"You're a Gemini? Sorry, no ASA for you." - Awesome stats review

Let's face it. Stats is incredibly dry, but pretty important to know when we want to back our drug recommendations with some data. Today, my preceptor gave a talk about the important things to know about stats in about 30 mins. Here goes stats in bullet points...

Statistics

• The null hypothesis = no difference between the groups.
• Type 1 error: We say A>B, but A=B
• Probability of making a Type 1 error = alpha = p-value = 0.05 (usually)
• Only consider Type 1 error if you REJECT your null hypothesis (there is a difference)
• Type 2 error: We say A=B, but A>B
• Probability of making a Type 2 error = beta = 0.20 (usually), and this is related to the POWER of a study (1-beta = power)
• Only consider Type 2 error if you ACCEPT your null hypothesis (there is NO difference found)
• Confidence Intervals: We're used to seeing 95% CI, related to p=0.05
• Prospective studies: if the interval includes 0, NOT significant
• Retrospective and Odds Ratio: if the interval includes 1, NOT significant
• Example: If Drug A makes you lose 10 pounds, and Drug B makes you lose 2 pounds. Then the difference is 8, correct? Say the results of this study say Drug A is better since it makes you lose 8 pounds more with a 95% CI of 6-10. What does this mean? It means that if you do this study over and over again, you're 95% certain that study subjects will lose between 6-10 pounds. If another drug has a result of 95% CI of 1-15, then which drug would you want to be on? The drug where you're 95% sure you'll lose at least 6 pounds, or the drug where you're 95% sure you'll lose at least 1 pound?
• Absolute Risk Reduction (ARR): If the risk of dying with placebo is 5%, and the risk of dying with Drug A is 2%, what is the ARR? Simple: 5-2 = 3%. **This is the clinically important number**
• Relative Risk Reduction (RRR): (Placebo risk - Drug risk)/Placebo risk. (5-2)/5 = 0.6 = 60%. This is the number that drug manufacturers will use to sell their product. Doesn't a relative risk reduction of death of 60% sound so much better than an absolute risk reduction of 3%? Don't be fooled! 
• Number Needed to Treat (NNT): This is the number of patients you would have to treat to get a specific effect. If the drug is to reduce death, then it's the number of patients you would have to treat to save a life. The formula for this is: 100/ARR. The lower this number the better. The way to interpret it is according to whoever is treating a specific patient, and things like cost of medication, length of treatment, etc. must be taken into account in order to make a good call. 
• Number Needed to Harm (NNH): This is the number of patients you would have to treat to get a specific adverse event to happen from the drug. If the adverse effect is causing death, then it's the number of patients that would be treated before a death happens. The formula for this is 100/percent increase in ADR. The denominator is found in whatever study you're looking at. So, if the study says Drug X caused 5% headaches, and 2% headaches on placebo, then the denominator would be the difference (3%). 
• Lastly, subgroup analysis is BS for the most part. If researchers don't find a difference between the groups they're analyzing, they might try and subgroup their analyses to find some kind of difference. For example, there was a study about to be published for the use of aspirin after a heart attack. However, the reviewers of that journal (Lancet) wanted the researchers to perform a subgroup analysis on patients with diabetes and without diabetes. The researchers also decided to do a subgroup analysis on which astrological sign the patients were (Gemini/Libra or other). There was no difference between diabetic and non-diabetic patients and their use of aspirin after a heart attack. However, apparently, if the patients were Gemini or Libra, they had a higher death rate after using aspirin. Pretty crazy! Oh and the astrological sign data is actually published in this article! 

COPD
GOLD guidelines: Stage 1 - FEV% predicted >80%. Stage 2 - FEV% predicted 50-79%. Stage 3 - FEV% predicted 30-49%. Stage 4. FEV% predicted <30% or respiratory failure symptoms. In COPD, anticholinergics tend to be better than beta-2 agonists. Steroids and theophylline are used as well.

Anticoagulation 
If a patient is having A-fib symptoms for more than 2 days, then the guidelines say to anticoagulate them 3-4 weeks before cardioversion, and then 1 month afterwards. If a patient is having A-fib symptoms for less than 2 days (and no mitral valve issues, prosthetic values, or history of embolism), then the patient will be fine with heparin throughout pericardioversion and 24 hours after cardioversion. OR the patient can be put on warfarin or dabigatran until therapeutic anticoagulation is reached (INR 2-3). If immediate cardioversion is needed, then patient would be put on heparin, and then warfarin for 4 weeks after cardioversion.