Discover and read the best of Twitter Threads about #statswithcoreim
Most recents (5)
2/ Let’s get started with a question: What type of bias do you think is present in this example? 
3/ The answer is Verification bias! Verification bias is present when:
✔️The reference standard (gold standard) is not used for all cases [= emoji] partial verification
✔️there is more than one reference standard 🟰 differential verification
✔️The reference standard (gold standard) is not used for all cases [= emoji] partial verification
✔️there is more than one reference standard 🟰 differential verification
1/ What are 3 Qs you can ask yourself when looking at non-inferiority trials?
Q1) Was the trial planned i.e. pre-specified as comparing an intervention which is non-inferior to control?
WHY?
Bc changing the analysis *afterwards* introduces bias #StatswithCoreIM
Q1) Was the trial planned i.e. pre-specified as comparing an intervention which is non-inferior to control?
WHY?
Bc changing the analysis *afterwards* introduces bias #StatswithCoreIM
2/ Q2) was the control treatment administered to the full std of care?
The trial relies on strict adherence to full standard of care for the control arm, otherwise the whole confidence interval shifts with a relative⬆️in benefit of the intervention compared to control
The trial relies on strict adherence to full standard of care for the control arm, otherwise the whole confidence interval shifts with a relative⬆️in benefit of the intervention compared to control
3/ Q3) Did investigators perform both a per protocol and intention to treat analysis?
Per protocol:
Exaggerates group difference when intervention is inferior
Less likely to result in false positive
Intention to treat:
Makes groups prognostically 🟰
Benefit of randomization
Per protocol:
Exaggerates group difference when intervention is inferior
Less likely to result in false positive
Intention to treat:
Makes groups prognostically 🟰
Benefit of randomization
1/ Let’s dive into #noninferiority trials for this edition of #StatswithCoreIM
But first, take a look at this figure for what superiority trials aim to assess: What’s better?
But first, take a look at this figure for what superiority trials aim to assess: What’s better?
2/ So how are NON-inferiority trials different?
They ask if a treatment is much worse than standard of care.
They ask if a treatment is much worse than standard of care.
3/ Let’s look at the possible outcomes of a non-inferiority trial.
✔️Superior and non-inferior
✔️Non-inferior
✔️Not non-inferior
✔️Inferior and not non-inferior
✔️Inferior and non-inferior
✔️Superior and non-inferior
✔️Non-inferior
✔️Not non-inferior
✔️Inferior and not non-inferior
✔️Inferior and non-inferior
1/ Welcome back to #StatswithCoreIM !
What would you tell this patient who inquires about lab cancer #screening test to help him “live longer”?
What types of bias can occur in determining whether a cancer screening test reduces mortality?
What would you tell this patient who inquires about lab cancer #screening test to help him “live longer”?
What types of bias can occur in determining whether a cancer screening test reduces mortality?
2/ Take a look at the bolded arrows below that illustrate that early detection doesn’t always mean better outcomes!
Length-time bias applies to slow-growing disease in which patients have a long phase without symptoms.
Length-time bias applies to slow-growing disease in which patients have a long phase without symptoms.
3/ Lead time bias applies to situations where patients are screened earlier, so they are diagnosed earlier, so they appear to live longer solely by nature of knowing they have the disease for a longer period of time.
Therefore, survival time⬆️⬆️
Therefore, survival time⬆️⬆️
1/ Hey #medtwitter, what do you know about diagnostic odds ratios and how they are used?
It’s time for another round of #statswithCoreIM!
Let’s start by considering dichotomous test characteristics:
It’s time for another round of #statswithCoreIM!
Let’s start by considering dichotomous test characteristics:
2/ Diagnostic Odds Ratios are different, and special, because they allow us to use one single number to describe how good a diagnostic test is.
3/ For example, let’s consider the DORs of some diagnostic tests of the flu:
😷Cough - 2.8
😷Fever - 4.5
😷Rapid flu swab ~ 15
😷Flu PCR ~ 100
(And stay tuned for another Steve & Janine podcast on the flu next Wednesday!)
😷Cough - 2.8
😷Fever - 4.5
😷Rapid flu swab ~ 15
😷Flu PCR ~ 100
(And stay tuned for another Steve & Janine podcast on the flu next Wednesday!)
