 year 2 to 3 years 4 to 5 years 6 to 10 years Longest follow‐up

Results overall favoring CABG at 5 years only;

Identified only in trials with a pre‐specifiedsubgroup comparison;

Meta‐regression model with dummy for pre‐specification (0=post‐hoc; 1=pre‐specified) notsignificant but funnel plot suggests that largerstudies with lower variance favor CABG.

RESULTS

DISCUSSIONThe large amount of existing reviews examining the effectiveness of CABG versusPCI did not return more consensus in results;

Many potential sources of bias identified, in particular combining very distinctfollow‐up duration and data from ex‐ante versus ex‐post subgroup analysis;

In particular for all‐cause mortality, there are conflicting results depending onwhether the diabetic subgroup analysis was pre‐specified; CABG best option inthe long‐term only in trials pre‐specifying diabetes;

Future Research should address:The implications of this result to other high‐risk patients and risk factors (e.g., kidney disease)If this difference is identified in trials comparing stents versus stentsPossible consequences for the estimation of relative cost‐effectiveness between revascularization strategies