Article Text


85 Prediction of response to biventricular pacing from dyssynchrony indices: the absolute limit on predictability, and its clinical implications
  1. S S Nijjer1,
  2. P Pabari2,
  3. B Stegemann3,
  4. V Palmieri4,
  5. N Freemantle5,
  6. A Hughes2,
  7. D P Francis2
  1. 1Imperial College Healthcare NHS Trust, London, UK
  2. 2Imperial College London, London, UK
  3. 3Medtronic Bakken Research Center, Maastricht, The Netherlands
  4. 4Ospedale dei Pellegrini, Naples, Italy
  5. 5University of Birmingham, Birmingham, UK


Background It may be incorrect to believe that, with a good echocardiographic marker of mechanical dyssynchrony, response to biventricular pacing (BVP) should be predictable with a high r2 value. Variability between repeat echocardiographic measurements, and between successive dyssynchrony measurements, may reduce r2. Both will mandatorily limit the achievable r2; we determine this “contraction factor”.

Method and Results We compared correlation coefficients of dyssynchrony indices with response markers, in externally monitored randomised controlled trials (EMRCTs) and highly skilled single centre studies (HSSCSs). ΔLVEF in CRT recipients comprises true CRT effect plus unpredictable spontaneous variability present in control patients (Abstract 85 figure 1, upper panel). The resultant depression in r2 is calculated. HSSCSs overstate r2 between dyssynchrony and remodelling response in contrast to EMRCTs (p<0.0000000001), whether response is LVEF (0.40 vs 0.01), ESV (0.26 vs 0.01); EDV (0.53 v 0.01). An “averaged” reported r2 between differing dyssynchrony markers to commonly used echocardiographic response markers is shown in Abstract 85 figure 1, lower panel.

EMRCT data shows maximal r2 between dyssynchrony and ΔLVEF is 0.57 (ΔESV, 0.54; ΔEDV, 0.50). Dyssynchrony indices' own variability further contracts observable r2 values (by x0.68). The overall ceiling to r2 is between dyssynchrony and ΔLVEF is 0.39 (ΔESV, 0.37; ΔEDV, 0.34). All EMRCT r2 values obey these statistical limits; 29% of HSSCSs results do not.

Conclusions HSSCSs suggest dyssynchrony markers strongly predict response to BVP but EMRCTs cannot confirm this. Natural variability forces observed correlation coefficients between dyssynchrony and response to be low. EMRCTs, being less susceptible to publication bias, reflect this reliably. Frequent citation (without verification in independent cohorts) of the most exuberant values, from HSSCSs creates mathematically unviable, unrealistic, expectations. Simply searching for progressively more extreme correlations is therefore misguided. Rationally, we should concentrate on improving test-retest reproducibility of markers of dyssynchrony and of response.

  • Heart failure
  • statistics
  • cardiac resynchronisation therapy

Statistics from

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.