Skip to main content

Table 2 Hazard ratios, absolute differences for matched and unmatched dataset

From: Thymectomy for non-thymomatous myasthenia gravis: a propensity score matched study

 

HR for R/MM status

Absolute difference in R/MM at 5 years

Probability HR > 1

Standard Cox models

Unmatched data

1.5 (CI: 0.8, 2.8)

−3% (CI: −10, 4)

--

Matched data

1.9 (CI: 1.6, 2.3)*

15% (CI: 1, 29)

--

Bayesian models

Matched data, uninformative prior

2.2 (CrI: 0.9, 6.0)

9% (CrI: −1, 27)

96%

Matched data, skeptical prior

1.2 (CrI: 0.8, 1.6)

2% (CrI: −1, 9)

79%

Unmatched data, uninformative prior

1.9 (CrI: 1.0, 3.3)

20% (CrI: 2, 38)

98%

Unmatched data, skeptical prior

1.2 (CrI: 0.8, 1.7)

4% (CrI: −3, 12)

86%

  1. Values for the standard Cox models are expressed with a 95% confidence interval (CI).
  2. The Bayesian models used 10000 iterations. The values presented are the median and 95% credible intervals (CrI).
  3. HR: Hazard Ratio.
  4. HR > 1 indicates increased likelihood of achieving R/MM, favoring thymectomy.
  5. R/MM: Remission or Minimal Manifestation Status.
  6. *p < 0.0001. p values are not part of Bayesian analyses.
Back to article page

Orphanet Journal of Rare Diseases

ISSN: 1750-1172