When i so easily learn new stuff through internet like this i cant help to wonder how the other generations learned the same through complex books or fast speaking live teachers... What a time to be alive
I got my 1st bachelor's back in the day before calculators and internet. Because it took longer to gather, calculate and analyze information, much less was expected. I would not trade the convenience and efficiency of today's technology for what we had back then. But I do miss the adrenalin rush of solving a problem or finding information after weeks of research. No one could take away that accomplishment. It is nice to have the information at our fingertips, but the thrill of the hunt is gone.
@@feurigerStern It's definitely another world than before 2000! (just saying as a current student who doesn't even know how the Western world looked like back then)
I graduated from Johns Hopkins back in 1999 and LET ME TELL YOU students are so LUCKY now. Now I work with Med Students and these here internets and youtube are saving my life!
Cool explanation, I have a quick question. Here, in the final example you have a HR of 0.64 which is the probability of having the event (death), thus, you have a 36% probability of not having the event (surviving). How do you interpret a HR of 1.95 in terms of probability?
Terry, Your videos have been so helpful! Thank you! I do have a question! How do you interpret a hazard ratio that is a whole number? For instance, 3.3 or 4.5? I understand them when they are below 1 (a decimal), but I am confused about what the number means when it is a number above 1.
Very nice presentation but does not emphasize how easy it is to misinterpret this data. In relative risk or hazard ratio, key words are 'relative' or 'ratio.' HR is still 0.64 (and 36 % reduction in risk) for 1,540 cases in controls and only 1000 for statins. In fact, it was 154 (~3 %) controls vs 100( ~1 % statins). An absolute reduction in 2 % is not as clear-cut as 36 % which means you need to analyze the details of the trial (and reports of side-effects) before you take a statin.
I'm trying to interpret a HR (Hazard Ratio) study done on smoking tobacco. The study looked at non-smokers vs. pipe tobacco smokers over the course of, like, 30 something years. For non-smokers, the HR was 1.0 (control group) For pipe smokers, the HR was 1.2 I'm interpreting this as "A pipe smoker is 20% more likely to get _some_ kind of cancer or complication due to pipe smoking, versus a non-smoker for that same cancer." But my question becomes: Is there a reliable HR study that shows non-smokers, pipe tobacco smokers, cigarette smokers, cigar smokers, and vape users? Thank you. Very easy to follow.
why it is needed to reduce the actual HR (.64) from 1??????? is that not possible to interpret as it is???????????????? how we know the correlation is negative?
This video FAILS because you cannot *interpret* hazard ratios (and relative risk) without understanding absolute risk. Thus this video is overly reductive as they MUST be taught together for a correct interpretation. It also fails to sufficiently emphasise that the reduction at any point in time is effectively a RELATIVE RISK calculation as opposed to an absolute risk calculation, and that these large relative-risk percentages are used by researchers and pharma to over-egg their findings and advertising. For example, you might have a hazard ratio of 0.5, which at any point on the curve is a 100% reduction in relative risk, but yet the absolute risk might simply change from 0.3% to 0.6% which in some instances would be regarded as a failure or at least a waste of time. A concrete example is with statins, which have such poor absolute risk reductions it’s surprising so many doctors prescribe them…it’s because most doctors don’t understand the difference between relative and absolute risk; and that hazard ratios are relative. Lastly, a crucial aspect of interpreting hazard ratios is when they can be used to show the correlation is not just an association but is actually causal. For example with smoking and lung cancer, a hazard ratio of around 30 has been reported in some studies, which shows you the magnitude we need to hit to demonstrate real significance. Using this scale one might question the significance of any hazard ratios below 5 or at least 3-4 in studies trying to identify the drivers of chronic disease.
great video. I got confused at the end when I was thinking to say: 64% as many of the patients in the treatment group will have an event at any time compared to the placebo group. But you talk about 34% so I got lost...
why isn't relative risk? You are comparing the risk of one group (treatment) to that of the another (placebo) and find out if there's reduction/increase in the onset of event
Great video ...helped to understand RR and HR better.RR is at any particular point only but the HR take the time to event into the equation ,so HR is during the period not at the end point. So can we calculate absolute HR like ARR? if so how?
Thank you Dr Shaneyfelt this tape is great i am using it in my preparation for USMLE step 3 watching your tapes again and again gives me hope about the exam .
Thank you. I shared this video with family members who are not medical researchers to help them understand a journal article on gastric cancer survival with and without chemotherapy. This video really helps, a lot of non-medical people don't like words. They prefer videos. Thank you.
amazing post! i have a question though if i may....i noticed that the p value is very significant at the end though the HR is included within the CI...is that ok? i mean we only worry about including the value 1 in the interval correct? just confirming things here:)
You mention other videos you've made that address certain topics more in depth: could you please either link to them IN the video or BELOW the video (or both). Please.
hi Dr, I really need your help in knowing how to calculate hazard ratio and confidence interval via cox regression manually(most times, the software is used in generating the beta coefficients, SE and -2log likelihood values, correlation matrix. then you're asked to compute those I mentioned) please guide me!!
The hazard ratio number seems to distort the implications of the study. Total events are 100 for the statin and 150 for the placebo. Divide each by 3200 - the number at the end - and you get 4.8% event for the placebo and 4.68% for the statin. So the real difference is very very small. It’s even smaller when you view the total picture. Yet the study could be hailed as reducing death rate from 150 to 100 or 33% or that skipping the statin increased rates by 50%.
Please double check your math. If you work relative to 3200, then 150/3200=0.047 and 100/3200=0.031 and 0.031/0.047=0.66, similar to the HR. In any case, why divide by the survivors (3200) rather than by the initial population (5137 and 5168)? This calculation would be more meaningful as it would be the integrated probability of an event occurring over the initial populations of subjects. If you work this through, you'll again get a ratio that is close to the HR. It seems to me the data _do_ show a reduction from 150 (placebo) to 100 (treatment) in the integrated events over 3.5 years. I'm new to this subject, so perhaps I'm confused, but it does seem like you have an arithmetic error in your comment.
@@ef2b My math may be off but why divide the percentages? It looks like the statin reduces your risk from .047 to .031. The. percentage difference might seem large but it doesn't seem worth taking a statin , with all its side effects, to reduce risk by .016 or 1.6%. RRR numbers are often misleading. It exaggerates differences. See ua-cam.com/video/7K30MGvOs5s/v-deo.html and google "relative vs absolute risk."
Thank you so much Terry for your video !! it was so helpfull ! But there is something I can't catch. During your example (LLA study) Why the number of at risk people decreased from 5K to 1,8k, whereas the number maximal of event in the placebo group is just 154 ? Thank you for your answer !!!
When i so easily learn new stuff through internet like this i cant help to wonder how the other generations learned the same through complex books or fast speaking live teachers... What a time to be alive
I got my 1st bachelor's back in the day before calculators and internet. Because it took longer to gather, calculate and analyze information, much less was expected. I would not trade the convenience and efficiency of today's technology for what we had back then. But I do miss the adrenalin rush of solving a problem or finding information after weeks of research. No one could take away that accomplishment. It is nice to have the information at our fingertips, but the thrill of the hunt is gone.
@@feurigerStern It's definitely another world than before 2000! (just saying as a current student who doesn't even know how the Western world looked like back then)
I graduated from Johns Hopkins back in 1999 and LET ME TELL YOU students are so LUCKY now. Now I work with Med Students and these here internets and youtube are saving my life!
@@feurigerStern The thrill now is in hunting for the wheat amidst all the chaff.
Thank you for aiding in the saving of my Clin Epi grade!!!
Excellent video explanation. Thanks Terry
Thank you for this simplicity on the explanation
Thank you! Good material and perfect example
Thanks for explaining it beautifully
great video, very very helpful
Excellent, Sir.. thank you
Thanks .Great clarity in expression. thanks again
Excellent video, thank you for uploading it!
what a great video!
very helpful refresher on hazard ratios. Do you have one on Kaplan-Meier curves?
Great video, thanks for the time you took explaining these concepts
very easy to understand
so amazing!
Great video, thanks.
Great Video. Thank you!
Very good explanation!
Cool explanation, I have a quick question. Here, in the final example you have a HR of 0.64 which is the probability of having the event (death), thus, you have a 36% probability of not having the event (surviving). How do you interpret a HR of 1.95 in terms of probability?
95% or 1.95 times more likely of having an event than someone in the other group. If it was 2.95 it would be 195% more likely.
@@UABEBMcourse Thank you very much for your answer!
clear and concise. Thank you.
Nice
Brilliant - thank you
thank you bro
Terryyyyy. Great video.
when the professor gave example of hazard ratio at 2:00 why did he emphasize proportionally for 2 and 0.5 HR?
how did you do it can you share with me , thank you
Terry,
Your videos have been so helpful! Thank you!
I do have a question! How do you interpret a hazard ratio that is a whole number? For instance, 3.3 or 4.5? I understand them when they are below 1 (a decimal), but I am confused about what the number means when it is a number above 1.
Something like risk is 3.3 times.higher than those in control group
@@UABEBMcourse or 330℅ increased risk. 1.3 means 30℅ increased risk,no?
Thank you!
Thank you
please upload more
10 years later and this video is still incredibly helpful for us students. Thank you so much!
thank you!
thank you
Very nice presentation but does not emphasize how easy it is to misinterpret this data. In relative risk or hazard ratio, key words are 'relative' or 'ratio.' HR is still 0.64 (and 36 % reduction in risk) for 1,540 cases in controls and only 1000 for statins. In fact, it was 154 (~3 %) controls vs 100( ~1 % statins). An absolute reduction in 2 % is not as clear-cut as 36 % which means you need to analyze the details of the trial (and reports of side-effects) before you take a statin.
Thank you! Very helpful
I'm trying to interpret a HR (Hazard Ratio) study done on smoking tobacco. The study looked at non-smokers vs. pipe tobacco smokers over the course of, like, 30 something years.
For non-smokers, the HR was 1.0 (control group)
For pipe smokers, the HR was 1.2
I'm interpreting this as "A pipe smoker is 20% more likely to get _some_ kind of cancer or complication due to pipe smoking, versus a non-smoker for that same cancer."
But my question becomes: Is there a reliable HR study that shows non-smokers, pipe tobacco smokers, cigarette smokers, cigar smokers, and vape users?
Thank you. Very easy to follow.
This is an excellent and simple tutorial, thank you.
love you man
why it is needed to reduce the actual HR (.64) from 1???????
is that not possible to interpret as it is????????????????
how we know the correlation is negative?
How Could I interpret or optimize a high value of HR in survival analysis (Exp : HR= 3.12e2, p=0.003)?
This video FAILS because you cannot *interpret* hazard ratios (and relative risk) without understanding absolute risk. Thus this video is overly reductive as they MUST be taught together for a correct interpretation.
It also fails to sufficiently emphasise that the reduction at any point in time is effectively a RELATIVE RISK calculation as opposed to an absolute risk calculation, and that these large relative-risk percentages are used by researchers and pharma to over-egg their findings and advertising.
For example, you might have a hazard ratio of 0.5, which at any point on the curve is a 100% reduction in relative risk, but yet the absolute risk might simply change from 0.3% to 0.6% which in some instances would be regarded as a failure or at least a waste of time.
A concrete example is with statins, which have such poor absolute risk reductions it’s surprising so many doctors prescribe them…it’s because most doctors don’t understand the difference between relative and absolute risk; and that hazard ratios are relative.
Lastly, a crucial aspect of interpreting hazard ratios is when they can be used to show the correlation is not just an association but is actually causal.
For example with smoking and lung cancer, a hazard ratio of around 30 has been reported in some studies, which shows you the magnitude we need to hit to demonstrate real significance.
Using this scale one might question the significance of any hazard ratios below 5 or at least 3-4 in studies trying to identify the drivers of chronic disease.
So clear. Thank you so much!
great video. I got confused at the end when I was thinking to say: 64% as many of the patients in the treatment group will have an event at any time compared to the placebo group. But you talk about 34% so I got lost...
so do you do 1-hazard ratio and this gives you the percenatge if you times by 100? and this tell you the relative risk compared to the placebo?
Thanks proffesor, especially grateful about the last comment regarding the difference between HR and RR.
Thank you! Step 3 tomorrow. Hopefully I get all question correct. :(
why isn't relative risk? You are comparing the risk of one group (treatment) to that of the another (placebo) and find out if there's reduction/increase in the onset of event
A really crisp, concise and understandable video. Thanks a lot !!
Hazard ratio with confidence interval with p value explain please
Great video ...helped to understand RR and HR better.RR is at any particular point only but the HR take the time to event into the equation ,so HR is during the period not at the end point. So can we calculate absolute HR like ARR? if so how?
Thank you Dr Shaneyfelt this tape is great i am using it in my preparation for USMLE step 3 watching your tapes again and again gives me hope about the exam .
Here cuz my step 3 medical exam wants us to know this 😢😢
thank you so much now I understand it alhumdillallah ^-^
When Hazard ratio = 0.00 what is the interpretation dr??????
Thank you!!! This was an excellent explanation. Very thankful I found I this video.
Thank you! Really great and simple explanation.
Couldn't be any clearer thank youuuuuu !
Thank you. I shared this video with family members who are not medical researchers to help them understand a journal article on gastric cancer survival with and without chemotherapy. This video really helps, a lot of non-medical people don't like words. They prefer videos. Thank you.
Thank you so much sir. God bless you
my god i just understood, what is HR
amazing post! i have a question though if i may....i noticed that the p value is very significant at the end though the HR is included within the CI...is that ok? i mean we only worry about including the value 1 in the interval correct? just confirming things here:)
You mention other videos you've made that address certain topics more in depth: could you please either link to them IN the video or BELOW the video (or both). Please.
Thanks sir
it is of great help to me
hi Dr, I really need your help in knowing how to calculate hazard ratio and confidence interval via cox regression manually(most times, the software is used in generating the beta coefficients, SE and -2log likelihood values, correlation matrix. then you're asked to compute those I mentioned) please guide me!!
thank you
Thank you
Thank you, Dr. Shaneyfelt for this easy-to-understand explanation!
The hazard ratio number seems to distort the implications of the study. Total events are 100 for the statin and 150 for the placebo. Divide each by 3200 - the number at the end - and you get 4.8% event for the placebo and 4.68% for the statin. So the real difference is very very small. It’s even smaller when you view the total picture. Yet the study could be hailed as reducing death rate from 150 to 100 or 33% or that skipping the statin increased rates by 50%.
Please double check your math. If you work relative to 3200, then 150/3200=0.047 and 100/3200=0.031 and 0.031/0.047=0.66, similar to the HR. In any case, why divide by the survivors (3200) rather than by the initial population (5137 and 5168)? This calculation would be more meaningful as it would be the integrated probability of an event occurring over the initial populations of subjects. If you work this through, you'll again get a ratio that is close to the HR. It seems to me the data _do_ show a reduction from 150 (placebo) to 100 (treatment) in the integrated events over 3.5 years. I'm new to this subject, so perhaps I'm confused, but it does seem like you have an arithmetic error in your comment.
@@ef2b My math may be off but why divide the percentages? It looks like the statin reduces your risk from .047 to .031. The. percentage difference might seem large but it doesn't seem worth taking a statin , with all its side effects, to reduce risk by .016 or 1.6%. RRR numbers are often misleading. It exaggerates differences. See ua-cam.com/video/7K30MGvOs5s/v-deo.html and google "relative vs absolute risk."
Brainstorming
It was very clear
"Intended for beginners". That's great; I'm a total beginner. And the sound quality is great - I can hear and understand every single word.
THANKYOU!
very brief but spot on.
Thank you so much!
Thanks, Dr. Shaneyfelt. A great and succinct explanation :)
Great explanation!
Thank you sir!
thanks you smooch
Thank you
Very good explanation! Thanks.
Thank you so much Doctor !!
You just saved me ♥️♥️!
Thank you Dr. Terry for the clear explanation
THANK YOU!!!!
Super simple
you are awesome , u made it very easy and simple , plz make more statistic videos
What???
Thank you for this video Dr.Shaneyfelt
Fantastic explanation so well presented thank you
So hazard ratio is more accurate than relative risk in other words.
Thanks
Nice to me
Thank you very much. You made me understand it in 5 minutes.
Thank you so much Terry for your video !! it was so helpfull ! But there is something I can't catch.
During your example (LLA study) Why the number of at risk people decreased from 5K to 1,8k, whereas the number maximal of event in the placebo group is just 154 ?
Thank you for your answer !!!
As events happen or people leave the study the number at risk goes down.
simple and very clear. Thanks for the video.
Thank you Terry ! it helps a lot !
thank you for your video, it is easy to understand!
Thank you Dr. Shaneyfelt for the video.
thank you for this simple and effective video
Thanks for the simple explanation and having the video uploaded!
thank you very much! it really helps