1
00:00:00,300 --> 00:00:04,470
Thank you very much. Can everyone hear me okay? Yeah, fine.

2
00:00:05,150 --> 00:00:14,879
Um, so I hadn't really thought about the western part of it that occurred to me on Thursday as a well-respected Western.

3
00:00:14,880 --> 00:00:18,930
I don't know if you're into Westerns, but I will seek to cover the good,

4
00:00:18,930 --> 00:00:25,830
the bad and the ugly as we progressed thinking about life as a trial statistician.

5
00:00:26,470 --> 00:00:31,920
Um, this is why I wanted to talk about. I'm going to give you a little bit of background about me,

6
00:00:32,010 --> 00:00:41,310
so maybe help explain some things I'm going to talk about or moan about and complain about and also where I'm coming from.

7
00:00:41,790 --> 00:00:45,239
Other trials too, would that a different, slightly different set.

8
00:00:45,240 --> 00:00:48,149
But I think there's enough in common they'll be interesting to everyone

9
00:00:48,150 --> 00:00:53,880
hopefully here and I don't want to finish on ugly the problem with the bar that

10
00:00:54,210 --> 00:01:00,480
some of the fellow that volunteered to write personal law to see if I can put

11
00:01:00,480 --> 00:01:06,350
it that way I start they'll might vaguely remember us as an undergraduate.

12
00:01:06,420 --> 00:01:17,460
Obviously I did a student project related to clinical trials, randomised controlled trials, which is prior to 1999.

13
00:01:18,600 --> 00:01:24,329
Um, and from that there was an opportunity to do a Ph.D. which was really a topic,

14
00:01:24,330 --> 00:01:29,550
but a colleague of Jill's at the time and then later a colleague of mine that I stayed on.

15
00:01:30,220 --> 00:01:34,620
Um, so my interest has gone through all my professional career.

16
00:01:35,250 --> 00:01:39,459
It's gone from very early on in my studies into clinical trials.

17
00:01:39,460 --> 00:01:50,880
It's mainly randomised controlled trials working up in Aberdeen in the Health Services Research Unit and then for a year I spent in Canada.

18
00:01:50,880 --> 00:02:00,630
So I got to see of a different environment and at about five years ago I moved down here to Oxford for the Centre for Statistics and Medicine,

19
00:02:00,930 --> 00:02:09,089
and happily enough people I was already working with from Aberdeen put in a department and there was an opportunity to move into that

20
00:02:09,090 --> 00:02:20,070
department which CSM so quite complex get story basically to two sets of people I worked for kind of moved to make a nice set up for me,

21
00:02:20,080 --> 00:02:28,400
a way to transition in quite a gentle way after my big break from obviously in after 15 years or so.

22
00:02:29,160 --> 00:02:32,330
I have been involved in all trials for over 15 years.

23
00:02:32,340 --> 00:02:35,080
But I know what you're thinking. I'm not sure.

24
00:02:35,700 --> 00:02:46,260
But the young fellow and I'm very vocal about 20 trials of various types, I think from very small one month bond sort of trials,

25
00:02:46,530 --> 00:02:54,089
if I can put it out in the old traditional sense as, as, as it was usually one man who did everything by himself.

26
00:02:54,090 --> 00:02:59,909
Clinician But then from about a bunch of that was pretty much it was slightly better than that but had my start,

27
00:02:59,910 --> 00:03:10,020
but it was only slightly better overall. But also they were moving into the much more diverse and better on high quality, large, expensive trials,

28
00:03:10,620 --> 00:03:18,839
sometimes too expensive maybe trials that we can get nowadays and from a variety of different funders and different types of treatment.

29
00:03:18,840 --> 00:03:27,510
So I mainly have work looking at non-drug treatments and specifically those that involve surgery, perhaps double the burden.

30
00:03:28,050 --> 00:03:34,050
And what people might be familiar with know the regulation of language, of systems, um,

31
00:03:34,230 --> 00:03:45,120
physiotherapy and other types of things which you didn't even know was something you could put on a trial and told from the beginning to time.

32
00:03:45,120 --> 00:03:53,190
I had an interest the there was lucky to work in a group that had an interest not just in doing trials but having to do trials on.

33
00:03:53,350 --> 00:04:04,440
And I have my PhD in topical these things kind of are there with you think your fear of them come back to haunt you some time for crossword puzzle

34
00:04:04,440 --> 00:04:13,989
so she got hopefully some sort of reputation for I also work in some of our generic sort of trial issues but very much with applied science.

35
00:04:13,990 --> 00:04:20,850
So you will know I can assure you see any form in this presentation might be a relief for this time on a monday.

36
00:04:22,530 --> 00:04:25,890
So there you go. I know I've been involved in lots of teams.

37
00:04:26,240 --> 00:04:29,560
Have you been introduced to that terminology or something you'd be familiar to?

38
00:04:29,640 --> 00:04:38,969
70 years or so they are monitoring phase and PSC or trial steering committees or the equivalent of management committees,

39
00:04:38,970 --> 00:04:43,560
etc. I've been involved in both of them bundling that only with the trials for research.

40
00:04:43,560 --> 00:04:48,450
So it's a fairly broad experience of trials,

41
00:04:48,450 --> 00:04:55,559
but very much on the academic side and then involved in the various different roles from both sides of the fence.

42
00:04:55,560 --> 00:04:59,850
So then more recently I've been managing some trial statisticians, so I.

43
00:05:00,360 --> 00:05:06,880
Do the hard work, really undermine other people I work with, do the hard work, and I think we try and help them occasionally.

44
00:05:08,960 --> 00:05:09,270
Okay.

45
00:05:10,110 --> 00:05:18,290
I'm happy to take any comments or queries, by the way, if things go wrong, but hopefully most things will change for some food for a better place.

46
00:05:18,450 --> 00:05:23,419
So. So what's so cute about clinical trials where you probably had a bit of a story,

47
00:05:23,420 --> 00:05:27,590
but as a statistician, there's probably a few ways you can think about.

48
00:05:28,630 --> 00:05:35,030
If I was to give you one phrase of analysis, it gives you a safe and great environment to do statistics.

49
00:05:37,580 --> 00:05:44,239
And one thing that you should never take for granted is that this is a design study.

50
00:05:44,240 --> 00:05:50,420
By definition, a clinical trial is some sort of experiment or study looking at the impact of the treatment.

51
00:05:50,420 --> 00:05:53,330
Therefore, that implies someone thought about it before that happened.

52
00:05:54,260 --> 00:06:00,090
Let me reassure you, unless you're in any doubt, people do come to you and they haven't thought about what they're doing.

53
00:06:00,140 --> 00:06:07,160
Just happen to think you might be able to explain to them why they're bothering, quite frankly, and why they do move over to you in the first place.

54
00:06:08,330 --> 00:06:12,200
Whereas in trials you're nice, safe in the sense as designed,

55
00:06:12,200 --> 00:06:20,000
and therefore also the data you're going to get is prospective and not just say so much horrible message from you.

56
00:06:20,780 --> 00:06:25,999
I'm generally speaking of funded studies as well to various degrees,

57
00:06:26,000 --> 00:06:35,720
but that also ensures a certain degree of thoughtful rigour and sort of due process oversight studies or regulation.

58
00:06:36,410 --> 00:06:39,050
I'll come back to that kind of regulation,

59
00:06:39,680 --> 00:06:46,460
but that is a very positive thing because it ensures data quality in a way that isn't true in many other areas.

60
00:06:47,270 --> 00:06:56,120
We do get very high quality studies or not clinical trials, but you get very many studies that are not high quality for clinical trials.

61
00:06:56,780 --> 00:07:04,970
And so this process of regulation and understanding what we're doing as a serious business leads to death.

62
00:07:05,030 --> 00:07:08,719
I would say, generally speaking of a better quality in a review process.

63
00:07:08,720 --> 00:07:11,380
And then lastly, I'd say it's for a while.

64
00:07:11,390 --> 00:07:19,250
We are doing something that's useful and adds to our understanding, understanding in a very altruistic generic sense.

65
00:07:19,250 --> 00:07:24,889
We're trying to gain insight so that we can help improve patient's health and we can make

66
00:07:24,890 --> 00:07:28,990
judgements about what treatment people should or shouldn't get and who should be treated.

67
00:07:30,620 --> 00:07:40,770
So we're addressing something important, but our the analysis are not very important and then it's very well,

68
00:07:41,180 --> 00:07:48,920
there are phases that people use in clinical trials, phase one, phase four, which works well for terms of drug treatment.

69
00:07:49,280 --> 00:07:53,840
But really there are very, very few actually even on the same page.

70
00:07:53,870 --> 00:08:01,160
This is a slide by some countries. Consider the phase of clinical trials which probably no one will agree with exactly with all things.

71
00:08:02,270 --> 00:08:11,530
But what we do have is is a real you can see just very easily the aim of what you're trying to do is on trial,

72
00:08:11,540 --> 00:08:16,519
trials vary quite a lot from very exploratory to efficacy.

73
00:08:16,520 --> 00:08:21,759
So called in part of the treatment post-marketing surveillance.

74
00:08:21,760 --> 00:08:28,159
There's language that might be used in an industry setting, but the size is very, very small studies.

75
00:08:28,160 --> 00:08:33,150
Two very large studies of population can vary. The design is building on that.

76
00:08:33,290 --> 00:08:40,880
And whether we even have basic things like the control group and the outcome and how we're going to assess the results of the study.

77
00:08:41,720 --> 00:08:48,830
And I have mostly focussed on what might be called phase three ones here and even rules are very,

78
00:08:48,830 --> 00:08:53,720
very mixed, but by far the same thing about being involved.

79
00:08:54,320 --> 00:09:00,379
And so I just want to emphasise a little bit about our systems because we need to remind

80
00:09:00,380 --> 00:09:05,060
ourselves that they are a really good thing to do because this is what most of us,

81
00:09:05,060 --> 00:09:12,260
the strong statisticians, spend most of our time to do it and you can view this in many different ways.

82
00:09:12,260 --> 00:09:16,879
You've probably had some of those today, so I'll just walk through very quickly.

83
00:09:16,880 --> 00:09:24,460
But you can describe a very simple language. You address the scenario of a depressed treatment or not treatment or treatment for people.

84
00:09:24,470 --> 00:09:27,580
The scripts are expected to be similar.

85
00:09:28,700 --> 00:09:31,999
It allows you to isolate the factor of interest. That is the treatment.

86
00:09:32,000 --> 00:09:36,470
Usually it prevents intentional, intentional bias.

87
00:09:37,760 --> 00:09:44,090
It provides clarity on inclusion and section. Point of participation is not something to take for granted.

88
00:09:44,780 --> 00:09:50,060
You do get into discussions about who is in my study and who should be in my study

89
00:09:51,110 --> 00:09:55,429
and all those things you do if you want to think of it more as a statistician.

90
00:09:55,430 --> 00:10:00,590
But I'm actually a different way that we can think about that or describe it.

91
00:10:00,680 --> 00:10:05,650
In fact, that's one of the, I think the exciting and good things about being trial sessions.

92
00:10:05,660 --> 00:10:08,300
Trials are fairly complex. They're also very simple.

93
00:10:08,760 --> 00:10:20,550
So the concept very simply is tossing coin or Ian Chalmers used to use and talk about children when they separate into two teams.

94
00:10:20,910 --> 00:10:24,510
Tick tock. Talk about this very simple concept.

95
00:10:24,510 --> 00:10:28,920
You're trying to roughly randomly create two teams. Two groups are going to be several.

96
00:10:29,310 --> 00:10:34,220
In that sense, you're trying to play their game here. You're trying to say the only difference is a treatment.

97
00:10:34,260 --> 00:10:42,900
Therefore, I told the team in March, that is true. But it's also true that we can think about in a more mathematical, complex sense.

98
00:10:43,470 --> 00:10:46,530
I'm the one where the word you'll often see is exchangeable.

99
00:10:46,620 --> 00:10:51,880
It just means basically that everything is the same except for the treatment.

100
00:10:51,900 --> 00:10:55,950
And you can write out formulas and justify it in that basis.

101
00:10:56,700 --> 00:11:02,879
There is a whole framework of statistics, classical statistics, which you will have been introduced,

102
00:11:02,880 --> 00:11:06,360
I would think, over in this question billion ones to some degree.

103
00:11:06,750 --> 00:11:12,540
But a really cool thing about randomisation is it justifies an analysis with no program assumptions.

104
00:11:13,500 --> 00:11:16,409
It's not necessarily what we do there today as a statistician,

105
00:11:16,410 --> 00:11:26,130
but actually there is a theoretical basis for doing an analysis for creating things like a feasibility that doesn't rely on anything else.

106
00:11:28,250 --> 00:11:37,220
And then there's also a very different way that is growing actually more recently in the literature, which is a causal way to think about it.

107
00:11:38,030 --> 00:11:42,050
And you can express things in terms of the observed or the counterfactual.

108
00:11:43,010 --> 00:11:48,180
But if you think about clouds, they are a little bit counterfactual in the sense that you create two groups.

109
00:11:48,620 --> 00:11:54,230
We can never really assess what we want to know, which is if you have a patient in front of you and you're a doctor,

110
00:11:54,860 --> 00:12:01,309
you you want to really know what would be the result they would have with the two treatments.

111
00:12:01,310 --> 00:12:06,550
So you get treatment aid, but you kind of go back in time and then say, okay, let's, let's give them treatment.

112
00:12:06,560 --> 00:12:08,000
And that's the kind of factual.

113
00:12:08,540 --> 00:12:16,520
But you can express trials, not when you can actually take that into another direction in terms of mathematics and thinking.

114
00:12:16,700 --> 00:12:19,969
So that's another way you can do it.

115
00:12:19,970 --> 00:12:25,170
And then lastly, have Bayesian statistics, which probably not touched on too much,

116
00:12:25,520 --> 00:12:31,069
but that's another way that you can think of randomised controlled trials and about you might think

117
00:12:31,070 --> 00:12:38,180
of trials in terms of refining your understanding or updating your prior by getting good data.

118
00:12:40,910 --> 00:12:50,840
Okay. So if we take that for granted, what are the other values or what the good things about being a trial starts to show?

119
00:12:52,970 --> 00:12:56,100
Probably the best place to start is high impact commodities.

120
00:12:56,420 --> 00:13:01,520
That's my heart. And you will see the language gold standard.

121
00:13:01,520 --> 00:13:08,390
You've heard of that that properties are the gold standard way to prepare people for something like this is number four.

122
00:13:08,780 --> 00:13:15,769
Clinical trials are the most definitive. To provide radiation therapy for three years is very 5 to 8,

123
00:13:15,770 --> 00:13:23,809
which is very nice ultimately in many regards when it comes to treatment about and comes to research about

124
00:13:23,810 --> 00:13:29,900
treatments or cities or clinical trials more generally are the standard which other research is held to account.

125
00:13:30,680 --> 00:13:35,089
And they are the most researched area. And roughly speaking,

126
00:13:35,090 --> 00:13:42,360
the methods in the process are most of the working of that is that they're

127
00:13:42,380 --> 00:13:46,730
generally the most well respected and they have the highest potential impact.

128
00:13:47,960 --> 00:13:51,260
And that impact can be in two ways. It can be a direct impact.

129
00:13:52,400 --> 00:14:00,440
If you work as a trial statistician, your work on a trial, your trial might theoretically change things straight away as soon as it's published.

130
00:14:01,610 --> 00:14:05,059
That's very exciting, but certainly is not true of most trials.

131
00:14:05,060 --> 00:14:08,930
So I'd be lying if I stood up here and told you that every trials will do that.

132
00:14:09,770 --> 00:14:11,310
Everything's going to change straight away.

133
00:14:11,930 --> 00:14:18,440
Nevertheless, it's also true that some of them do have a dramatic impact from this, a potential there for your study.

134
00:14:18,440 --> 00:14:26,329
So it is also true that the impact of your study might be more subtle or indirect, but outcome come over time.

135
00:14:26,330 --> 00:14:35,239
You are contributing a trial result which might be then taken with a body of evidence that might eventually end up in a systematic review.

136
00:14:35,240 --> 00:14:42,680
You would hope for maybe a better analysis, and through that you're contributing again to the practice and the updating of practice

137
00:14:42,680 --> 00:14:48,260
and the understanding that irrespective of what the result is in many regards,

138
00:14:49,640 --> 00:15:00,709
this is a sample of a paper that was by a colleague of mine in the department, and he was involved in a study called Draft.

139
00:15:00,710 --> 00:15:04,640
I don't expect you to be able to read the book, but the title you probably can read,

140
00:15:04,640 --> 00:15:09,080
which says Do large pragmatic randomised trials change clinical practice?

141
00:15:10,610 --> 00:15:20,540
What is interest? This is a graphical issue. What they show here is the use of different treatments for the condition or contrast.

142
00:15:22,640 --> 00:15:30,650
This is orthopaedics, this is trauma. And they've got wires, which is a blue line plate, which is a red light, and then they've got other.

143
00:15:30,980 --> 00:15:40,190
So the offer is not what they were evaluating and the format stayed pretty constant over this period of ten years, which is shown on the x axis.

144
00:15:40,790 --> 00:15:46,330
But what happens is there's a quite dramatic change, or particularly an uptick in the Wire,

145
00:15:46,580 --> 00:15:52,100
which was one of the two interventions is trial play, it being the other.

146
00:15:52,580 --> 00:15:59,420
And because this is a very large trial and then influenced involving the large

147
00:15:59,420 --> 00:16:03,980
teaching hospital by purely doing the trials seem to actually influence the practice.

148
00:16:04,190 --> 00:16:11,810
People become more aware of the alternatives, we're told, which is not always true in surgery about the options are available.

149
00:16:12,380 --> 00:16:20,959
I'm actually started using and then when the results of the trial actually came out about 10% even more so the trial very quickly,

150
00:16:20,960 --> 00:16:24,050
even before it was actually completed, seems to have influenced practice.

151
00:16:24,050 --> 00:16:32,630
And then when the results came out to the issue of the use of wire, this has continued in this quite dramatic increase.

152
00:16:33,620 --> 00:16:42,440
This is another one. Again, forgive me for using Peter examples because it's where I work most familiar with of recent.

153
00:16:42,770 --> 00:16:52,759
There's another study which again in this trial is called for for all I want to highlight here is a conclusion which may or may not be able to read,

154
00:16:52,760 --> 00:17:03,950
but it says proper has had an impact on the surgeons clinical practice both for changing that and underpinning existing non profit practice.

155
00:17:04,550 --> 00:17:14,000
So trials, aside from the correct results, they also can be an embodiment of good practice in many clinical areas.

156
00:17:14,000 --> 00:17:20,930
Surgery is certainly one and here they're highlighting that although not everyone accepted immediately the result of the trial,

157
00:17:21,200 --> 00:17:27,970
everyone knew about it pretty much showed, a lot did, and many recognised that the practice was being applied.

158
00:17:28,220 --> 00:17:32,180
So it's just another example of how the trials can have a big impact.

159
00:17:32,510 --> 00:17:40,030
This for me personally is my most recent example, and if you get a paper in The Lancet, you like to tell people it doesn't fit in with The Lancet.

160
00:17:40,740 --> 00:17:49,979
Hence why I have to say that this just came out just last month, but it's got some uptake in some newspapers and whatever.

161
00:17:49,980 --> 00:17:56,730
So it will be interesting to see, but the word is that it's actually starting to influence people's focus.

162
00:17:56,730 --> 00:18:04,350
The problem here. This is about shoulder impingement, which was a growing and reasonably common shoulder operation.

163
00:18:04,530 --> 00:18:09,780
We did what you might call a placebo form of the shoulder operation.

164
00:18:10,080 --> 00:18:16,650
You can ask me afterwards about it, but you're interested that we didn't show that there was much benefit for the real surgery.

165
00:18:17,010 --> 00:18:18,330
So it's quite controversial.

166
00:18:19,240 --> 00:18:28,890
And if you're into ultra metrics, it's got something like 1300 results for medical school, which is pretty good, given that most people have a ten.

167
00:18:32,770 --> 00:18:33,970
Okay. Later on,

168
00:18:34,900 --> 00:18:45,670
I think something that's hidden because a lot of statisticians who are not really areas on trials is the variety of work and issues that are raised.

169
00:18:47,500 --> 00:18:51,670
And I think that, you know, almost any type of analysis, at least to some degree,

170
00:18:53,920 --> 00:18:59,680
we can argue any kind of very simple things like Fisher's exact facts that,

171
00:18:59,710 --> 00:19:10,210
you know, the names or T test and quite robustly and confident of the use of some aspirin and probably some clinical trials in randomised trials.

172
00:19:10,630 --> 00:19:11,690
Having said that though,

173
00:19:11,750 --> 00:19:21,190
there's such a wide variety of issues and concepts that are addressed in statistical literature that's common to some of the trials.

174
00:19:21,190 --> 00:19:23,820
So issues such as clustering,

175
00:19:23,830 --> 00:19:33,460
if people are familiar that the idea that some observations are more similar as a group to each other than they are to every other observation.

176
00:19:34,300 --> 00:19:39,160
And so you get a variety of statistical methods like multilevel modelling and GS.

177
00:19:39,160 --> 00:19:44,920
These are things that people familiar for other areas and are options.

178
00:19:44,920 --> 00:19:46,690
You have the longitudinal aspect.

179
00:19:46,690 --> 00:19:54,579
Of course, most cases you have different types of outcomes, quite different types of programs which bring up different types of analysis.

180
00:19:54,580 --> 00:20:01,390
Just for any regression you can think of, you can find a way to analyse in a trial somewhere.

181
00:20:01,810 --> 00:20:06,570
And then you also have more recently a lot of the causal work which I alluded to earlier,

182
00:20:06,590 --> 00:20:11,169
this whole dream of growing literature online, which interests enough.

183
00:20:11,170 --> 00:20:24,190
A lot of it has come from economics and the interest of doing looking at the impact of policy in health economics

184
00:20:24,190 --> 00:20:32,680
and trying to isolate factors that actually inspire the use and analysing clinical trials and increasingly so.

185
00:20:33,790 --> 00:20:36,099
So don't be misled by thinking.

186
00:20:36,100 --> 00:20:42,700
It's all just one very simple thing when you do the same type of analysis of the research and past array of different things.

187
00:20:44,710 --> 00:20:50,440
And the core analysis, as I say, can be straightforward, but there's many interesting lawyers.

188
00:20:52,750 --> 00:21:00,500
Our next topic is, I think it's generally accepted that you're a valued part of a team, the level of study,

189
00:21:01,690 --> 00:21:07,590
your value for your expertise, even if sometimes about corruption like people need to your office,

190
00:21:07,960 --> 00:21:12,640
for example, let's say speak to a statistician looking for the final flight and then 6 hours later,

191
00:21:12,640 --> 00:21:16,690
at least to get sample size, they might think vaguely.

192
00:21:16,690 --> 00:21:22,269
You should do something more than that. But more often than not, increasingly is my experience.

193
00:21:22,270 --> 00:21:25,719
Maybe I've been fortunate is that people do recognise your expertise,

194
00:21:25,720 --> 00:21:32,170
they do recognise the value and to recognise more generally as a trial statistician,

195
00:21:32,170 --> 00:21:36,630
you understand things about the study, about how to do what's the right way to do it.

196
00:21:36,760 --> 00:21:40,970
How could you improve a study? What's the right way to go about it?

197
00:21:40,990 --> 00:21:48,460
Sometimes from the more getting rid of them than the number actually coming, then actually the problem will continue.

198
00:21:48,870 --> 00:21:55,300
And I think also that's been my experience that you get to work with some excellent scientists.

199
00:21:55,570 --> 00:21:59,740
The whole process can be quite some things.

200
00:21:59,770 --> 00:22:03,370
And so for the clinical trial, it's a long drawn process.

201
00:22:03,370 --> 00:22:09,350
But one positive is that people who tend to get there at the end often are fairly normal.

202
00:22:09,490 --> 00:22:21,819
This will help to converse, but often very good for experienced researchers and people who have good scientific understand.

203
00:22:21,820 --> 00:22:28,940
So they're quite voluble people to be around and just spend and be part of the team when you will work.

204
00:22:28,940 --> 00:22:32,230
For lots of professional people, particularly when you're in the law study,

205
00:22:34,300 --> 00:22:41,320
there will be professional managers of the project, will get the professional data managers as well to have there,

206
00:22:41,320 --> 00:22:48,549
although will be programmers who specialise often in doing files and not expertise,

207
00:22:48,550 --> 00:22:54,620
is a very exciting thing to be part of and they're often along with nice people, I think.

208
00:22:56,290 --> 00:22:59,409
And then what is fun is that you're there to take over.

209
00:22:59,410 --> 00:23:06,040
I would say a unique trial, there is a sense of trials, is a robust methodology, and that's true.

210
00:23:06,040 --> 00:23:15,159
But the application of it every time I would say is unique in some ways that the trials of little life to those trials and events happen,

211
00:23:15,160 --> 00:23:23,120
whether it's personnel related or otherwise. And the makes it quite an interesting thing to be part of an unpredictable what's going to happen.

212
00:23:23,630 --> 00:23:28,150
So when you're working employing together to get to the finish line.

213
00:23:28,270 --> 00:23:30,610
And so that's really enjoyable.

214
00:23:34,860 --> 00:23:41,610
One of the special things having a baby in a house station is that as well as being part of the team, you have a very unique perspective.

215
00:23:43,470 --> 00:23:50,760
You're one of the few, maybe the only people sometimes who other than maybe the C.I. who's involved from the very beginning to the very end.

216
00:23:50,760 --> 00:23:57,999
But we start possibilities there for you from the very kernel of an idea or working group, the PICO type idea,

217
00:23:58,000 --> 00:24:01,630
or all the way to thinking about funding and the grant application and getting

218
00:24:01,670 --> 00:24:05,489
that and then stirring up a study and then running out and all the rest.

219
00:24:05,490 --> 00:24:07,319
And then to the analysis and the reporting.

220
00:24:07,320 --> 00:24:16,610
And also when it's running as a statistician, you get through the data, you get to understand how things all connect up.

221
00:24:16,620 --> 00:24:22,050
And I find that not everyone who's involved in travel really has that perspective.

222
00:24:22,980 --> 00:24:27,180
So it's really quite a privilege, I think, in some ways to do that.

223
00:24:27,180 --> 00:24:32,960
And it allows you to improve and to all think you have something useful to say about data collection and how it forms and complete.

224
00:24:32,970 --> 00:24:34,470
You have something useful to say,

225
00:24:34,920 --> 00:24:42,840
but databases and how they're set up and how the data comes out and how it's processed and about data quality and all these other things.

226
00:24:45,410 --> 00:24:53,910
But um, here is a picture of a team and I still enjoy being part of it, but it's not because there've been some dramatic, glamorous things.

227
00:24:53,910 --> 00:24:56,219
It's not ended up in The Lancet by any means.

228
00:24:56,220 --> 00:25:02,940
But what was very fun was it was people who wanted to do a trial that were new to it, but there was a keen interest.

229
00:25:02,940 --> 00:25:05,969
And so I put off also being part of the picture.

230
00:25:05,970 --> 00:25:08,880
I think this was tweeted by the Journal. You wouldn't believe it,

231
00:25:11,790 --> 00:25:20,939
but I personally have enjoyed working with that group and it was a new area and they were just very interested in keen to do what's nice to say also.

232
00:25:20,940 --> 00:25:26,670
So we did a pilot study and now we've got funding from the property company, so we're continuing to work together.

233
00:25:29,100 --> 00:25:37,489
Another nice thing as a special thing about being across cloud services is you get to know the answer for

234
00:25:37,490 --> 00:25:42,750
a number of people first to know they are monitored and I'll come back to them on the negative side.

235
00:25:42,750 --> 00:25:48,570
On the positive side, you get to know the answer. Everybody wants to know it's that's great.

236
00:25:49,260 --> 00:25:51,260
And then the final answer as well,

237
00:25:51,270 --> 00:25:57,839
that's the most special thing because the results in between can change and you don't get always to see the full picture,

238
00:25:57,840 --> 00:26:05,220
but you will be the first person who sits down with the results and you will know for anyone else really which way this thing is going to go.

239
00:26:06,060 --> 00:26:11,790
So we'll see. So we knew we and the other of this, we knew that this was going to be quite controversial.

240
00:26:12,330 --> 00:26:16,640
That's quite exciting. I quite like this quote.

241
00:26:16,690 --> 00:26:23,999
So I learned that some things I think it is a slightly funny position as a statistician that you have this information,

242
00:26:24,000 --> 00:26:34,790
but you've got to not reveal it. But it is quite an exercise like well, okay, but scale of work I three see.

243
00:26:34,800 --> 00:26:39,950
So if I'm going to say the negative only because it's very familiar with me on the scale.

244
00:26:40,440 --> 00:26:45,570
So soft spots are not bad, but it's quite a long work to produce a good fair one.

245
00:26:45,900 --> 00:26:50,129
This is what we had for so others will produce much longer document.

246
00:26:50,130 --> 00:26:55,890
There's a lot of work involved, lots of analysis.

247
00:26:56,930 --> 00:27:00,750
I tend to think of primary outcomes or maybe a few of our outcomes.

248
00:27:02,640 --> 00:27:08,250
We have nine analysis that's of only the primary in the main paper.

249
00:27:08,280 --> 00:27:14,550
We also have supplementary paper at 18 and also some primary outcome 20 year of secondary outcomes.

250
00:27:15,420 --> 00:27:18,420
So that's 46 analysis.

251
00:27:22,230 --> 00:27:26,280
Now one of the things we're publishing, which is very interesting,

252
00:27:26,290 --> 00:27:31,769
is you keep producing the same thing in different formats and every time it gets revised you have to double check.

253
00:27:31,770 --> 00:27:41,520
And every time someone's touched it, it's changed. And so I consider recently those 618 numbers in the press that we needed to check.

254
00:27:42,300 --> 00:27:48,900
Now, you would think that if you checked the ones in the proof, you might think that you can just rely on those numbers being right.

255
00:27:48,930 --> 00:27:52,810
You'd be wrong. So it doesn't matter how good a journal is, you will find.

256
00:27:52,830 --> 00:27:57,989
In fact, we found two numbers which we hadn't asked for any changes and there are other ones

257
00:27:57,990 --> 00:28:00,860
where we had asked for changes and they haven't changed that the way we asked them.

258
00:28:01,200 --> 00:28:07,559
So we found two numbers that were correct and the original proof that were then wrong.

259
00:28:07,560 --> 00:28:11,220
And so you have to do that checking every time.

260
00:28:11,610 --> 00:28:17,610
That's in addition to all the checks that we did on the multiple versions prior and multiple collaborators.

261
00:28:18,330 --> 00:28:26,200
So it just gives you a feel for the scale of work at the end, which then you if you think through back the way to vote, how much?

262
00:28:26,200 --> 00:28:32,550
So with the analysis, it's a lot of time and a lot of effort, mortality, whether some people seem to go away with what?

263
00:28:32,790 --> 00:28:42,130
Come as they're kind of all study but certainly not to clinical trials is a long process.

264
00:28:42,160 --> 00:28:50,469
My first involvement in the study was November 2010 to get involved in the drama published November seven,

265
00:28:50,470 --> 00:28:57,190
about seven years from initial search, which was 2016 to publication 16 months.

266
00:28:57,850 --> 00:29:03,430
So I actually when I knew the results, we knew the results of the team that 16 more months to get it published.

267
00:29:04,240 --> 00:29:09,220
I thought it was crazy business and everything, but that was it was only two journals,

268
00:29:11,380 --> 00:29:19,210
one journal to face and plan about a year finally to side with the back and forth with about 40 page briefs,

269
00:29:19,300 --> 00:29:25,300
submission and response to reviewer comments before they decided they were going to reject it.

270
00:29:28,360 --> 00:29:36,800
So even when you think you're tired, we're not actually done. Which brings me to my next point, which is increasing demands sops.

271
00:29:36,910 --> 00:29:41,890
I put those as necessarily legal. I guess they are there.

272
00:29:42,190 --> 00:29:47,760
So who knows how many sops and I don't even know the results that they have.

273
00:29:47,770 --> 00:29:52,270
Nobody ever knows how many. So. So it's always changing and the SOPs are always changing.

274
00:29:54,340 --> 00:30:00,909
I am not most fond of sports, but they do and are necessary and increasingly necessary.

275
00:30:00,910 --> 00:30:03,120
And part of what clinical trials like.

276
00:30:03,130 --> 00:30:09,790
But particularly if you work on drug trials and it is something as a trial statistics you have to accept, I'm afraid.

277
00:30:10,240 --> 00:30:16,730
But as part of your life on the frontline, you are part of your life very little.

278
00:30:16,870 --> 00:30:18,489
And they very much symbolise.

279
00:30:18,490 --> 00:30:26,410
It's not really fair to say this, but it's also very much symbolise the increasing scrutiny and expectations that have come about for good reasons,

280
00:30:26,710 --> 00:30:35,290
mostly about trials and about safe conduct, about people's data, and about handling things appropriately and robustness.

281
00:30:36,250 --> 00:30:41,020
And if you have consults and photo ops, they will be very useful.

282
00:30:41,020 --> 00:30:44,370
When you need to defend yourself and defend your study more,

283
00:30:44,380 --> 00:30:57,370
you've got to put personal expectations on data alongside that and there's been a much greater emphasis on it.

284
00:30:58,390 --> 00:31:00,640
But when the scale, when you put it together scale,

285
00:31:00,880 --> 00:31:08,500
it's very demanding about crafts in the process and as a statistician to sign off and check and make sure and it's true what I said before,

286
00:31:08,830 --> 00:31:17,880
that you are you are the person who can look at all these things, but it falls on you and you're the person has to make sure it's done right.

287
00:31:17,890 --> 00:31:25,720
And you have a thousand data items in the database to double check, and you have to check that they've all got the right courtroom as well.

288
00:31:25,900 --> 00:31:32,170
It's totally new. So and that's, I would say, increased in the time I've been involved.

289
00:31:32,530 --> 00:31:40,149
And then also more so about how it's getting stored. And this is an area that's still developing and will be increasing for the next five,

290
00:31:40,150 --> 00:31:45,430
ten years about where data ends up and about the process and want to make it available to others.

291
00:31:45,430 --> 00:31:48,070
So even when you finished, you haven't finished.

292
00:31:48,800 --> 00:31:55,540
We've still got other responsibilities that will go on as a trial statistician beyond what you think of as your involvement at full time.

293
00:31:58,150 --> 00:32:04,450
And then on top of that, there's some more documents. Again, for good reasons, all have the right place,

294
00:32:04,450 --> 00:32:09,910
but it's just more things that you're involved and more things you have to make are consistent and standard.

295
00:32:11,650 --> 00:32:18,100
And so it does create work, good as they are. Another bad side.

296
00:32:18,110 --> 00:32:22,780
This is what someone told me that recently I was an independent statistician.

297
00:32:22,780 --> 00:32:26,440
And you see, this is all the old first place you stop in the middle of a war.

298
00:32:27,460 --> 00:32:29,920
That is a statisticians. But in general.

299
00:32:29,930 --> 00:32:37,329
But it's very much true as a trial statistician don't usually get the the lead the first off the ground with the

300
00:32:37,330 --> 00:32:44,130
one you usually get to do the presentations that give you all the glory you will be stuck somewhere in the middle.

301
00:32:44,140 --> 00:32:50,260
The good thing is you get involved in multiple studies, but the bad thing is people will sometimes think, well, what did he do?

302
00:32:51,160 --> 00:32:59,050
Did he actually do anything? I assure you I did. But there are no convinced because you are far, far, far away from this fairly often.

303
00:32:59,590 --> 00:33:05,350
So to some quite depend on the perspective of people who are high up in universities.

304
00:33:05,350 --> 00:33:14,770
They're not convinced you're doing them or rational independent become the promotion more difficult.

305
00:33:14,770 --> 00:33:20,380
I'm talking in an academic sense, but perhaps the same might be true of an investigative degree.

306
00:33:20,950 --> 00:33:24,249
You may not always see you as a leader.

307
00:33:24,250 --> 00:33:27,490
And so it does say a problem of think, which is fair to say.

308
00:33:28,930 --> 00:33:33,370
I think you can get there in the end, but it tends to be a little bit more difficult or slower.

309
00:33:33,370 --> 00:33:39,279
And I think reinventing the wheel, this is my other bugbear of travel appeal.

310
00:33:39,280 --> 00:33:43,959
Every time I'm doing something with someone from before and done it better, but somehow I'm still doing it.

311
00:33:43,960 --> 00:33:48,580
And what can work of the and it's been solved.

312
00:33:49,960 --> 00:33:54,610
But we can always be assured, of course, the good things that we always go round with.

313
00:33:54,880 --> 00:34:01,630
So you will get there in the end for those seem like the groups waste of time but there's a path.

314
00:34:01,960 --> 00:34:12,100
DMC is very good to do. DMC don't get me wrong, but they are a lot of work and you get quick turnaround.

315
00:34:12,220 --> 00:34:15,360
So you're expected to do things maybe in a couple of weeks and turn out a

316
00:34:15,360 --> 00:34:19,750
report or 30 pages and not make any out of it and then people pore over them.

317
00:34:20,410 --> 00:34:24,830
It's never as automated as you feel that she is the last thing totally true of DMC.

318
00:34:24,830 --> 00:34:31,330
You feel like you've done one 2003, but that for a member of a poster you can get another better.

319
00:34:31,780 --> 00:34:39,069
This is the Sword of Damocles. There is a study of up an operating corp of Damocles group, which is a great piece of work,

320
00:34:39,070 --> 00:34:49,330
which was giving guidance on Dmcs and how to set them up and also just barely hum the terms of reference.

321
00:34:49,330 --> 00:34:56,380
You can use authority. And so that idea is the sword of Damocles of sort of deciding where the study continues or not.

322
00:34:56,410 --> 00:34:58,959
I'm suggesting the Sword of Damocles in a different sense.

323
00:34:58,960 --> 00:35:05,800
And that's all your spare time to do interesting things, just about to get chopped off your life.

324
00:35:07,150 --> 00:35:11,380
And perhaps most greatly you get no direct academic credit for it.

325
00:35:11,410 --> 00:35:14,459
If you're there at the end, then you will get credit somewhat.

326
00:35:14,460 --> 00:35:21,070
But no one will give you any credit. Really. No one will really thank you or hardly anyone sees the results of your labour.

327
00:35:22,000 --> 00:35:27,819
But three or four people that you may have heard the definition of insanity,

328
00:35:27,820 --> 00:35:31,180
which is doing the same thing over and over and expecting different results.

329
00:35:31,180 --> 00:35:42,280
But it is what you did in the DMCA. It is quite ironic, I think, but I move on to further analysis and some studies I've taken it.

330
00:35:42,280 --> 00:35:49,270
No one's particularly interested in the screws because there are different types of screws and it's very important to have different types of screws.

331
00:35:50,590 --> 00:35:58,780
But there is a reason why some studies or some studies on further analysis or further analysis of the symbol fireflies.

332
00:35:58,780 --> 00:36:01,540
Most of us are not that interested in the types of fireflies.

333
00:36:02,830 --> 00:36:07,300
And so I wanted to give you a quick list of five things I would say I've learned to watch out for.

334
00:36:08,080 --> 00:36:11,710
One is the observational comparison. A The randomised controlled trial,

335
00:36:11,950 --> 00:36:18,129
the main of investigators have come to me later after they've seen the main results and

336
00:36:18,130 --> 00:36:22,390
basically talk their way around saying could you not just do the observational comparison?

337
00:36:23,440 --> 00:36:33,579
That is incredible. There is also the there is a positive study and they're waiting for discovery to be as applicable.

338
00:36:33,580 --> 00:36:37,840
False, positive or false negative discovery.

339
00:36:39,100 --> 00:36:44,320
That's this case of well, we'll just keep analysing it in various ways and eventually we'll get the results we want.

340
00:36:45,850 --> 00:36:50,290
There's also the I have data, please make me a paper or an X people to turn off.

341
00:36:50,290 --> 00:36:53,349
The study was done that they associated with it.

342
00:36:53,350 --> 00:37:00,450
They hadn't really thought that what they were just going who for imaging data or something like that and then you're supposed to progress

343
00:37:00,450 --> 00:37:10,719
the study in your spare time because no one ever gives you time to do the completely unrelated substudy I think is a very dangerous one.

344
00:37:10,720 --> 00:37:14,770
And then the ill judged a fairly new one to me recently.

345
00:37:15,450 --> 00:37:26,970
Came across a study for the BMC member where to see how felt it was a good idea of the two Ph.D. PhDs were closely tied to the progress of the trial.

346
00:37:27,180 --> 00:37:30,690
Because we all know trials never, ever needed extensions.

347
00:37:30,690 --> 00:37:33,300
The recruitment periods are not always, always on time.

348
00:37:34,140 --> 00:37:44,790
And therefore we were in this awkward position of stations failing to get access to the data very well to start is still ongoing because they said,

349
00:37:44,790 --> 00:37:49,380
well I needed to cement for my my PhD in very idea.

350
00:37:53,300 --> 00:37:59,660
Which leads me on to data requests. And this person up here, that's how it feels like as the or system.

351
00:37:59,670 --> 00:38:04,870
You just pick things for the various spokes and eventually that thankless task.

352
00:38:04,880 --> 00:38:11,150
And also you spend a lot of time clearing up after the people and the trial.

353
00:38:11,150 --> 00:38:16,250
It can never end. Sometimes it can get in the way. It's not it's not the ones you have to live with, thankfully.

354
00:38:16,250 --> 00:38:22,489
But other people have a bright idea or layer that they want the data for something else to get long term.

355
00:38:22,490 --> 00:38:27,049
Both the study time comes back to life and then there's more work involved.

356
00:38:27,050 --> 00:38:36,260
Meta Analysis. If I got a request last year for 35 years after the study was published for a really evaluation of one of the outcomes.

357
00:38:36,830 --> 00:38:42,100
So that takes quite a bit of time to remember. Well, how come that we're talking about even more study in our time?

358
00:38:42,110 --> 00:38:44,420
I find though get access to the data.

359
00:38:44,660 --> 00:38:54,760
I've moved cities by that point and then to read Coda and then produce the difference in which case you send it to them.

360
00:38:54,810 --> 00:38:59,720
They don't apply, which made it real for you feel really worthwhile.

361
00:39:00,710 --> 00:39:08,090
But I think that it did include an analysis of more systematic review and then you might get the last one, the next trial.

362
00:39:08,300 --> 00:39:12,959
If you do a good job or a half decent job, people will come to you. Their night is always electronic.

363
00:39:12,960 --> 00:39:18,500
They could see eyes as an extra and you might get rushed into doing things to help along

364
00:39:18,510 --> 00:39:24,409
with some device that can be a good opportunity or it can be quite burdensome burden,

365
00:39:24,410 --> 00:39:30,980
some thing that's not on your desk and so you get pulled in lots of different directions.

366
00:39:31,440 --> 00:39:35,050
I put all these in red because everyone wants it.

367
00:39:35,090 --> 00:39:41,210
It's all over here. And this all happened to me in the last couple of months and it's all different.

368
00:39:41,570 --> 00:39:48,290
So different things going on and you can feel a bit overwhelmed because no matter how well you plan,

369
00:39:48,290 --> 00:39:54,320
different culture will be working on different trials and they will have their own timescales on agendas you can't control.

370
00:39:54,650 --> 00:40:00,410
So there are moments where you feel a bit like this guy up here, the ugly.

371
00:40:00,950 --> 00:40:06,620
But I move towards the end. Probably the most uncomfortable thing is the bearer of bad news.

372
00:40:09,200 --> 00:40:12,950
The classic one is the sample size and it is much bigger than you expect.

373
00:40:13,970 --> 00:40:18,890
Akin to 60 isn't really that many or variations on that thing.

374
00:40:19,610 --> 00:40:27,020
The trial result is negative, so despondency you can create in people as they see the result, which is quite absurd.

375
00:40:27,350 --> 00:40:34,610
And then just this could be people's plan over the next five years work to the premise of this intervention that

376
00:40:34,610 --> 00:40:40,100
if you spend the last five years working on and then you're showing that it doesn't seem to actually work,

377
00:40:43,190 --> 00:40:46,880
and you also spend a lot of time saying, no, no,

378
00:40:46,900 --> 00:40:55,100
we can just change the analysis because you don't like the result of about one third of and recently said we didn't want to be part of study.

379
00:40:55,310 --> 00:41:00,580
You didn't realise the result could show that something didn't work everywhere.

380
00:41:01,470 --> 00:41:06,830
So no one what seems to make sense to you is statistically a bad idea.

381
00:41:06,890 --> 00:41:12,680
We do spend quite a while unfortunately explaining what seems quite intuitive actually isn't a good idea.

382
00:41:15,020 --> 00:41:22,309
You can analyse the data halfway through because you're bored keen to go to conference one Christmas to come only variations on a theme.

383
00:41:22,310 --> 00:41:26,470
But you do get people who feel the company could be not just send an outsider.

384
00:41:26,480 --> 00:41:30,590
Can we not just give us. I know it's taking a long time, so it could be not.

385
00:41:33,080 --> 00:41:37,880
And then probably the worst thing is that, Houston, we have a problem because of your role.

386
00:41:38,360 --> 00:41:43,669
They can they monitoring emails or other things. You see data where you observe the randomisation.

387
00:41:43,670 --> 00:41:49,820
You're the person whose fault the programmers made an error in the randomisation code.

388
00:41:50,660 --> 00:41:55,940
Therefore the randomisation has not been working properly or you're the person response when

389
00:41:55,940 --> 00:42:00,499
someone asks you to analyse something in the hopes that the data doesn't exist because

390
00:42:00,500 --> 00:42:07,120
nobody actually remember to put it in the CRF or you're the person who worked out the

391
00:42:07,220 --> 00:42:12,020
one point we were asking about Saturdays and every time we're asking for three days,

392
00:42:12,350 --> 00:42:15,440
why would we choose to talk about sororities and Fridays?

393
00:42:15,440 --> 00:42:18,500
But it's the same study. Surely we just want to pick one day.

394
00:42:20,720 --> 00:42:27,170
So you unfortunately are the person who finds these things often and then has to go to people and explain to them.

395
00:42:27,500 --> 00:42:31,700
And it's almost like you've created this problem where really you're just the person who finds it,

396
00:42:32,210 --> 00:42:36,740
but then you're also left trying to resolve it as best can be done and can.

397
00:42:39,560 --> 00:42:43,240
Then I think my last one is ignorance.

398
00:42:43,240 --> 00:42:46,480
These are bandits, by the way. This is like you are confused.

399
00:42:48,670 --> 00:42:52,299
This comes in various forms. This is my great moment to some reviewers.

400
00:42:52,300 --> 00:42:56,440
You spend all this time doing all this work and then you get little statements

401
00:42:56,440 --> 00:43:00,970
where people just basically scrap your whole study and make it all the rubbish

402
00:43:01,630 --> 00:43:05,860
and like a couple of sentences which don't even make that much sense in my

403
00:43:05,860 --> 00:43:12,909
opinion or I was recently as part of a group were accused of confirmation bias.

404
00:43:12,910 --> 00:43:18,200
Who knows what confirmation bias is the one one or two people.

405
00:43:18,200 --> 00:43:24,339
Well, I was guilty of anybody, which was news to me or we were as a group because the person wrote a nice,

406
00:43:24,340 --> 00:43:30,010
eloquent statement about how they'd been so clever to in about two sentences in the whole paper.

407
00:43:30,520 --> 00:43:33,849
But we can go to confirmation bias. And so again,

408
00:43:33,850 --> 00:43:40,930
they just discarded the whole study on the basis two sentences which could just be rewritten if they really had an objection to two sentences.

409
00:43:41,650 --> 00:43:48,370
So it can be very frustrating. I think it's a tough decision by the skill and effort process.

410
00:43:48,370 --> 00:43:57,070
It's time to produce something that's rigorous and has produced data that people can share and get out if you really don't like what you've done.

411
00:43:57,640 --> 00:44:00,790
And yet people will still just describe it fairly simply.

412
00:44:02,590 --> 00:44:06,999
And then I was also in a group that was accused of academic dissembling.

413
00:44:07,000 --> 00:44:13,480
Dissembling is also not a word I was familiar with, but let me suggest that's not a nice thing to be accused of.

414
00:44:13,630 --> 00:44:17,890
And that, again, related to how we were dealing with adverse events.

415
00:44:18,940 --> 00:44:21,139
So it was only a small component of a report,

416
00:44:21,140 --> 00:44:29,410
but nevertheless we were keen for the academic dissembling someone friend to reporters, to the front door of the university.

417
00:44:31,690 --> 00:44:37,840
You do get co-operators in randomised control because people want to be involved who have a remarkable level of ignorance.

418
00:44:40,180 --> 00:44:47,490
I have heard surgeons again, I'm always using surgeons, the other people around us, but I didn't realise it being a randomised controlled trial then.

419
00:44:47,500 --> 00:44:54,340
I didn't choose the treatment that I've ever done when I was there.

420
00:44:54,340 --> 00:44:59,650
People talk about other people who are way ahead of everyone else, including blah blah people.

421
00:44:59,920 --> 00:45:06,129
I'm just making this not clear who is on the statistics for this person who is way ahead of everyone else.

422
00:45:06,130 --> 00:45:14,950
I have never done a trial in their life, but they were a mathematician, so therefore they must know everything by them.

423
00:45:16,090 --> 00:45:20,020
I'll never side as a reviewer and you can make investigators very unhappy.

424
00:45:20,020 --> 00:45:32,260
And I've had one very eminent photo op and basically try and pressurise me in writing a letter to retract my review because he didn't like it.

425
00:45:34,090 --> 00:45:40,840
So good thing think it was very important to. But the bad thing is because the law is at stake sometimes.

426
00:45:40,840 --> 00:45:46,360
I'm not saying this is a normal you can have a lot of pressure put on that special and

427
00:45:47,440 --> 00:45:51,550
it comes out to the ignorance of the people who are people who should know better,

428
00:45:51,940 --> 00:46:00,460
but they do quite things. So to finish some more period, I like trials.

429
00:46:00,820 --> 00:46:05,200
I'm glad it worked because it was opportunistic, but I got involved in trials.

430
00:46:05,200 --> 00:46:09,999
I would encourage you to work in trials in whatever way works for you.

431
00:46:10,000 --> 00:46:13,330
And I would say there are glimmer of opportunities.

432
00:46:15,100 --> 00:46:18,620
There is opportunities to get publications. In fact, quick wins. I was very lucky.

433
00:46:18,640 --> 00:46:24,440
My first trial we got involved and end up getting published in the BMJ, but also which was very nice for me.

434
00:46:24,440 --> 00:46:30,129
I didn't have to be there all the way through it. I just turned up and somehow they didn't have a statistician at the very end.

435
00:46:30,130 --> 00:46:31,330
I needed one.

436
00:46:31,780 --> 00:46:40,120
I got involved early on, so a friend of mine, six months after being champion, wanted me to get things out of the trials the first happened,

437
00:46:41,260 --> 00:46:48,460
but also you do get to be involved in protocol publications and the results that are sob stories, good studies and all that.

438
00:46:49,600 --> 00:46:53,860
And so it's a good way to build up your CV and just get involved in research.

439
00:46:54,970 --> 00:47:02,200
There are opportunities involved in large grants. Then hundreds of thousands and even millions and projects are not uncommon.

440
00:47:02,410 --> 00:47:09,400
There aren't many of them. So that's true of and also that you could be involved in multiple studies.

441
00:47:11,500 --> 00:47:16,729
There are opportunities to specialise, although it's a very big area and a lot of statisticians,

442
00:47:16,730 --> 00:47:19,629
there's never enough from everything to know about this.

443
00:47:19,630 --> 00:47:26,800
If you know a bit you know a lot more than a lot of these collaborators, as I've just demonstrated to you on this very early on,

444
00:47:27,130 --> 00:47:31,990
an opportunity for you to specialise and become an expert in a soft type of study,

445
00:47:31,990 --> 00:47:37,810
whether that's cost of college or whether it's primary care trials or whether it's cost of primary care time as well have about.

446
00:47:37,860 --> 00:47:43,649
Of adopted designs or whatever is very easy to come up with a scenario where you can

447
00:47:43,650 --> 00:47:48,840
specialise and have some expertise that very few other people will be that comfortable with.

448
00:47:49,080 --> 00:47:56,700
No one will do every type of trial and service career opportunities and you do get opportunities to do metallurgical research alongside.

449
00:47:56,700 --> 00:48:01,269
It does raise operational issues as you go along. So well respected there.

450
00:48:01,270 --> 00:48:07,380
If people are interested, if you can make a small difference in clinical trials because there's so many of them going on,

451
00:48:07,770 --> 00:48:12,600
you can make a small difference that's important. So people will publish and people will be interested.

452
00:48:12,840 --> 00:48:19,740
So if you can find a little way about doing trials, something making slightly better people will care.

453
00:48:21,090 --> 00:48:28,920
And I'd say you have an opportunity to make a difference, browse through and have make an impact on a patient's health.

454
00:48:29,790 --> 00:48:36,110
So my final thoughts. If your question is to be or not to be a trial statistician, I don't know.

455
00:48:36,130 --> 00:48:39,270
The answer is because it's very much a personal thing. I'm very quiet.

456
00:48:39,270 --> 00:48:43,259
I how to come on. I still work in there. This is how I spend my days.

457
00:48:43,260 --> 00:48:52,620
But you have to answer a lot of questions in your cell. But I do quite like this, for there's never a surefire cure, really, except during good work.

458
00:48:52,980 --> 00:48:57,750
My view is that doing trials is good work, but it starts with.