How to interpret a survival plot

[@dakshadhingra]

Cannot thank you enough, Darren. This is truly awesome!

[@FavorpassGoal]

cos 4 mins is all I have to loose on YT. Thanks!

[@robertasampong]

Excellent explanation. Thanks Darren!

[@quintonwilson8565]

finally an intelligent person found. My professors are absolute garbage.

[@drsks6]

Thank you, nice brief explanation

[@syazwanfadeli4725]

Thank you sir!

[@pranayaryal]

Thanks Darren this was great

[@Sammy.a1287]

Very useful, thank you!

[@parlifunk1]

Thanks for that explanation!

[@devjat9110]

Short and sweet

[@tomhefferon8420]

Great video, makes perfect sense!

[@terrylaitw840]

Ez and concise. Thx

[@sharmainequake]

If median survival cannot be expressed for a cohort by the Kaplan Meier analysis because >50% has not experienced the outcome (eg death), then can we calculate median survival directly by any other way?

[@BettinaRyll]

Great video, thank you for that. Apart from the annoying EVENTS. This is an overall survival plot, so the EVENT means a patient's DEATH. Patients need to understand these curves because they tell them where the highest chances are to survive.....calling it an 'event' is obviously the correct technical term and might cater for non-patient sensitivities but doesn't help REAL patients to understand why this is so, so, so important.

[@darrendl1]

Thanks Bettina. I actually went back and forth on that one, not sure which was more appropriate. The plots and the related statistical methods are applied to all kinds of time-to-event data, not just mortality, but I do see and appreciate your point.

[@BettinaRyll]

There is hardly anything less factual than death. Calling it 'event' then leads to comments like 'not enough events have occurred to xxxxx' (that was one ASCO session on OS in Melanoma) EXCUSE ME? The point of medical research should be to save lives not to regret that not enough patients have died- this is obscene. I have lost too many people I cared about to Melanoma- including my husband and some dear friends- so it is time to realise what these data points actually correspond to as this e.g. also affects the way new trials are designed. Easy to be cool on 'events'- not so easy when it's called by what it is, especially when it is 'death'. Apart from that, great video- already on twitter- and I hope you are planning to do more? www.melanomapatientnetworkEU.org

[@BettinaRyll]

@@TheBlazingRiver oh help, are you new to this and have ever listened to how these graphs are discussed? If it's all about statistics, surely you can call a shoe a shoe. And these are DEATHS. And it very much matters whether you discuss *events* and everyone in their mind thinks *events* *parties* *weddings* *concerts* *whatever* while in reality we talk about people who died on that very trial. People like people from our community, so sorry to pop your bubble but we are the folks who can put names on those euphemistic *events*- they are our losses and our funerals. So- wake up.

[@mkoberoi]

amazing

[@zainababid9653]

Hey, thank you for the video. I want to ask if I need to make progression free survival graph of two drugs. I have 0 without progression and 1 with progression. Will I take the events with progression?

[@mamjad6274]

Could you please explain NR?

[@drpratikppatil7198]

perfect

[@siatemu199]

Hellow! Thanx for the video. Lets say i dont have any censored observations in my data.Meaning all experienced an event at different time intervals. What will happen to kaplan meier curve?

[@griftereric83]

if each event causes the graph to drop, why are there lines that have small spikes particularly at the end of the line?

[@darrendl1]

Those marks indicate when an observation is "censored", which means they drop out of the sample before they have a chance to experience the outcome (or hit the end of the follow-up period).

[@stinkavox002]

Great synopsis- what does it mean when the lines cross again?

[@darrendl1]

Crossing survival curves can indicate "non-proportional hazards", whereas "proportional hazards" are a key assumption of cox regression, which is by far the most common way to analysis these kinds of data. However, if you are just looking at the plot, crossing curves don't have any special meaning beyond what each individual curve indicates.

[@johannesfeder3485]

I don't understand the number of people at risk. Using the lowest line, the one you referred to most (Ipi >2xULN), as an example; after 24 months about 10% of the population didn't experience the event (let's say death). The corresponding number of people at risk is 1. Does that mean that only 9 people died up to that point? I don't understand to what total the 10% refer and how the number of people ar risk refers to that total.

[@GeorgeAscue]

Thaaaaanks

[@k-bala-vignesh]

Hi Darren, thanks for the video. A conceptual question- If all my data were censored, would it be meaningful to use Kaplan Meier estimation?

[@darrendl1]

If all the observations were censored, it means that nobody experienced the event, and so your survival "curve", produced by Kaplan Meier or any other method, would just be a flat line at 100% survival.

[@ranga_vasu]

facebook link not working

[@joannabusinessaccount7293]

Why not just say “death” rather than “event?”