Randy Neil Biostats 06: USMLE Step 1 Biostatistics - Exam Review Questions

You need true positives and false negatives .

Formula: TP / (TP + FN)

Absolute risk reduction = control event rate − treatment event rate

Power decreases and the risk of a type II error increases.

The result is not statistically significant.

Answer: B) Patients with negative tests who still have cancer

Sensitivity = TP / (TP + FN), so you need the false negatives.

Answer: B) 0.20

ARR = 15/50 − 5/50 = 10/50 = 0.20.

Answer: B) Statistically significant but clinically weak

The p-value is significant, but a 0.3-day difference is unlikely to be clinically meaningful.

Alright guys, so let's go a little bit quicker. It says, which of the following is an necessary to calculate sensitivity of this test. Now whenever I see the word sensitivity or specificity of positive value, we're always going to draw the box, okay? And I'll just put reality and test positive negative, positive negative. A new blood test for insect prostate cancer is evaluated in 300 males. A needle biop to the prostate gland is done on all men with serum prostate specific energy and concentrations greater than five. So a test is done if it's more than five. 100 men under go the biopsy. So we're going to put that there. I'll be through 100 men under go the biopsy. That means 200 did not. 90 of those are found to have prostate cancer. So let's just put cancer here, okay? And so how many people had cancer of those? They said 90, right? 90 of the 100 had the positive cancer. And five are found to have chronic prostateitis. Which of the following is necessary to calculate the sensitivity. Now we know sensitivity is top left going down. So one of these answer choices has to describe that box, right? Because sensitivity is 90 over 90 plus whatever that number is, okay? We just call it X for now. Is it A, the incidence of chronic prostateitis in the general population? Is that described with this boxes? It does not. So it's not going to be that. Is it B, the number of men with test results greater than five or normal byopsies specimen? No, because remember, if it was greater than five, then that would have been a positive test, right? So is it C prevalence of the chronic process cytosine and the general population? No, that does not describe this box. Is it D? A prostate biopsies a men with test results equal or less than five, right? Yeah, because this is the group that underwent the test. If it was greater than five, you fell into this test. If it was less than five, you

fell into this box. And this is the box that we needed to figure out sensitivity. So it's going to answer choice D. Now again, they see what they did here is they needed you to say that you need to find this box versus any of these other guys. And this box was described in words and all the step exam is real popular in wanting to do that. So know how to work this contact problem. This one says, which of the following best represents the absolute risk reduction for vomiting among patients in the as it's reminds me of group. So absolute risk reduction. We got to know our basic formulas is event rate, event rate control minus the event rate um placebo or treatment. And so is the absolute value. So really, I don't care which one comes first. A randomized control trials conducted to assess risk for development of gastroenterate intestinal adverse effects using as it's a medicine compared to a ridthrode medicine in the treatment of pertusses and children. They 100 children with pertusses enrolled 50 received as it's a medicine and 50 received a ridthrode medicine. So you see, zithro we're going to say 50 misses you already know I've really had to write this down. But 50 and 50 result show vomiting among five patients in the as it's a mind-sing group. Okay, five patients at a nausea of vomiting and then fifth compared to 15 patients in the a ridthrode in your risk for a group. So which of the following best represents the absolute risk reduction? Now since since there's only two of these, it doesn't matter which one you're going to call the treatment versus plus plus versus placebo. We just got to take the difference between the two. Now the event rate, the event rate, this is put the ridthro first since it's a bigger number. So make our lives a little easier. So it can be 15 out of 50. Okay, minus the other treatment, placebo, could be zithromycin which was going to be the

event rate was five people had the nausea out of the 50. Okay, and so when we subtract these, you know it's basic, you know basic math here. We've got 10 out of 50, right? And now this goes back to your high school algebra. You got to have a common denominator and all that kind of good stuff. And so you're going to get 10 out of 50 or 0.2. Okay, that's how you solve that one. So that would be, it's choice B, B is in boy. Okay, but you had to know the formula. Okay, absolute risk reduction. You can't rate control, minus the rate placebo, all under the absolute value flag. Okay, this one says which the following is the most appropriate conclusion about the effect of regular exercise on the risk of hit fracture. So we look up here with their answer choices, we're talking about statistically significant or increased risk. So in a cohort study a vitamin, the relative risk ratio for hit fractures among those who exercise is 1.2. Okay, so relative risk is 1.2. Now that's greater than 1. So that means there is some type of risk there. And so it's, it's risk for people who exercise. So what it's saying is if you exercise, you would have a 1.2 times greater chance to fracture your hit. But we have to look, always look at this, what is the confidence interval? What do we not want that confidence interval to contain? We don't want to contain 1.0. Because if that confidence interval ever contains 0, like 1.0, so if it was like 0.8 to 1.2, we don't like that, right? Because somewhere in there there was no difference between whether someone exercise or not. But what this is saying is that when data came back, you either had a 1.1 to 1.8 times higher chance of having a hit fracture if you exercised 95% confidence interval. So what does that data mean? What does this 1.2 mean? Statistically not significant? No, because it is significant, right? We had a good conference, 95% and this

endorsed it that it was actually good. So anything that has not significant is going to go away. So is it statistically significant overall decrease risk for statistically significant overall increase risk? Well the fact is it's greater than 1. So it's going to be an increase risk now. If all these numbers were a little bit different, say this thing was like 0.8, okay? Say it was 0.8 and then the confidence interval was 0.6 to 0.9, then you could have said if you exercised there's a 0.8 or less risk than doing nothing to hit fracture your hit. In that case it would have been a decrease. But the fact is this one was 1.2, it's 1.2 means it's greater than. So you're going to go with the answer choice D is in dog. Okay, now when you get something like this you better put a big smile in your face and know we're going to get these right. All right? I don't even read this garbage at top. I go straight to the box. Specificity, okay? Bottom, right? Going up, whatever we circle goes on top, whatever line we draw between the boxes goes on bottom 95 plus 5. So it's 95 over 100 and that's going to give you 95% answer choice E. Remember, go back and read that first video sensitivity, specificity. Positive, perfect value, negative, perfect value. This is the most guaranteed is going to be on the step exams and we've got to nail this, get it right. This one says five year old, well actually I've read the question, which of the following is the most appropriate interpretation of these study results. Okay? Whenever I see this I always start the first thought I do. I scan and see if there's a confidence interval in here. I don't see one off hand in this, but what I do see is a p-value and I say, okay, well p-value looks pretty good right because we care about the p-value being 0.05 or less because that means there's a chance, it's the chance of something occurring by, quote, chance. So

we want that to be a small number. So it says a five year of boys brought to the physician by his mother because of the two-day history of low-grade fever, cough, and runny nose. His temperature is 10,0004. Examination finds her to systemate diagnosed and come and cold. The physician refers to her randomized double-blind placebo control trial that he valued at the efficacy effectiveness of a new drug for the treatment of the cold. The meantime for the resolution of symptoms for the patient receiving the drug was 6.4 days. So if you receive the drug, your symptoms lasted for 6.4 days compared to the meantime 6.7 if you took a sugar pill. So there wasn't much difference, 6.4 versus 6.7, not too much, even though it had a good p-value. So what are my answer choices? The findings are clinically and statistically significant. I like statistically, but I mean clinically, 6.4 versus 6.7 is not too impressive. The findings are clinically and significant, but statistically significant. Okay, I do like that one. The findings are clinically significant, but statistically insignificant. No, because this is statistically, it was good. Okay, so it can't be that one. The findings are neither clinically nor significant. Okay, so the only answer choice that matches the status statistically significant, but clinically insignificant. When we look at that, answer choice B as in boy. Okay. Now, this one says it starts like about this type 2 error in all this. So anytime I see this, I automatically start thinking about the null hypothesis and in our box stuff and remember on your exam, you're going to get a little scrap sheet of paper essentially where you can, it's like laminated and got markers and stuff like that. And I always just kind of drew stuff on it. Just to get me a reminder, I always do this, alphabet of 1 minus beta, it was also known as the power, right? So anyways, says it

says if the study has been performed in a population of only 500 patients, which the following would have been most likely to increase. Okay, let's see what they're asking. A study is conducted to assess the effectiveness of a new drug for the treatment type 2 diabetes. A total of 1,000 patients, okay. With type 2 diabetes, our enrolled. Patients are randomly assigned to receive the new drug or standard treatment. The alpha and beta values for calculating probability are 0.05 and 0.2 respectively. Now you had to know alpha and beta, right? Alpha is going to be right here. Also known as the type 1 error. So we're going to put 0.05 there. It's like p value, right? And then beta, also known as the type 2 error. And that's just flat out for memorization. Gotta watch those other videos. Goes right there. So without even looking at this, what is the power, right? It's 1 minus beta. So it's 1 minus 0.2.0. So I got everything in the box. So results show that the new drug is significantly better than standard treatment. If this study has been performed in a population of only 500 patients, which of the following would have been most likely to increase. Okay, so what they're saying is they're going to try to reduce, you know, they're going to reduce the, what are they doing? If they're taking it down like this, they're going to reduce the power, right? When your number goes down, power, when your, the end goes down, the power also goes down. Okay? Right? Because that's what they always say. How do you make power go up? You increase the number. But in this situation, they're going from 1,000 down to 500. So power is going to power is going to decrease. So what would happen to Alpha in this situation? You know, probably Alpha is going, you know, it would probably, Alpha would probably go up. But that's not my very answer choices. What would it increase the chance of a type 1

error? Um, you know, actually if we did that, if we increased Alpha, that would actually decrease our chance of making a mistake, right? So it gives more leeway. So that, that would actually go down. So it's not going to be that choice. Chance of a type 2 error. Well, let's think about this. If the, we said if the end goes down, the power goes down. And if this goes down, if it went, if it went to 0.7 or 0.6, that means this guy's got to go higher. So the chances of making a type 2 error went up. So I really liked that one. And we know the power of the study isn't likely to increase. The power actually went down. And then since the TV specialty had nothing to do with this box, so I'm going to eliminate those. So the only answer that we're, that's going to work on this one is going to be answer choice, uh, be as annoying. But you see what they did? They decrease the power. And if you know that these things just add up to 1 and I decrease the power, then this guy beta or the chance of a type 2 error is going to go up. Answer choice, uh, be. All right. This one says, uh, which of the following is the relative risk? Good. All right. We like relative risk, uh, death from, from this type of cancer and individuals who take drug X compared to the individuals who do not take drug X. All right. Good. So it says a physician is conducting a richer perspectives review of a trial involving the use of drug X in patients with this specific disease. It is known that drug X is associated with increased probability of cancer in patients who use the drug. You know, blah, blah, blah, blah, blah. A total of 600 individuals, okay, with a specific disease or a clout new trial. So I'll automatically know if I'm dealing with this stuff, I know it's something I got a draw box on this. And you can draw the box however, you know, go with the words, but remember it's whatever's mentioned first is

what you write first. I know that doesn't mean much to you if you haven't seen that first video, but let's just work it out. So we know we are going to label our box reality and over here is actually the test positive negative, positive negative. It says there's 600 individuals who have this disease, okay, who are part of this, recluding the trial of the participants, 200 received drug X. So let's just put that one up here. Let's just put drug X up on our reality thing, okay, 200 of those N received received the drug, okay. So that must mean that 400 did not, okay, I can buy that, okay, and 400 did not, good 100 individuals who received the drug died. So let's just put, this was put died over here, okay, so died here. And so other 200 who received the drug, 100 of those individuals passed. All right, that's fair. And 100 individuals who did not receive the drug died of the same type of cancer. All right, so they're saying that of those 100 who did not receive the drug, who did receive the drug didn't die, okay. So based on this data, which of these is the following risk of death, okay, well I'm kind of missing something here, right, right, they did die. 100 individuals who received drug X died, all right, so 100 who received the drug died, got it, and 100 individuals who did not receive this drug died, okay, wait, so this is actually in the wrong box because the 100 individuals who did receive the drug, who did not receive the drug died, so this is the die category. And these guys did not receive the drug, these 400, so it's 100 of those guys died, okay. I, ironically, I still get 100 in this box because there was 200 people who received the drug, 100 of them died, 100 didn't, 400 people who did not receive the drug out of those, 100 people died, so by default, how many people in this box? 300, okay, I want to submit an error there with my explanation, but I'm going

to slow it down for a second to make sure we're clear. There's 600 people in this study, okay, and then all of those participants 200 receive the drug, that means 400 did not receive it. 100 of the people who received the drug died, okay, and 100 individuals who did not receive the drug died. Now, based on the data, which of the following is the relative risk of death, okay, death, right here, okay, these people died. From this type of cancer and individuals who take the drug, compared to people who did not take the drug. Now, when I do relative risk, I have to say, so the people who, and this is important because how did that, how did this read? Relative risk, people death, when the cytokines are who took the drug, compared to people who did not, people who took it versus people who did not take. Now, if you remember correctly, we said that a case control had a lot of stuff, there's one number of one number is associated with the odd's ratio, and that's one number over one number, relative risk is one number over two. So the people who took the drug is 100, right, 100 over two numbers. That's 100 plus 100. Over the people who did not take the drug, that's 100 over 100 plus 300. Now, again, I explained that in that first video, but you got to have this down, okay, because they're talking about relative risk here, and you got to go to this one number over two. What they said odd's ratio, you'd put one number over one number, okay, and that's a foreign language to you. Watch that first video. So when we solved this, we got 100 over 200 divided by 100 over 400. Now, if you do all your your simple math, you all your basic stuff, 400 over 100, okay, all these, all these guys cancel out, and I'm left with 400 over 200, or I'm left with two. So my relative risk of people who took it, of the people who take drug acts compared to people who did not of dying is a relative risk

of two. So is it A individuals who take the drug X have an equal risk of dying? Nope, because if that was the case, there'd been one individuals who take drug X have four times the risk of dying? Nope, it's not four, it should be two. Individuals take the drug X have three times, no, individual is take the drug X two times risk of dying, right? There, okay, the risk of dying yet, it can be determined. Very important problem, guys. They could have said relative risk, they could have said odd's ratio, it doesn't matter. You got to be able to draw this box. I could have labeled this thing opposite, it doesn't matter, because if you know, if you know how you can just, whatever goes on top is whatever you said first, the people who took the drug went on top, over the people who did not take the drug, okay? It seemed like a long problem, but at the end of the day, this is when you can guarantee to get right on your exam. All right, this says, when otherwise covarious are controlled, covarious, which of the following is the most appropriate conclusion, okay, and talks about all this risk, okay, let's just read it, because talking about here's some odd's ratio stuff, and confidence interval. Now, and I see confidence interval, what do I scan for? I scan for something that's garbage because it crosses one, okay? So I don't like to use that data. I'm going to use that same strategy when it comes to you drug ads. It says, a case control study is conducted to assess the risk factors predicting in patient mortality among geriatric patients with community-card pneumonia. Results of the study include the odds ratio shown below, okay? Remember odds ratio? It's so it's almost like saying there's a greater risk rate. It's a 3.3, 2.4, and 1.3, which are calculated from a multi-variable logistic regression equation, okay, too much information. So, when all these are controlled, which of

the following is the most appropriate conclusion regarding the use data? Is it the risk of inpatient mortality? Is greater for patients with hypotension than for those without? Okay, so it looks like, so if it looks like the risk of hypotension is 3, compared to that. So I mean, that one actually without hypotension, that looks pretty decent. So I'm going to go to question mark by that when I hate to jump on that and just move on to the next one. The risk of inpatient mortality is increased by hypoxemia. By more, more by hypoxemia than by hypotension? No, because hypoxemia is only 2.4, 3 greater risk, whereas the hypotension is over, it's a 3.3 times greater. So it's definitely not that one. The risk of inpatient mortality is increased when there's a pulmonary infiltrate. Now look, how come I don't like that? I don't even want to look at that, okay? Because why? The data is garbage because of the confidence interval crossed 1, because at some point during that study, it actually was, it was a one-to-one ratio. I mean, it didn't have a difference. So I don't like to use anything that has a confidence interval. I don't like to use any of the information as a confidence interval that crosses 1. So automatically, I'll take a look for me. The risk of inpatient mortality is significantly affected by all of them. No, I can't say that because again, I can't use this guy because of this right here across as one. So I mean, that, so the correct answer is going to be in-sure choice A. The risk for inpatient mortality is greater for patients with hypotension than for those without hypotension. Okay, 3.3. Okay, in-sure choice A, this one, all right, kind of a two-bagger here. Go straight to the questions, which of the following is the most accurate interpretation of these data, taking a small pulmeto. So again, you've got to interpret this chart. A lot of times, you can just

answer the questions just by looking at this because that's a pretty big paragraph to read. So when I look at this, it says patient reported versus position reported. There's a placebo data between the two and a salt and the, I mean, the treatment was the salt pulmeto between the two. The statistical significance actually looks pretty good, right, because the p-value is very low. We like that. So which of the following is the most accurate interpretation of the data? So at this point, I don't even have to really, I don't think I need to read this. I like looking at the charts. Improvement is seen in both physician assessed and patient-ness inpatient reported. Well, it looks like placebo versus treatment went from 58 to 74% in patient and 38 to 73 in physician. Okay, so that's decent. But I don't want to get scared off on this yet. Patient reported symptoms are more improved than physician. So are they saying that the patient improved more than physician where the numbers are a little higher, but what was the actual difference? You know, if I did 74 minus 58, six, one. So there was a 16% change in the patient line and then the physician one went from differences 63 to 38 is 25. So actually, there was a bigger change in the physician one than compared to the patient one. So we definitely knew it's not that one, kind of backwards. Statistical significance is not important compared with symptom improvement. I don't know, that's kind of a, that's a bit much. I don't think you can actually go prove that on that one. Statistically, significant changes in the physician assessed symptoms do not result in decrease symptoms. You know, we can't make that statement there. That's too much kind of a, gosh, there's a fancy term for it. But we can't, we can't pull that statement from from that data because statistics can change the physician. Do not result. Well, yeah, we don't know

as far as the significance, we have to compare those numbers. No conclusion can be drawn from the present information. I can't really say that. So I think we can actually, this, this first one looks actually pretty good. Improving is seen in both the physician assessed in the patient recorded, right? Remember what we said. It's almost like the drug at thing. It's, we can only put the answer that we can prove. And then only one of these answer choices that we're, that we're decent that we can prove is going to be actually answer choice A. This one says, based on the physician assessment, this physician assessment data, the number of patients who need to be treated with supplemental to show significant improvement in one patient is which the well they're asking number needed to treat one over the absolute risk reduction. But be careful because it says the physician, a physician assessment data, I'm of course asking me the second one, right? They want you to jump all over that first one. But the difference between treatment, no treatment, right? You've been rate control, my spend rate, uh, treatment. So anyways, it's going to be 0.63 minus 0.38. Okay, I'm going to take the difference between those guys. Um, heck, I already did it right there, I technically. Um, so that's going to be 0.25, which is actually I always like to think it's 25 over 100. And if I did the inverse of that, okay, whatever your math style is, again, number four, right? If I do this, the answer is going to be four. So the number needed to treat, this is going to be a, uh, four. So you really got to like these, uh, it's number needed to treat stuff. Very, very talk per question. And so we get this one, it looks like a parallel, our last one here. It's talked about statistical power in the study. Okay, when any time I see this whole power thing, I better go back to my whole null hypothesis box. I do

this. I do this. And I always put reality on top, I put a test, and then I can kind of start plug-in things in. A randomized control trial is conducted to assess the effectiveness of a new combination of drug and typhoid tense of therapy, drug experts compared to the standard hypenterpretensive single therapy. Study participants include 140 women, 70 percent, and 60 men ages 30 to 60. With baseline blood pressure 115, 95, or higher, the investigators find anti-hypertensive therapy is affected to treatment resulted in a blood pressure measurement below. Okay, so 140 over 90. So okay, when designing the study, the investigators set the probability of wrongly finding that drug experts more effective than standard therapy is 1%. So that would, they essentially said, look, we're going to set our, our p-value, or alpha, error is going to be at 1%. Or 0.01, right? The probability of wrongly finding that drug experts is more effective than standard therapy. What they're saying is any reality, there would be a no association, but the test said that there would be 1. So then they say they set the probability of wrongly finding the effectiveness of two drugs is the same as 10%. So if their test said, wrongly that there is no association, but in reality there was 1, they're saying that's going to be 0.1. What is the power? Now, no power is this box. Now, so here's the thing on this. They're going to do this to you. They're going to see if you know how to write in this no hypothesis stuff, but you got to know how to fill in the box and they're going to put it in words for you. But if you just find out, they're going to do this. They're going to say there's an alpha or p-value and then they're going to say here's the beta, but they're going to describe beta and words, and then they're just going to most likely they're ask you for power, or they'll give you the power and ask for

the beta. So in this one it's going to be the asking for the power. And all we know is we take 1 minus the 0.10 and we're going to get 0.9 and that's going to be 90% answer choice D. So guys, it's kind of all that I have. You know, I took these questions from the essentially from that USMLE.org webpage and just took all the ones that were biased. So they're expecting us to know these things. But again, go through them quick. The more problems that you see, the more that you can work, the better out that will be for the exam. So I hope to help and we'll see in the next video.