nf0279: i'm namex and i'm very pleased to be lecturing in this course and i saw some of the things on the board and i thought probably what i'd do is start with a single equation and this is the only equation you're going to see in this lecture and it's on the board there now now what does that tell you does it look even vaguely familiar to anyone [laughter] no i've probably got it wrong i thought it was something like the equation of relativity and the real reason i put it up there is that Albert Einstein was one of the very famous people to have an abdominal aortic aneurysm and i'm going to talk to you about abdominal aortic aneurysms so a more clinical flavour and how we use trial error and statistics in a real health problem i've sorted myself out with how to use this but i don't know whether i've got any sort of pointer nf0280: yes there's a mouse nf0279: i use the mouse nf0280: yeah nf0279: er well that's great 'cause the screen's gone off so i er oh okay right fine i've got you this is a cartoons of the aorta this is your aorta's nice and young and healthy these are the main branches of th-, the renal arteries this is the diaphragm as you get older the i mean i suppose one of the most obvious facial characteristics when people get older apart from greying hair like mine is wrinkles well just like your face wrinkles your blood vessels wrinkle too in a sense and they start sagging and this is what we call ectasia but in some people when they get older they actually balloon out their abdominal aorta and you probably remember from your anatomy if you've got that far that this bit the but-, the iliac bifurcation is about at the level of your umbilicus this nice little swelling of the aorta doesn't do anything you don't feel it usually no problems in fact many people have no problems until it grows so big that it bursts and the diameters here whoops normal aorta somewhere between one-point-six and two- point-two centimetres and we start to call it an aneurysm when the diameter here exceeds three centimetres and they can grow really very big up to fifteen centimetres or more so this may not provide any symptoms but when this happens and the blood pours out into your belly onto the retroperitoneum extreme pain and collapse and the survival rate is very poor we can detect aneurysms using ultrasonography even though a person may not know they've got them if we do an ultrasonogram of the aorta we can see here this is an example of a very large aorta and in the middle here you've got colours 'cause this is colour flow showing the blood flow so here you've got an aorta slightly irregular but probably about seven centimetres in diameter if you go and do this to the general population and you choose men over sixty years you'll find that five per cent of them have a small swelling of the aorta or an occult abdominal aortic aneurysm so it's not rare it's quite common and rupture of an aortic aneurysm is a common cause of sudden death and often you will see on a death certificate died from a heart attack or myocardial infarction actually it would be very difficult to discriminate if you didn't know anything about the history whether it was from a sudden major heart attack or a ruptured aortic aneurysm now the natural history of this condition is that once you've got a swelling it gradually increases in size and just like a balloon you might suspect that as it increases in size there's a bigger chance of it bursting or rupturing we've actually quantified this in over seventeen- hundred patients and the way we've done this is to break these aneurysms down by size we've put them into three groups very small small and larger and we've looked at a crude rupture rate per a hundred person years very very small for these aneurysms becoming significant for these aneurysms and more than one in four likely to rupture for these larger aneurysms i understand you haven't learned about hazard ratios but this is another way of quantifying the risk as to how much the risk increases for each centimetre in diameter of the aorta and because various other things might influence rupture such as smoking status and blood pressure we've needed to adjust for those and from these series we looked at a hundred-and-three patients with ruptured aortic aneurysm and this was their fate twenty-six or about a quarter died without even reaching hospital fifty-three arrived in hospital but were too sick or ill to contemplate corrective surgery corrective surgery was planned in less than a quarter and er only a very small proportion about ten per cent were alive thirty days after the surgery so pretty dire consequences if your aneurysm ruptures what does the surgeon do well the surgeon opens you up and has a look at your aneurysm and this is one here and he opens it up and he corrects it by inserting a Dacron tube or she preferably she's have got much better hands and more nimble hands than men whatever namex might say in namex [laughter] so this is a cartoon of the operation now then you can't do this operation without some risks and it's quite difficult to ascribe a risk to this operation 'cause you can do this operation electively that is if you find by chance a large aneurysm in a patient say an eight centimetre one you might think the risks of this bursting are so high i need to do an operation to correct it if you go and read all those learned journals you will see that the risk of this operation depends very much on the type of study whether it is prospective or retrospective whether it's population based hospital based or comes from a very selected hospital such as the National Heart and Lung Hospital and perhaps what you'll notice here is that the As the population based studies indicate that the mortality rate for this operation is quite high it's er probably somewhere in the region of eight per cent if you go and look at hospital based studies particularly retrospective ones the mortality rate drops and selected series published in the nineteen early nineteen-nineties from a very smart important hospital suggested that the mortality rate in really good surgeons' hands was only about two per cent now of course how do you report your r-, mortality rates you usually keep a record of consecutive patients that come to see you with this condition you can start that record when you like and you can stop it when you like so supposing that i have approximately two-hundred patients like this come to see me each year at the beginning of two-thousand-and-two the first two patients operated on died within thirty days after surgery so i thought it would be better to start my series on the first of February two-thousand-and-two we then did a hundred operations and all the patients survived and did wonderfully by the time we came to patient a hundred-and-three in October two-thousand-and-two we had a little bit of run of bad luck and a few more patients died so when we wrote up this series for a learned journal we decided only to include those consecutive patients between the first of February two-thousand-and-two and the thirtieth of September two-thousand-and-two 'cause this gave us a perfect record it's very difficult to work out what the true mortality rate is for this operation but quite clearly given that there is a significant mortality rate if you find one of these aneurysms in somebody what do you do do you let it grow and rupture or do you offer them an operation and where's the balancing point in terms of the size of this aneurysm and what else does it depend on and if you ask a vascular surgeon this person who operates on these they'll think it's going to be depend on the age perhaps the sex how fit are these patients can they still walk about have they got very good lung function and other physiological parameters so to find out about this and this was now about twelve thirteen years ago we did a survey of vascular surgeons in Britain and it's very useful to get information via professional bodies and it turned out that surgeons really had no idea what to do what was the best policy if they found a patient who had one of these swellings which was between four and five-and-a-half centimetres in diameter they didn't know this is a grey area and therefore it's an excellent area for a clinical trial and you might ask whether it was better to have a policy of early elective surgery or just monitor the size of the aneurysm there and see which was the best treatment method with respect to how many deaths occurred vascular surgeons can't do this by themselves they need a lot of skills to put together a clinical trial including a statistician very good to have an important statistician however much you may think that st-, statistics is very difficult once you're doing a big complex trial it's useful to look at more than mortality and we wanted to look at quality of life of two treatments the costs of two treatments and the cost-effectiveness of two treatments also allowed us to look at other things such as the risk of aneurysm rupture perhaps some biological things as genes predicting prognosis or the rate of aneurysm growth and it allowed us to do some real statistical work which the statisticians did on how you model aneurysm growth and depending how the time goes as to whether i'll talk about that later design is always best when it's simple for something like this and from our poll of vascular surgeons this was the grey area this was the area where they didn't know what to do so they could put their hands on their heart and say to the patient i don't know what to do i think you might have just as good a chance with early surgery or observation surveillance for growth and i'd like you to join us in a trial to see what's best for patients we had to try and work out before we started how many patients we would need and because you'd seen that on the group of people doing this the most populous group was vascular surgeons there was the automatic assumption that surgery would be best for these patients and i said it's quite difficult to know what the operative mortality is but they thought they'd go with the figures that suggested if you did an er operation electively two per cent of these patients would die within thirty days thereafter and because this is an older group of people there's a general mortality which they thought was six per cent per annum they thought if you just had your aneurysm watched you probably had a two-and-a- half per cent per annum chance of your aneurysm rupturing and most of you would die if your aneurysm ruptured plus the normal six per cent per annum and on that basis we thought that after five years the survival and the survival is here on the vertical axis would be seventy-one per cent in those that had an early operation and sixty-two per cent in those that were just monitored to see whether they grew bigger to arrange where surgeons were convinced you needed an operation i understand that this is something you haven't dealt yet with yet is power but this is something we needed to have any chance of showing this definitively we had to recruit about a thousand patients we had to be careful about how we randomized these patients we couldn't allow any fiddling we couldn't have a series of closed envelopes in each centre where we were going to recruit patients 'cause we had to do this through lots of hospitals and in fact we did it through ninety-three hospitals in Britain and so we used a central computer for randomization to avoid bias measurement trial and error and this is where we start talking about error and again this is a cartoon of using an ultrasound probe to measure both the maximum diameter and to see whether we could work out the ratio of the normal diameter here to the maximum diameter here and although it would be very nice to do this it turns out that here the er images are so fuzzy it's very difficult even for experienced people to measure and if you do it here you can measure the diameter either from front to back A-P anterior posterior or from side to side and because of the actual physics of ultrasound you can measure it more accurately from anterior to posterior but we wanted to design this trial to work through five regions in Britain and each one have a dedicated trial coordinator who are going to look after the patients and measure the size of their aneurysm so a question that came to our mind was could we be sure that different people would measure an aneurysm in the same way so we actually did a little experiment taking different people who knew how to use an ultrasound machine observers giving them an array of patients with different diameter aortas to see how well they did and we h-, the first study we did we had ten patients and then we increased the number of patients and we also looked and i'll tell you about that later on er looking measuring aneurysm diameter using a different modality computed tomography or C-T right here's the ultrasound the one you saw before you can measure this either in this dimension transverse or this dimension A-P using A- P and a large number of different patients we took just two single experienced ultrasonographers and we said measure these aneurysms for us and record them and this shows that data this is the difference in aneurysm diameter recorded between ultrasonographer A and ultrasonographer B against the mean aneurysm diameter the mean of those two observers you can see that the distribution of results is around the zero line but for instance in this particular case observer A measured the aneurysm diameter at point-four of a centimetre bigger than observer B whilst here they had measured it point-four of a centimetre smaller this type of display of data is known to me at least as a Bland Altman plot after two quite famous British statisticians Martin Bland and his partner Altman not partner domestic partner statistical partner [laughter] and it's a very useful way of displaying measurement data for error so we looked at this and we were appalled how could we hope to recruit patients from all over Britain measure aneurysms to make sure that they didn't grow bigger than five-point-five and have this done reliably how do you think we could improve on this sm0281: nf0279: one suggestion sm0282: use the same person to measure all your aortas nf0279: yes now if we're going to recruit people in Edinburgh and Plymouth sm0282: nf0279: they're going to be pretty busy aren't they and interestingly a good way to reduce error and it must be very obvious to you when i say it is training in fact we wanted to make this trial work using just five people and we found if we trained them and trained them and trained them you use this part of the line to place your cursor for your measurement you angle the probe this way so you make really sure you've got the maximum diameter we found that we could do much better and here are just ten patients and five coordinators the ones we used for the trial and you can see that you still get a spread of aneurysm diameter this is now the patient number and this is the aneurysm diameter measured by five different coordinators in different colours and you're still getting a spread of about point-three point-four millimetres but it's a lot better than the previous one with no training and no training together so i haven't shown you all of this but you need to measure the anterior posterior diameter you can't measure diameter that re-, reproducibly and this measurement error is really important in the way you design any work you do and what about computer tomography this is a C-T scan of an abdomen and this is the abdominal aorta here this red line is the transverse diameter this grey stuff round the side here is mainly laminated thrombus filling up the centre of the aorta and this is the outline it looks a little bit dark partly and white because partly because this is calcification in the aortic wall here we have a limited number of patients with known C-T aneurysm diameter we've taken our five observers we used in the trial and we've asked them to measure it by ultrasound using the anterior posterior diameter and this is what we come up with first of all you will see that this line lies above the zero so the difference between ultrasound and A-P A-P in the C-T diameter is usually positive I-E we appear to be measuring the aneurysm larger when we use ultrasound than we u-, when we use C-T scan in addition there is something we haven't seen before in that there appears to be a positive skew up here that the difference between ultrasound and C-T appears to increase with increasing aneurysm diameter so here we have some evidence that there's some magnification or relative magnification associated with ultrasound but we don't actually know which the gold standard is and which is better this is again is a kind of Bland Altman plot and i suggest to you is much more informative than a straightforward correlation line if we'd done a correlation line we'd have seen something lice nice like this oh fantastic agreement between the two different sorts of measurements and when you think about it you're measuring exactly the same thing you're measuring how wide the aorta is and just one you're using a C-T scanner and one you're using ultrasound so if it didn't agree very well well it wouldn't look very good would it and you ca-, not very informative at all so measurement repea-, random error reproduceability very very important but because ultrasound is safe it's cheap it's non-invasive and there are very few false positives or false negatives and it can detect ninety-nine per cent of aortas it's the chosen method for screening and surveillance of aortic aneurysms right so we did this trial we recruited patients w-, slightly more than the thousand we thought we needed over a four year period we followed them up to see how they did we flagged these patients with the Office of National Statistics and this is a very valuable resource that we have in Britain whereas you can get at the end of a research project the date of death and the cause of death of patients entered into this 'cause they're all recorded centrally in Southport and this was the principal results of the trial as reported in the Lancet in nineteen-ninety-eight on the vertical axis here we had the proportion of patients surviving this is the time in years and this is the number of patients at risk at each time interval this kind of display is known as a life table we can see here in yellow the patients undergoing surveillance of their small aneurysm as opposed to those with early surgery and what we'd predicted was that the blue line would h-, be up here at five years and the yellow line would be down here dow-, way down here at five years we didn't find that fairly f-, steady attrition in the surveillance arm surgery arm marked attrition early levelling off bearing in mind that out here having started with more than five-hundred patients in each arm of the trial not very many patients with so few patients the error around here is probably quite large but the statisticians the trial steering committee the journals and the public were convinced enough to s-, that this result suggested that there was no difference in survival if you operated early on this condi-, for this condition or whether you left it till later now very often one trial isn't en-, sorry we also looked at the health service cost 'cause i said we wanted to look at more things and this were the costs of the two treatments here early surgery and surveillance and you can break this down into how much the surveillance costs how much the hospital treatment for repair costs how much other health service costs are and the bottom line for this is that surgery is a much more expensive treatment option than surveillance and of course we live in a cash-strapped National Health Service so this is important it perhaps wouldn't have been so important if quality of life had been different in the two treatment arms but there was almost no different in quality of life between the two treatment arms so therefore with a similar survival similar quality of life but one treatment costing much more than the other what do you think the National Health Service wants surveillance and in fact vascular surgeons across Europe were quite happy with this result and for these reasons probably wouldn't would no longer recommend elective surgery for patients with aneurysms less than five-and-a-half centimetres but you see here that the operative mortality rate in this population based study was far more than we'd used in the power calculations we'd expected two per cent taking from best vascular surgeon series and in fact it was five-point-six per cent and when since we started the trial there were several other population based studies reporting the mortality for this operation turns out it for some reason it's lowest in Western Australia and i don't think it's just 'cause they have the most fantastic surgeons in the world in Perth or that it's the most fantastic place in the world to live i think very much probably most of these are very similar and the mortality associated with the elective repair of this condition is probably somewhere between five and six per cent it's not such a safe operation if you thi-, think of it another way probably somewhere between one in seventeen and one in eighteen patients that you operate on electively is going to die so we'd er managed to convince Britain and Europe that perhaps you shouldn't operate on these small aneurysms the Americans of course weren't convinced they don't have a National Health Service they have insurance based private care for most of their patients and they mounted a very similar trial started the same time as the British trial and it's always nice to beat the Americans 'cause they didn't report their results until two-thousand-and-two four years later but they found exactly the same thing er almost exactly the sa- , er s-, s-, same numbers of patients that in terms of mortality although this is plotted as survival no difference depending whether you operate early or you don't you just watch it but however this was four years later and by that time because of the Office of National Statistics we had four years more information on the patients that we'd had in the British trial and what i'd said to you before was that we didn't have very many numbers out here we didn't we only had about fifty-two and sixty-three by the time we get to two-thousand-and-two we've got large numbers of patients out here we've removed some of the noise in these curves and we can see that these curves appear to be starting to separate ah and the vascular surgeons shout with glee now we've got evidence that we can operate on these people 'cause they like operating but actually caution caution caution it doesn't really show that at all and you can look at this data another way you can integrate the areas under those curves and you can look at the average life expectancy of patients since their ori-, original enrolment in the trial and you can see that if you have early surgery that's your life expectancy if you have surveillance that's your life expectancy and there's absolutely no difference between the two of them again right out at the long time periods we don't have that many patients there's a large degree of error but there are some interesting reasons as to why those curves might diverge at a late time points and those reasons are principally attributable to the fact that the operation is quite a major trauma patients are in hospital for more than a week and the one thing it really f-, does force them to do is to stop smoking and most of these patients smoke or have smoked or will tell you they've stopped smoking and are still smoking a prolonged hospitalization of seven to ten days really enforces that and a lot of the late benefit we might see might be simply attributable to the fact that those who have their operation early stop smoking early so if you come to a consultation at the end of last year or even at the beginning of this year and what's the evidence we've now got two trials and i think the evidence from both trials is don't operate on these aneurysms when they're small the risks of your patient dying afterwards are too high recently we've seen reported a trial of screening 'cause obviously one of the important questions would be well 'cause we can detect these and five per cent of people over sixty have got them men do we need to do a screening programme for them a screening trial was reported recently and suggested might be cost-effective er but of course if a screening trial's really going to be cost-effective you need to find these things when they're small and you need to stop them growing how can we stop them growing if we don't know how to measure growth now there's a suggestion from retrospective studies that small aneurysms grow s-, more slowly than medium s-, size aneurysms and the largest aneurysms grow fastest of all we looked at growth rates of again seventeen-hundred patients this was their starting aneurysm diameter so that you can see that the main aneurysm diameter at the start is somewhere about four-and-a-half centimetres this was the nine longest series for aneurysm growth and the first thing you can see is just by looking at the shapes is that actually they're all different and actually one of them even gets smaller tells you we don't know very much about this condition some of them seem to fit nicely to exponential curves some to straight lines if you look in the literature the only thing that had ever been used before for modelling aneurysm growths and taking data points and working out how fast do they grow was linear regression modelling and i'm sure you'll have heard of that trying to draw a s-, the best straight line between your points and i've drawn three lines on here four lines and you can see how bad they are for most of the cases if you draw just straight lines there's another problem when you've got real people and you've got continuous measures of data and i'm going to talk about aneurysm diameter but it could relate to anything and that's to do with truncation of series what makes you stop looking at someone going to consider a conceptual patient here or a virtual patient who starts with an aneurysm diameter of four centimetres has that diameter measured every six months no change no change no worries suddenly eighteen months there's a high reading and they've gone to the five-point-five centimetre threshold where you might think about surgery the person's nice and fit still plays two rounds of golf a week or whatever and the surgeon says come in for an operation now perhaps if that patient had been coughing and spluttering the surgeon would have said i think we need to have a some sort of fitness programme before i think about operating on this come back and see me do this this and this we'll have some physiotherapy come back in so many months next time they come back the reading is ah just under four centimetres trial and error measurement error we can't measure it exactly ah not to worry now we'll keep on following you up and this is what happens and you can see that if you draw straight lines between these points you get quite slow aneurysm growth if you draw it between these you get quite high aneurysm growth you truncated your series here and you've got an artifically high growth rate using this approach of course that's not the only thing that happens to patients patients die let's take another patient four measures and suddenly they die from a myocardial infarction so you simulate their growth you get a nice line across here but if they hadn't died perhaps this would have happened and the growth rate would have been quite different so by not having this continuity that you can have in laboratory experiments where you can go and watch cells dividing for days on end people are different and there are other problems when it comes to trying to measure things and work things out statistician very clever no equations here use some flexible modelling i don't understand what any of this means and if you do you're all h-, absolute heroes and heroines but we can try and make some pragmatic rules using this modelling we want to know how to use data in clinical practice and we can therefore say that if you've got an aneurysm of four centimetres on average after five years seventy per cent of them will have reached the five-point-five centimetres threshold where you can consider surgery if you start at four-point sorry i showed you four-point-five centimetres if you start higher er bigger find aneurysms bigger you'll probably know that your chance of reaching the fi-, the threshold the surgery within five years is almost a hundred per cent for big aneurysms so it can tell you start to be able to give you information so that you can sit down with a patient with a five centimetre aneurysm and say well i suspect your aneurysm's going to keep on growing and the likelihood is that probably in about five years' time we might have to think of an operation but don't worry about it at the moment and there's another reason not to worry about it at the moment now i'm going to go quite quickly 'cause having said that i want to get to the end and i know that it's nearly lunchtime but just to reinforce let's show you what smoking does to aneurysm growth flexible modelling most of these patients with aneurysms have smoked this is the growth rate millimetres per year people who never smoked only twelve of those might as well forget about them ex-smokers some of these probably still smoking smokers don't like to tell you they can't stop smoking growth rate about two-point-four millini-, metres a year those who are actively smoking two-point-nine millimetres a year big difference so if you keep smoking you're going to grow faster you're going to come to surgery faster your lungs are going to be in worse shape and you're probably more likely to die if you'd used linear regression modelling this is what we'd have seen linear regression modelling exaggerates the effects of smoking and the possibility of giving up smoking in terms of helping your aneurysm growth and if we're going to be fair to patients we have to tell them this is what happens and not this and the reason this has arisen is probably because of truncation of patients right good reasons for your patients to wait technology is moving very very fast you no longer have to repair aneurysms using conventional open surgery and you can use endovascular repair you can insert a device through the femoral artery sometimes under local anaesthesia anchor it here with expandable stents probably don't anchor it here y-, most of the new devices come down into the iliac arteries and look a bit like this pair of trousers or like this and you can s-, insert these through quite small catheters through the femoral artery new technology how do we find out whether it works blind faith from the manufacturers Johnson and Johnson to the success again making more money i'm now going to charge you oops i'm now going to charge you five-thousand pounds for one of these whereas a little bit of Dacron tube i sold you previously was only fifty quid fantastic this is one of those er endovascular grafts being placed with a measuring device you can see all the little buttons going up there in the aorta to try the thoracoscopy when you place one and actually in population terms the health of the population we've got to find out whether the new technology is any good so we're starting all over at the beginning again although it's now been ongoing for two or three years with a new trial this time surgeons don't know whether it's better to treat these larger aneurysms with either the conventional open operation which everybody's been trained to do or with the new technology and this is being evaluated again we have to be aware of trial and error and measurement and we're very reliant on our statisticians to keep us in order give us the correct trial design and analyse the results correctly and even though i haven't referred to you any equations statisticians are absolutely vital for getting it right and getting the right information for the patients thank you for your attention