Harm and Causation: When Bad Things Happen to Good Studies - - PowerPoint PPT Presentation

Harm and Causation: When Bad Things Happen to Good Studies

DANIELLA ZIPKIN, MD KEN GOLDBERG, MD

Exposure Outcome

Cohort Study

Exposure Outcome

Case Control Study

+ _

Exposure Outcome

Exposure Outcome Confounding

Diet Soda Type II Diabetes

Confounding

Have to think of it Have to measure it Have to quantify it Adjustment occurs in a regression analysis

Validity in a Cohort study

Does everyone in the study begin and end with the same chance of having the outcome develop, aside from the exposure of interest?

What does it mean for everyone to have an equal chance of the outcome?

(1) Were patients similar for prognostic factors that are known to be associated with the outcome (or did statistical adjustment address the imbalance?) (2) Were the circumstances and methods for detecting the

• utcome similar?

(3) Was follow-up sufficiently complete? Accounted for drop

• uts?

Let’s assess the validity

• f this

paper together!

Hazard Ratio as an alternative to Risk Ratio

Define “risk” (or event rate):

RISK = Nevent/Total N

A little about Survival Analysis

2 4 6 8 10 12 14 16 18 20 1 2 3 4 5 6 7 8 9 10 11 Time since trial start Subject Number

Survival Analysis 2

2 4 6 8 10 12 14 16 18 20 1 2 3 4 5 6 7 8 9 10 11 Time since enrollment Subject Number

Events Time Imagine two studies which end up at the same place, but take very different paths to getting there…

Events Time

The HAZARD RATIO is equivalent to the RISK RATIO at any point in time

Define Hazard Ratio:

HR = “RR at any point in time during the study”

An Example:

My hairdresser’s dilemma “Can I ask you a doctor question?”

Odds ratios and risk ratios

Define “risk” (or event rate):

RISK = Nevent/Total N

Define “odds”:

ODDS = Nevent/Nwithout event

3 1

Time to play with numbers…

The Case: Your “friend” has just been through a grueling week immersed in this strange way of life called “EBM”. Your friend is exhausted. Your friend needs a drink. But your friend has a golf game scheduled for tomorrow at 8am and really does not want to be hungover. Which bar should your friend go to in order to minimize the chances of being hungover? You don’t know the answer yet, but to prepare for next year you run a study: You post a study assistant at the door to the two most popular watering holes in the area and enroll volunteers who will accept a call the next day to get the hangover report. Over the course of 3 months you enroll 100 people at each bar.

Sample teaching abstracts