Biostats UW question querry

[aspiringmd1015]

---

Members don't see this ad.

A/c to UW, for case control studies, when selecting the control group, it doesnt matter if youre exposed to the RF or not, ie to select patients in the control group, you have to choose pateints just based on the fact that they dont have AML, regardless if theyre exposed to the RF or not
Here: kids w/ aml exposed to some industrial exposure, so the kids w/aml w/exposure are added as the cases, and the COntrols arent the kids w/exposure with NO aml, but are Kids that just dont have AML, regardless of their exposure status, and UW's explanation on why was ******ed. Any takers?

 

[Nabin]

Whenever research is done, we want to make sure the cases and controls are representatives of the population. Even the control should represent the population which gave rise to the cases, since it is the same population from where the cases appear. You want the demographics to be equally distributed among them as much as possible. Difference in exposure reporting can occur at so many levels, and its affected by many personal and social factors. . You want a sample which will represent the whole population and minimize any bias in selection. I forgot the UW explanation but when I think of it right now, this is the best I could come up with.

 

[aspiringmd1015]

I hate biostats w/a passion.

 

[aspiringmd1015]

can someone also eqxplain why NNT is 1/ARR. im trying to understand the equation

 

[Nabin]

You do a trial and take 200 people. You give 100 people drug A to treat a certain disease they have. You don't give anything to the other 100 (They don't look so happy!). At the end, you find that 8% out of the 100 treated patients died, whereas 10% of the untreated patients died. (Ignore the fact this drug is an absolute waste of time and money!)
So, you see that the death of 10-8% i.e. 2% patients was prevented by this drug A. So, when treating 100 people, you prevented the death of 2% patients, i.e. you lessened mortality by 2 patients (2% of 100).
You treat 100, you prevent deaths of 2.
You treat 1, you prevent deaths of (2/100) patient(s).
You treat 50, you prevent deaths of (2/100) * 50 patients, i.e. 1 patient. This means in order to prevent 1 death, you have to treat 50 of them.

Seen another way, NNT = 1/ARR = 1/2% = 50. You need to treat 50 to reduce the death by 1 patient.

 

[aspiringmd1015]

okay so im extremely **** at math, lost me after you treat 1, you prevent deaths of 2/100

 

[Nabin]

okay so im extremely **** at math, lost me after you treat 1, you prevent deaths of 2/100

Just go simple. Its more about interpretation than math. You treated 100 of them and could prevent the deaths of 2 patients. Its just asking how many you need to treat to prevent the death of 1 patient. 100 for 2, 50 for 1 - simple unitary method: this just to understand where to go about relating ARR with NNT. The numbers may not be so simplistic while doing questions. So, when you see the bigger picture using that formula gets you the number.

 

[aspiringmd1015]

okay just got it, so the ARR will give you 2%, so 2/100, flip it and you get the NNT. which is 100 patients, and 2 will be prevented, or 50 patients and 1 will be prevented. got it! for those who are wondering, the UW biostats bank is worth it, despite it being 40$ for only a few questions, the explanations are nice.

 

[Keto]

Hi

the NNT (or number needed to harm or number needed to screen) are one of the most important numbers we have in medicine.

So we learn, that everybody gets aspirin immediately after a major heart attack*. For Step 1, we study the mechanisms of action of aspirin, the indications and side effects, but the question remains: how many people will you save from aspirin? How many people do you have to treat, in order to prevent one person from dying? That's where NNT comes into play, which answers exactly this question.

*its 42
(ISIS2: Randomized trial of intravenous streptokinase, oral aspirin, both, or neither among 17187 cases of suspected acute myocardial infarction. Lancet. 1988 Aug 13;2(8607):349-60.)

Great website is http://www.thennt.com/

 

[iamgonnaeat]

Here is a good one to help anyone that's having trouble with biostats;
You can get full access for 10 bucks, it's pretty cheap. I think it was worth it