Author + information
- Christopher M. O’Connor, Editor-in-Chief, JACC: Heart Failure∗ ()
- ↵∗Address for correspondence:
Dr. Christopher M. O’Connor, Editor-in-Chief, JACC: Heart Failure, American College of Cardiology, Heart House, 2400 N Street NW, Washington, DC 20037.
Over the 30 years that I have spent participating in heart failure (HF) clinical trials, it is amazing how many trials have finished very close to the pre-specified nominal p value of 0.05.
The CHARM (Candesartan in Heart failure Assessment of Reduction in Mortality and morbidity)-Preserved trial evaluated the effects of the long-acting angiotensin receptor blocker candesartan compared with placebo in patients with symptomatic HF and ejection fraction (EF) ≥40% with a primary outcome of cardiovascular death or hospitalization for HF. The intervention demonstrated a covariate adjusted benefit with a hazard ratio of 0.86, confidence interval [CI] of 0.74 to 1.00; p = 0.051 (1). Given this finding, candesartan did not meet its primary endpoint and was not integrated into the guidelines for potential benefit in patients with HF with preserved EF.
The DOSE (Diuretic Strategies in Patients with Acute Decompensated HF) trial compared a bolus versus continuous infusion of furosemide, and compared high-dose (2.5 × the patient’s home oral dose) versus standard-dose furosemide in a 2 × 2 factorial design (2). There was a trend toward greater improvement in the primary endpoint of patient global assessment of symptoms in the high-dose group, with p = 0.06. Most of the secondary endpoints conferred benefit in the high-dose strategy. Nonetheless, this finding did not lead to guideline endorsement of high-dose diuretic therapy in acute decompensated HF.
Recently, we learned the results from PARAGON-HF (Efficacy and Safety of LCZ696 Compared to Valsartan, on Morbidity and Mortality in Heart Failure Patients With Preserved Ejection Fraction) (3,4), a trial of combination neprilysin inhibition/angiotensin receptor antagonist in HF with preserved EF. A total of 4,822 patients with New York Heart Association functional class II to IV, EF ≥45% were randomized to sacubitril-valsartan versus valsartan. The primary endpoint was total HF hospitalizations or cardiovascular death. The rate ratio for the primary endpoint was 0.87 (95% confidence interval: 0.75 to 1.01; p = 0.06). During the presentation of the late-breaking results at the European Society of Cardiology meetings, there was a big gasp in the audience and signs of disappointment when the primary endpoint was shown. In contrast, the DAPA-HF (Study to Evaluate the Effect of Dapagliflozin on the Incidence of Worsening Heart Failure or Cardiovascular Death in Patients With Chronic Heart Failure) results were presented a few minutes later, showing that in 4,744 patients, dapagliflozin reduced the primary composite endpoint of worsening HF or cardiovascular death by 26% with a hazard ratio of 0.74 (95% confidence interval: 0.65 to 0.85; p < 0.001).
What if we changed our societal view of what statistical significance means? Many clinicians in the community believe that the p value provides a simple, reliable, objective triage tool for separating truth from the unimportant. P values are used to assess how unexpected the observed effect size is if the true state was no effect and no difference, that is, the null hypothesis. More importantly, what we want to know as clinicians and the community is the effect size or magnitude, and how precise is the effect size (7). In randomized controlled trials, should we focus on the odds ratios, hazard ratios, rate ratios, and the confidence intervals, and use clinical judgment over p values? What would the world look like today with the previous 3 examples if we re-examined the way we interpret the results in a pre-specified fashion? Our patients deserve another look of how we interpret clinical research.
- 2019 American College of Cardiology Foundation
- Solomon S.D.,
- Zile M.,
- Pieske B.,
- et al.
- O’Connor C.M.,
- deFilippi C.
- Busko M.
- Stiles S.
- Mark D.B.,
- Lee K.L.,
- Harrell F.E.