Internal Medicine Portal: Archived - Journal Club Blog

Journal Club: PIONEER-HF

Angiotensin–Neprilysin Inhibition in Acute Decompensated Heart Failure

Known as the "PIONEER-HF" Trial, discussed August 9, 2019

Related trials: PARADIGM-HF, CONSENSUS, SOLVD, CHARM-Alternative

Key clinical take-aways from this article: 

• We need to know it: In patients hospitalized with acute decompensated heart failure with reduced ejection fraction, initiation of sacubitril-valsartan therapy reduced NT-proBNP levels when compared to enalapril
  • Sacubitril-valsartan therapy decreased NT-proBNP concentration compared to enalapril therapy at 4 and 8 weeks, without significantly different rates of medication related adverse effects.
• Basically compared an ARNI vs an ACE
• Small sample size compared to previous study (PARADIGM-HF); only had 880 patients compared to ~8000 in multiple sites in the US. It also had a different endpoint of comparison, looking at a biomarker
  • Journal Club would have liked to see an increased number of patients and a longer follow up than 8 weeks for more informed decision making
• Also concerns about population studied
  • Population ONLY looked at blacks and whites (no other races)
  • Majority males (25.7% / 30.2% female in ARNI vs ACE) - could be argued that this is a population similar to those that we see at our hospital
  • NYHA Class IV made up a small population of the study (the ICU patients we'd like to look at)
  • Serum Creatinine grouping didn't seem to include the ESRD patients, and we have a number of patients with CKD and heart failure. Are these results valid for those with CKD?
• Funding bias - Funded by Novartis, the makers of this drug (drug costs ~$10/pill, roughly $300/month/patient) and found very much in favor of it
• Post-Hoc Analysis is a little sketchy for endpoints
• Take-aways - it may actually be reducing rehospitalization, but it could also be increasing symptomatic hypotension; we can't conclude it does no harm, and we need a longer study with more endpoints. 

Learning points from article: Post-Hoc Analysis

• The group performed 6 exploratory subgroup analyses (presumably post-hoc/hypothesis-generating) analyses of certain HF-relevant events, including rehospitalization  [see Table 2 in the article]
  • Post-hoc probabilities:
    • Post hoc analysis of the width and magnitude of the 95% confidence interval (95% CI) may be a more appropriate method of determining statistical power.
      • Note that table 2 does not include p values - it has confidence intervals (which need to be adjusted)
      • The Bonferroni correction can be used to adjust confidence intervals. If you establishes m confidence intervals, and wish to have an overall confidence level of  1-α, each individual confidence interval can be adjusted to the level of 1 - (α/m) 
        • Probabilities in post-hoc tests are cumulative e.g., three comparisons at 0.05 level produce a cumulative probability of type I error of 0.15. Controls for Type 1 error across multiple tests. 
        • The probability of each test must equal α / the number of comparisons to preserve the overall significance level (Bonferroni correction) <-- regular p values
          • Example:  If there are 3 groups being compared at the 0.05 level (for a total of 3 comparisons), each must be significant at p = 0.017.  
        • Therefore, for our article, the confidence intervals should be adjusted - we have 6 post-hoc subgroup analyses at an α of 0.05. 1-(0.05/6) = 0.992 CI
          • Okay, so we now have a wide confidence interval. Normally, we're interested in the precision of the point estimate and want a small confidence interval; but we're also worried about the truth/reproducability. As this confidence interval gets wider, we're 99% confident that the true value lies within this range, so it's okay to have a wider range
          • We do have to reconsider the numbers on Table 2 - they'll be a wider stretch than what's stated (so some of the numbers may in fact now cross 1 when they didn't before). There's calculators for this.
        • Authors noted at the bottom of table 2 that this is a limitation and "inferences drawn from these intervals may not be reproducible."

Citation information:

Velazquez EJ, et al. "Angiotensin-Neprilysin Inhibition in Acute Decompensated Heart Failure". The New England Journal of Medicine. 2018. ePub 2018-11-11:1-10.

https://www.ncbi.nlm.nih.gov/pubmed/30415601

Further Reading:

A general introduction to adjustment for multiple comparisons.