Studying the Outcomes in Patients with Tricuspid Regurgitation Treated with Valve Repair or Valve Replacement: Interpreting the Survival Pattern on The Long Term by Application of Artificial Intelligence Methods

TL;DR

This study uses artificial intelligence to analyze long-term survival in patients with tricuspid regurgitation treated with valve repair or replacement.

Contribution

The study applies the IPDfromKM method to compare valve repair and replacement outcomes across six trials.

Findings

Valve repair showed significantly better survival than valve replacement (HR 0.6098).

Heterogeneity was significant in valve replacement groups but not in repair groups.

NYHA class III or IV was a negative prognostic factor for valve replacement patients.

Abstract

The reconstruction of individual patient data from published Kaplan-Meier survival curves is a new technique (often denoted as the IPDfromKM
method) for studying efficacy in cases where multiple trials are available, and
the endpoint is long-term mortality. In patients with tricuspid regurgitation,
both valve repair and valve replacement have been proposed to improve prognosis;
6 controlled clinical trials (CTs) have been conducted to compare the two
therapeutic options mentioned above. The objective of our analysis was to study
these six trials through the application of the IPDfromKM method.

In the present report, we applied the IPDfromKM method to carry
out a pooled analysis of these 6 CTs to investigate the effectiveness of valve
repair vs valve replacement and to assess the between-study heterogeneity from
this clinical material. After reconstructing individual patient data from these 6
trials, patients treated with valve repair were pooled together and their
Kaplan-Meier curve was generated. Likewise, patients treated with valve
replacement were pooled together and their Kaplan-Meier curve was generated.
Finally, these two curves were compared by standard survival statistics. The
hazard ratio (HR) was determined; death from any cause was the endpoint.

These 6 CTs included a total of 552 patients; in each of these
CTs, the patient group treated with valve repair was compared with another group
treated with valve replacement. Our statistical results showed a significantly
better survival for valve repair compared with valve replacement (HR, 0.6098; 95% confidence intervals (CI), 0.445 to 0.835; p = 0.002). Heterogeneity was
found to be significant in the 6 patient arms undergoing replacement, but not in
those undergoing valve repair. In valve replacement, the classification of
patients in class III or IV of New York Heart Association (NYHA) was the main
negative prognostic factor.

Our analysis confirmed the
methodological advantages of the IPDfromKM method in the indirect comparative
analysis of multiple trials. These advantages include appropriate analysis of
censored patients, original assessment of heterogeneity, and graphical
presentation of the results, wherein individual patients retain an important
role.

Figures