1
00:00:01,920 --> 00:00:08,850
Welcome to everyone. My name is Rafael PEREIRA, a medical statistician working in the Department of Primary Health Care.

2
00:00:09,000 --> 00:00:12,360
I've been involved with the centre for the last five years, I think.

3
00:00:12,870 --> 00:00:20,670
And today we're going to be talking mainly about rapid critical appraisal, particularly of controlled trials,

4
00:00:21,780 --> 00:00:29,700
which is sort of the second or third step of your evidence based medicine steps.

5
00:00:30,240 --> 00:00:35,340
So before I go into this, I hope you all received an envelope with something in it.

6
00:00:35,850 --> 00:00:42,840
Please keep it short. Don't open it. The second thing I if you haven't, please ask only for one.

7
00:00:43,710 --> 00:00:44,550
If you haven't got one,

8
00:00:44,850 --> 00:00:55,110
second thing I need you to have is to have your workbooks because we're going to do a bit of work that involves reading and appraising.

9
00:00:56,060 --> 00:01:02,370
You have it with you ask, all evolve has some spare copies as well that you can borrow if there's loads and those of you.

10
00:01:02,400 --> 00:01:06,600
Then I'll ask you to work in groups of two or three.

11
00:01:09,180 --> 00:01:13,800
All right. I'll need a lot of patient, patient participant participation.

12
00:01:14,250 --> 00:01:25,080
Okay. So as Paul mentioned earlier on, we're really looking at five different steps in the evidence based medicine to do evidence based medicine.

13
00:01:26,280 --> 00:01:29,600
Mike is on the OC.

14
00:01:30,030 --> 00:01:33,630
Is the mike on the slide?

15
00:01:34,020 --> 00:01:37,380
Can you hear me now? I'll predict, I hope.

16
00:01:38,340 --> 00:01:46,130
Okay. Paul has mentioned or mentioned in the morning that the five different steps of evidence based medicine,

17
00:01:46,350 --> 00:01:50,129
the first to formulating an answerable question and tracking down the best evidence.

18
00:01:50,130 --> 00:01:59,580
The first one, sorry, was mentioned at the beginning of the first plenary session and was also part of the group sessions right now.

19
00:02:00,480 --> 00:02:03,660
The second step tracking down the best evidence will be done this afternoon.

20
00:02:03,840 --> 00:02:10,530
Okay. Today, we're going to right now we're going to be focusing mainly on critically appraising the evidence for the

21
00:02:10,530 --> 00:02:15,240
validity of the evidence that we find with the studies that we we've identified are actually valid,

22
00:02:15,510 --> 00:02:18,960
the impact of the size of the benefit. And finally, the applicability.

23
00:02:19,110 --> 00:02:24,060
Well, the findings that we have in this particular study can be applied in our setting.

24
00:02:27,670 --> 00:02:32,440
Now, when we're looking at using evidence based medicine, generally,

25
00:02:32,440 --> 00:02:41,680
we think of a series of items that we need to check and decide whether they're actually there or not.

26
00:02:42,070 --> 00:02:46,260
To define it, these particular studies are actually useful or not.

27
00:02:46,270 --> 00:02:50,860
So if you type a critical appraisal checklist and randomised controlled trials in Google,

28
00:02:51,190 --> 00:03:00,610
as you can see you get in 2005, you get 36,000 articles going up massively, 2006 it's getting a little bit better.

29
00:03:00,610 --> 00:03:09,640
By 2008 we only have 97,000 articles that proposed different checklist of doing critical appraisal of randomised controlled trials.

30
00:03:09,940 --> 00:03:15,399
And this is a sort of a typical checklist with 20 different items that you go

31
00:03:15,400 --> 00:03:18,700
from the beginning to the end with your trial and with your study in hand,

32
00:03:18,780 --> 00:03:22,270
in the other hand, and you go take one, take two, take three.

33
00:03:22,480 --> 00:03:26,440
Probably by the time you finish reading all of them, we've forgotten what your question was in the first place.

34
00:03:26,500 --> 00:03:32,739
Okay, so what are we going to try to do today is give you a framework of two simple acronyms

35
00:03:32,740 --> 00:03:37,690
to try to remember how to do critical appraisal and first critical appraisal,

36
00:03:38,050 --> 00:03:47,560
particularly of randomised controlled trials. And I think during the small group exercises you all looked at exercise,

37
00:03:47,560 --> 00:03:56,920
one of your abstracts had to deal the dealt with this particular clinical question which is in people who take long haul flights,

38
00:03:56,920 --> 00:03:59,830
does wearing graduated compression stockings prevent DVT?

39
00:04:00,160 --> 00:04:05,560
So the first question I want to ask is how many of you have had a long haul flight in the last couple of years?

40
00:04:06,370 --> 00:04:14,950
Just don't know how many of you up for stockings, compression stockings.

41
00:04:17,110 --> 00:04:25,300
Okay. And those are you didn't let's see how many of you those of you that didn't, how many of you would think about wearing compression stockings?

42
00:04:28,070 --> 00:04:31,640
Okay. Let's see if we can change that towards the end of the session, or maybe not.

43
00:04:32,960 --> 00:04:36,650
So for those of you who don't know what compression is talking, so I have an example.

44
00:04:37,760 --> 00:04:44,690
So these are the typical compression stockings worn by one of the teachers in this week's course.

45
00:04:44,870 --> 00:04:49,580
And I need a volunteer to wear these stockings.

46
00:04:52,210 --> 00:04:58,240
Washington. They've been very clean. Yeah.

47
00:04:58,260 --> 00:05:03,150
Three or four can do it. And give the talk at the same time someone could.

48
00:05:03,270 --> 00:05:08,430
Actually, you're in perfect position because it's very similar to what type of seats you were getting on an aeroplane as well.

49
00:05:11,770 --> 00:05:14,790
Yeah. Two on the flight. Could you put both on? Thanks.

50
00:05:14,820 --> 00:05:28,140
You. Okay. Now, in your workbooks, you can find in page 73 of your workbooks.

51
00:05:35,320 --> 00:05:38,740
What search function to try to answer this particular question.

52
00:05:39,490 --> 00:05:43,810
Okay. And they came up. Well, we came up with a particular study.

53
00:05:47,020 --> 00:05:54,460
By Ska in 2001, and the full study of the full paper can be found page 95.

54
00:05:54,780 --> 00:06:04,950
So if you go to page 95, what books? Okay.

55
00:06:05,730 --> 00:06:10,560
You'll be able to see the full study as reported in The Lancet in 2001.

56
00:06:11,520 --> 00:06:19,950
For the rest of the presentation, I'll be asking you to come up with the answers for the different parts of the quest of of the critical appraisal.

57
00:06:20,220 --> 00:06:25,500
Should be able you should be able to find in the paper right there.

58
00:06:25,800 --> 00:06:36,210
Okay. So the first thing to do is to think about whether this particular study that we found is relevant and can answer the question that we have.

59
00:06:36,870 --> 00:06:46,020
And for that, we've already generated a structured question using a picture from a clinical setting.

60
00:06:47,490 --> 00:06:56,819
And what we want to do is do exactly the same to generate the picture or the question that was trying to answer whoever wrote this particular paper.

61
00:06:56,820 --> 00:07:02,520
So we weren't going to set up to find the high school for this particular trial.

62
00:07:02,790 --> 00:07:06,600
So I want you to have a look. I'll give you 20, 30 seconds.

63
00:07:06,810 --> 00:07:11,520
Search around in the and the article to identify the participants,

64
00:07:11,820 --> 00:07:17,070
the intervention group, comparison group and the outcome that they were interested in.

65
00:07:53,120 --> 00:07:56,360
It's not a solo exercise. We obviously can confer with your neighbours.

66
00:07:57,260 --> 00:08:02,020
You're happy to talk, but it is a critical appraisal exercise.

67
00:08:02,030 --> 00:08:10,939
So let's let's crack on. The other thing that I want you to notice here, I've set up the PI connects to a series of geometrical shapes,

68
00:08:10,940 --> 00:08:19,160
and this is what is called the Gape framework developed by Rod Jackson.

69
00:08:19,580 --> 00:08:28,790
And the way he came up with this series of figures is he was looking at where we are, the Xbox generation.

70
00:08:29,060 --> 00:08:36,560
And this is pretty much what he's that's a lot of people that use play with Xbox and these are actually the figures that come in,

71
00:08:36,560 --> 00:08:37,700
the experts, a triangle,

72
00:08:38,120 --> 00:08:48,980
a circle, a square, and then the light from the triangle will represent precisely the population with unidentified target that you want to look at.

73
00:08:49,190 --> 00:08:54,260
And then how it narrows down to who you actually included in your trial.

74
00:08:54,680 --> 00:08:59,990
Then you have a circle which sort of creates the population that you encounter.

75
00:09:00,500 --> 00:09:08,660
So we put into the trial into two groups, the intervention group, whoever you want to, to apply whatever therapy it is that you want to apply.

76
00:09:08,840 --> 00:09:13,190
And then a control group or comparator group and the circle with dividing in the middle

77
00:09:13,370 --> 00:09:19,049
sort of highlights the fact that actually looking at pretty much the same individuals,

78
00:09:19,050 --> 00:09:25,190
so the same type of comparable groups with the only difference that these two groups have being the intervention as such.

79
00:09:25,340 --> 00:09:29,720
So anything that we find different between these two groups will be down to the intervention.

80
00:09:30,260 --> 00:09:36,169
And finally, an outcome which is a square just typical at two by two table with what we have,

81
00:09:36,170 --> 00:09:40,129
whether they have the intervention or whether they were intervention of the

82
00:09:40,130 --> 00:09:43,580
control group and whether they have the outcome or they didn't have the outcome.

83
00:09:43,670 --> 00:09:51,469
So that's the framework. And it's very useful because as you go around, for example, through a paper and you find things about the population,

84
00:09:51,470 --> 00:09:54,870
you can just put a little triangle next to to the, to the paper.

85
00:09:54,910 --> 00:09:57,650
And they said, okay, this is something about the population, something about the population.

86
00:09:57,890 --> 00:10:00,950
You find something about the intervention as such describes that you can just put

87
00:10:00,950 --> 00:10:06,319
a little a little circle and identify the the intervention of the control groups,

88
00:10:06,320 --> 00:10:09,710
that way of descriptions of intervention and the same thing with the outcome.

89
00:10:10,190 --> 00:10:12,980
Okay, so look at this particular paper.

90
00:10:13,070 --> 00:10:25,820
What was the population that were interesting in in looking at estimates of passengers compared with 41 years of age over 50 years,

91
00:10:26,240 --> 00:10:30,260
the most history with no past history women.

92
00:10:32,330 --> 00:10:39,680
Okay. Okay, so that's fairly specific. So passengers of long haul flights, they specify the this 8 hours.

93
00:10:39,680 --> 00:10:47,540
Okay. Passengers of long haul flights longer than 8 hours over 50 returning to the UK in the next six months.

94
00:10:47,720 --> 00:10:54,530
Six weeks. Okay. So it's fairly specific. Does that match up with the P that we have in our question, critical question.

95
00:10:57,750 --> 00:11:03,390
Reminder people take long haul flights. Well, maybe it's a bit too specific because they're looking at over fifties.

96
00:11:04,290 --> 00:11:13,080
Right. Maybe just the right the right time. So they're looking at a very fairly tight group of people by way of intervention.

97
00:11:13,110 --> 00:11:20,370
What was the intervention that they're looking at? Didn't specify the type of stockings.

98
00:11:22,530 --> 00:11:25,560
So you would be able to go and buy these stockings. Okay.

99
00:11:25,760 --> 00:11:31,980
That's quite important as well in terms of whether we could actually generalise the findings or replicate the findings.

100
00:11:32,790 --> 00:11:36,480
Finally, what was the comparator that they used did not work.

101
00:11:38,550 --> 00:11:45,840
Okay. So nothing. All right. Okay. Again, I think it matches roughly a clinical setting.

102
00:11:46,110 --> 00:11:52,620
And finally, what outcome do they look to? Symptomless, DVT.

103
00:11:52,740 --> 00:11:57,030
How is that assessed? Okay.

104
00:11:57,120 --> 00:12:02,069
We'll come back to that in a second. So all of these things that we've identified in a way,

105
00:12:02,070 --> 00:12:08,580
the picture of the study that we've retrieved and what we need to decide is whether the

106
00:12:08,580 --> 00:12:14,040
picture of this particular study matches the picture that we had in our structure.

107
00:12:14,040 --> 00:12:17,880
Question. If it doesn't, we can take it away. We don't need to do anything else.

108
00:12:18,480 --> 00:12:22,170
Okay. That said, if it does, then we can go to the next step.

109
00:12:22,980 --> 00:12:29,430
We decided it can be relevant. So now we want to assess whether it is valid with internal validity or not.

110
00:12:29,820 --> 00:12:35,460
Okay. So we have a call for relevance and we have another acronym for Internal Validity.

111
00:12:35,700 --> 00:12:39,330
And this next acronym is Rambo. Okay.

112
00:12:42,510 --> 00:12:51,450
And it stands for, in the case of therapy studies, recruitment, allocation, concealment,

113
00:12:52,740 --> 00:12:58,380
maintenance measurements, and where these measurements were blinded and objective.

114
00:12:58,680 --> 00:13:02,730
And we'll go through each one of these bits in more detail in a second.

115
00:13:02,820 --> 00:13:08,910
Okay. These are the things we need to identify in order to determine whether the study study

116
00:13:09,180 --> 00:13:15,260
was done to a high quality for then the results to be of any use to us whatsoever.

117
00:13:15,270 --> 00:13:24,690
If we decide that any of these things are not done adequately, then we might or this particular study might be introducing some kind of bias,

118
00:13:25,350 --> 00:13:30,330
which would mean that the results they find, no matter how big the trial was,

119
00:13:30,750 --> 00:13:35,130
the results that they found might not be answering the relevant question.

120
00:13:35,670 --> 00:13:41,310
And that's why we want to do this. Before we do, we look at any anything related to the actual results of the study.

121
00:13:41,610 --> 00:13:45,960
And normally you would find all of these things in the methods side of of of the of the paper.

122
00:13:46,260 --> 00:13:49,319
So you look at the methods side of the paper to find things about recruitment,

123
00:13:49,320 --> 00:13:56,430
whether they did randomisation, how is the maintenance done and the type of measurements that were involved?

124
00:13:58,590 --> 00:14:03,959
So Tamarod on Rambo is the thing that we're going to be focusing on for doing critical appraisal or validity,

125
00:14:03,960 --> 00:14:08,280
internal validity of a randomised controlled trial. And just to break it into small bits.

126
00:14:08,670 --> 00:14:09,930
First we look at recruitment,

127
00:14:11,070 --> 00:14:19,380
try to identify in a way we're matching the Piku and the Rambo doing the internal validity and the relevance of the study.

128
00:14:20,610 --> 00:14:26,310
So in parallel. So the first step is to find out who did the studies that the subjects represent.

129
00:14:28,020 --> 00:14:32,639
Then we'll find out about a location, whether the assignment of treatments with randomise or not.

130
00:14:32,640 --> 00:14:36,750
We'll explain why this is important, where the groups similar to the trials start.

131
00:14:37,530 --> 00:14:43,890
Then we look at maintenance, which is where the groups are treated equally, where outcomes ascertain and analyse from those patients.

132
00:14:44,670 --> 00:14:45,420
And finally,

133
00:14:45,420 --> 00:14:52,920
the measurements and we will discuss why blinding is an issue and when blinding is an issue and what type of blinding is important or not.

134
00:14:53,220 --> 00:15:00,780
And finally, whether the measurements were objective or standardised and the objectivity and the blinding of the of the outcomes are tied in.

135
00:15:01,140 --> 00:15:06,150
So you might be more interested in blinding, particularly when the outcome is not objective or fairly subjective.

136
00:15:06,240 --> 00:15:09,660
And finally, we look a little bit about the the results and study statistics.

137
00:15:10,650 --> 00:15:18,390
So let's focus first on the steps one and two of the are the recruitment and allocation.

138
00:15:18,400 --> 00:15:25,290
I want you to have a look at, again, the article and try to identify who they managed to recruit and.

139
00:15:27,730 --> 00:15:37,000
What how do they actually work through their allocation. So how do they randomise and whether groups around the trial trial stuff.

140
00:15:37,990 --> 00:15:44,990
And in many cases you might find these in figures and tables rather than lengthy descriptions.

141
00:15:44,990 --> 00:15:49,990
So have a look at try to make it fairly. You'll find that in practice.

142
00:15:49,990 --> 00:15:55,420
With practice, you might get this really, really fairly quickly. I'll give you again another 30 seconds.

143
00:16:40,820 --> 00:16:50,080
Okay. Let's move on. So in terms of who they collect, who they identify, did anyone find?

144
00:16:53,050 --> 00:16:57,040
Well, we already talked about what type of population they were interested in finding.

145
00:16:57,250 --> 00:17:05,260
So these were people in long haul flights over 50 with no previous history of deep vein thrombosis.

146
00:17:05,680 --> 00:17:09,310
Did they get those people to anyone?

147
00:17:09,890 --> 00:17:13,030
I mean, it's not related to this, but they actually collected.

148
00:17:14,170 --> 00:17:21,340
Because sometimes a study would say this is the type of population we want to identify and the people they actually get in is completely different.

149
00:17:22,270 --> 00:17:26,000
Okay, so the pi coin, the original question might not be.

150
00:17:26,020 --> 00:17:32,080
May not match the population or the people that weren't actually in the trial.

151
00:17:32,380 --> 00:17:38,170
What about here? Do they actually match? 50%.

152
00:17:38,680 --> 00:17:42,760
50%. That's okay.

153
00:17:42,970 --> 00:17:49,730
The only one identified sort of how they recruited. Newspapers took.

154
00:17:55,250 --> 00:17:59,600
Okay. And that was to determine whether they were okay.

155
00:17:59,870 --> 00:18:09,060
So in a sense, it seems that, yes, that part was tight and fight was an issue to decide whether there were 50 probably.

156
00:18:09,860 --> 00:18:13,430
What about long term, long haul flights? That was a third part of the inclusion criteria.

157
00:18:17,360 --> 00:18:22,610
Okay. And why they included people on the aircraft.

158
00:18:22,910 --> 00:18:32,020
I mean, I've just quickly gone through the. Okay. And I was trying to see why I think was the question, was prevention simple?

159
00:18:32,600 --> 00:18:40,729
I mean, I'm I felt like they wanted to see a higher frequency indication that this expired high frequency goes above 50.

160
00:18:40,730 --> 00:18:44,030
But knowing the opposite, this is why you chose specifically.

161
00:18:44,150 --> 00:18:46,219
Okay. Okay.

162
00:18:46,220 --> 00:18:53,540
So that they might that might be something that if you were to apply these findings to someone else, if they were under 50, you might disable it.

163
00:18:54,230 --> 00:18:58,220
I'm not entirely sure. Yeah. Okay. Generalisability of the question.

164
00:18:59,900 --> 00:19:03,150
Okay. Yeah. That that could be maybe one of the shortcomings.

165
00:19:03,170 --> 00:19:07,930
Maybe you wanted to know further, but I think in a way, you've answered the question yourself.

166
00:19:08,600 --> 00:19:11,690
But I think sometimes I should, because at the hospital, when they said that,

167
00:19:12,140 --> 00:19:17,180
they think that that was not ideal authority to exclude individuals at highest risk.

168
00:19:17,990 --> 00:19:25,580
So, again, I'm not really sure what are trying to say, but I expect that they did at one point a greater frequency,

169
00:19:25,970 --> 00:19:29,210
but yet at the last part, they excluded girls who are at high risk.

170
00:19:29,480 --> 00:19:35,570
And I think in this case, the reason for excluding those at higher risk is to in a way so that previous episodes.

171
00:19:36,170 --> 00:19:41,860
Yeah. And questions in a previous. Yeah. Yes.

172
00:19:42,230 --> 00:19:49,130
And also to make it more generalisable to the overall population, I guess, and to decide whether the long haul flight was positive or not.

173
00:19:52,740 --> 00:19:56,490
That's right. Okay. Let's see how it is played.

174
00:19:56,490 --> 00:20:00,900
The sample size is high sample size. I think they don't mention that.

175
00:20:01,720 --> 00:20:05,490
Oh, sorry. They do mention that. We've got we'll go into that much later on.

176
00:20:07,440 --> 00:20:12,419
So in terms of whether they collect the right sort of people,

177
00:20:12,420 --> 00:20:18,540
I think they did at least identify a group that they decided or they said they were going to identify.

178
00:20:19,620 --> 00:20:25,560
The next step is whether they actually did the right type of allocation.

179
00:20:25,890 --> 00:20:34,530
That is, whether they randomise these individuals to the two groups in such a way that no systematic bias could be found.

180
00:20:34,530 --> 00:20:42,120
That is, I will always watch all the men in one group and all the women in the other group that could potentially have a problem.

181
00:20:42,450 --> 00:20:48,750
So to do they identified what did they say, how they chose the two groups.

182
00:20:52,170 --> 00:21:00,540
Okay. So they were randomised the randomised volunteers by sealed envelope to one of two groups.

183
00:21:00,810 --> 00:21:04,830
Okay. You're happy with that. I'm not sure I'll be randomised.

184
00:21:05,610 --> 00:21:14,760
They just had an envelope. Okay. That was not unless this is related to our envelopes.

185
00:21:14,970 --> 00:21:18,270
Okay. So we have our envelopes at hand.

186
00:21:18,510 --> 00:21:35,810
Well, I need to now open them. Amanda and I.

187
00:21:40,580 --> 00:21:47,570
So you should have one of two. You've been randomly allocated to fruit pastilles or fruit gums.

188
00:21:47,720 --> 00:21:51,860
Right. Okay. And there were envelopes.

189
00:21:54,200 --> 00:21:57,560
Did anyone switch their envelope around?

190
00:21:59,810 --> 00:22:04,280
Okay. Did anyone open the envelopes before I asked? Okay.

191
00:22:04,280 --> 00:22:13,730
A couple. So we might we might have some problems in terms of of how they were how randomisation was was.

192
00:22:17,150 --> 00:22:21,470
Of course I don't want there. We have contamination swabbing.

193
00:22:25,160 --> 00:22:29,810
Yeah. Okay. And these are all things that could potentially happen in a randomised controlled trial.

194
00:22:29,960 --> 00:22:41,240
Okay. So thinking about what we've just done, we've done our own media randomisation.

195
00:22:41,260 --> 00:22:48,730
I want to ask two, but we have two particular outcomes and is related to long haul flights who have been to Australia.

196
00:22:50,020 --> 00:22:54,010
Those that have been to Australia can be raise your hands. Okay.

197
00:22:54,460 --> 00:22:59,800
How many sorry? How many of those that have been to Australia have the gums, fighting gums.

198
00:22:59,800 --> 00:23:04,390
One, two, three, four, five, six, seven, eight, nine, ten, 11, 12.

199
00:23:08,410 --> 00:23:16,240
And how many of those have the still? Three, four, five, six, seven, eight, nine.

200
00:23:18,940 --> 00:23:26,140
Okay. Okay. So in this case, I would maybe say that randomisation, roughly work.

201
00:23:26,170 --> 00:23:30,760
We have roughly the same numbers in each one of the groups.

202
00:23:31,130 --> 00:23:38,440
Okay. We have 12 and ten that if we want to do a proper statistical test, we probably will find that it's not statistically significant.

203
00:23:39,580 --> 00:23:46,900
What about who how many of those who had the guns made us noble in the last 24 hours?

204
00:23:50,180 --> 00:23:57,490
That comes. One, two, three. How about a positive?

205
00:23:59,760 --> 00:24:03,870
One, two, three, four, five, six. Twice the number.

206
00:24:05,130 --> 00:24:14,700
Okay, so here we have an outcome that's actually less likely to happen, and we have twice as many.

207
00:24:14,900 --> 00:24:16,979
Probably if we were to test this thing,

208
00:24:16,980 --> 00:24:25,290
maybe we actually show that there's a statistically significant difference between the two groups for that particular outcome.

209
00:24:25,440 --> 00:24:28,770
And this was a completely random you have to agree.

210
00:24:30,240 --> 00:24:35,460
And this is, again, this sort of illustrate the type of things that you might find in a trial,

211
00:24:35,820 --> 00:24:39,510
even though you have randomisation, even even if you do randomisation properly,

212
00:24:39,720 --> 00:24:47,799
just by chance, you might find that some characteristics in one group happen to be higher than another group of high school status.

213
00:24:47,800 --> 00:24:51,480
So why should we be a double cross? Should be for arbitrary five.

214
00:24:54,030 --> 00:25:00,990
Intention to treat as well. That's another thing that you need to. So there are some crossover here.

215
00:25:05,090 --> 00:25:14,140
Okay. All right. So how do we ensure allocation, concealment?

216
00:25:14,170 --> 00:25:18,430
There are several different methods, so you can try to deal with this.

217
00:25:19,360 --> 00:25:27,099
And generally the one that's decided to be the best is to use some kind of central computer randomisation,

218
00:25:27,100 --> 00:25:36,280
where when there's someone actually doing the randomisation for you that is outside that person, that's including the participants in the trial.

219
00:25:36,910 --> 00:25:45,610
And generally, we for example, if you're entering someone to the trial, you phone a central person doing the randomisation.

220
00:25:45,610 --> 00:25:49,899
They tell you, oh, it's labelled blah, blah, blah, two, three, four or five,

221
00:25:49,900 --> 00:25:54,340
or it belongs to the placebo group, and therefore it has to receive this type of treatment.

222
00:25:54,340 --> 00:26:02,890
And because there's that barrier between the person enrolling a participant and the person doing the randomisation,

223
00:26:03,040 --> 00:26:10,240
then in a way you make sure that there's no problems, that the concealment is actually done adequately.

224
00:26:10,810 --> 00:26:21,580
Now the second step would be used, and another approach would be to use, for example, envelopes, and this is seen as could be affected by tampering.

225
00:26:21,580 --> 00:26:29,590
And here we have a little example with if someone is very interested in finding out what group they're going to be allocated to,

226
00:26:29,920 --> 00:26:36,070
you might even look through the envelope, try to how many of you, for example, try to feel what the.

227
00:26:38,200 --> 00:26:44,710
Okay. And maybe if you had this take away was fruit pastilles versus something completely different, you would have swap don't know.

228
00:26:45,340 --> 00:26:55,140
So envelopes can be tampered with. And for that particular reason, they're seen as potentially suspect, but it's still using randomisation.

229
00:26:55,150 --> 00:27:04,020
And finally, we have some studies that are actually not randomised when the allocation is is done by some characteristic of the participants.

230
00:27:04,030 --> 00:27:09,490
I could be, for example, the date of birth when they come in at alternative alternate days.

231
00:27:09,880 --> 00:27:18,010
And the problem with that is that you've already allocated the individual just just because of that particular characteristic.

232
00:27:18,010 --> 00:27:23,890
And if that characteristic has an association with something else that might affect your findings in your trial,

233
00:27:24,130 --> 00:27:28,570
in a way, what you're doing is you're creating two groups that are different.

234
00:27:29,380 --> 00:27:32,470
They're not exactly the same. They're different because of one particular characteristic.

235
00:27:32,950 --> 00:27:39,670
And in a randomised controlled trial, the objective is precisely to create two groups that are exactly the same except for the intervention.

236
00:27:43,830 --> 00:27:53,370
So the question is, in this particular trial with a group similar to the trials, does they give us any information about that, whether they say.

237
00:27:56,960 --> 00:28:04,430
So if you have a look at your table one. Are you happy with the information they provide you?

238
00:28:04,880 --> 00:28:09,350
Do you believe that the two groups are two groups created on roughly the same?

239
00:28:12,170 --> 00:28:17,450
Do you find any differences with gender?

240
00:28:17,960 --> 00:28:23,360
Okay. Number of women. So the gender, of course, is actually about all this stuff.

241
00:28:24,020 --> 00:28:28,430
Okay. So one thing that I should ask, how many of you always fly business class?

242
00:28:30,440 --> 00:28:33,620
Okay, so you can all stay and this is relevant to everyone.

243
00:28:34,100 --> 00:28:39,730
So that may be important. Why? Because the three of you know that there's space to move around.

244
00:28:40,910 --> 00:28:48,720
Okay. So to say, what happened to those two women that were excluded?

245
00:28:48,740 --> 00:28:52,850
Are you happy with that? Okay. That was part of the inclusion exclusion criteria.

246
00:28:56,240 --> 00:29:03,200
Were those two in the same group? I remember. I think that I'm not sure if they specified.

247
00:29:03,830 --> 00:29:11,690
So besides gender, was there any other thing that you identify in terms of the baseline characteristics that might be different?

248
00:29:15,350 --> 00:29:29,060
But despite the. The last two the last to feel positive and positive now is that is that a genetic markers that right.

249
00:29:29,370 --> 00:29:30,950
Yeah. Okay. Okay.

250
00:29:31,280 --> 00:29:43,010
And again, those were done to try to identify if there was some one of these genetic markers where associated with deep vein thrombosis.

251
00:29:44,060 --> 00:29:50,930
Um, the non stockings were slightly higher and one stockings were higher, slightly higher in the other.

252
00:29:51,350 --> 00:29:57,230
I think a direct test might not show any fact, but it's similar to what we have here.

253
00:29:57,770 --> 00:30:00,640
So why is this important?

254
00:30:00,650 --> 00:30:09,960
We don't want to do a direct comparison between these two things or work out the p value of differences in the event rate of each one of these things.

255
00:30:10,310 --> 00:30:16,280
Because that's not really what we're interested in, but we want to identify that potential differences between the two groups.

256
00:30:16,640 --> 00:30:21,560
Why? Because we one need we might need to take that into account when we're interpreting our results.

257
00:30:21,920 --> 00:30:29,140
So what are we doing? Finding out, doing our analysis. We might need to, for example, identify that women, slightly more women in the stockings group.

258
00:30:31,070 --> 00:30:35,760
Does anyone know? Yeah, well, because.

259
00:30:35,880 --> 00:30:41,100
Don't you think that. Because you want to. For that.

260
00:30:44,030 --> 00:30:56,260
Four, two, one, two, one, one, six. And 45 out of 115 up the hole, I imagine, is a whole group.

261
00:30:57,160 --> 00:31:03,820
Yep. And much as I would say good points, if they had more varicose veins in my head, I wouldn't have an impact.

262
00:31:04,030 --> 00:31:08,530
So anyone know if women or men are more prone to having deep vein thrombosis?

263
00:31:09,160 --> 00:31:12,190
So I'm not clinicians. Okay.

264
00:31:12,670 --> 00:31:17,780
So if if women are more prone to having having the pain from both this and the ones in the stockings,

265
00:31:19,090 --> 00:31:25,480
what we we expect in our findings to would be favouring stockings.

266
00:31:25,720 --> 00:31:29,500
Going against stockings. Yes. Okay. Okay.

267
00:31:29,680 --> 00:31:36,940
So, if anything, this imbalance that we find might be biasing against stockings.

268
00:31:37,120 --> 00:31:37,420
Right.

269
00:31:37,870 --> 00:31:43,809
And this this other type of things that you need to take into account, we need to be thinking about when you look at the baseline characteristics,

270
00:31:43,810 --> 00:31:48,910
not only if there's a potential difference, but also in which direction those differences might bias.

271
00:31:48,910 --> 00:31:53,920
Here, we felt. Okay. So.

272
00:31:57,710 --> 00:32:07,300
So I could give you participation in that one quarter to the emails and all this.

273
00:32:08,970 --> 00:32:15,560
Okay. So it may be that men were less likely to agree to participate.

274
00:32:15,620 --> 00:32:25,620
That's true. That's true. Do you think that's just because men don't want to wear compression stockings?

275
00:32:27,390 --> 00:32:32,010
But John, John, John was the only one that volunteered, so that was against it.

276
00:32:32,820 --> 00:32:38,220
But it may be that participation was less an issue in terms of getting men to do it.

277
00:32:38,580 --> 00:32:46,530
I think if we had a more efficient group would still be who would expect to do this to have more?

278
00:32:47,920 --> 00:32:54,930
So, of course, that's right. So this is the damage was evident in the study.

279
00:32:55,410 --> 00:33:02,850
It would if we if we find if our finding is the opposite, yet if our finding is the opposite,

280
00:33:02,850 --> 00:33:12,929
that the stockings have less than sorry, this particular light, this particular issue would bias against stockings.

281
00:33:12,930 --> 00:33:21,239
So if we find that stockings actually effective, then we might be more likely to believe it if it cuts out the characteristic.

282
00:33:21,240 --> 00:33:29,880
So some people might be more willing to cooperate, theoretically saying, yeah, okay, now I know what the study is doing.

283
00:33:30,630 --> 00:33:36,800
Yeah. Do I think they mentioned that, don't they, that they highlight the fact that because these are all participants.

284
00:33:37,250 --> 00:33:46,510
Yes. So we've looked at the first two parts of our Rambo recruitment allocation.

285
00:33:46,520 --> 00:33:49,280
Let's look at the third one maintenance. And for this,

286
00:33:49,280 --> 00:33:57,530
we want to focus on whether the groups were treated equally or not and whether the outcomes were

287
00:33:57,530 --> 00:34:04,190
ascertained in the same way for both groups and ascertained from most patients and most patients.

288
00:34:04,820 --> 00:34:06,830
Why is this important? Okay.

289
00:34:07,820 --> 00:34:17,180
Let me talk very briefly about what could happen if we have groups in a randomised controlled trial that are actually treated slightly differently,

290
00:34:18,800 --> 00:34:29,150
not because of information but because of something else. There was a randomised controlled trial of vitamin E in preterm infants that's aimed to

291
00:34:29,960 --> 00:34:33,800
check whether there was an effect of vitamin D preventing ritual landfills to replace it,

292
00:34:33,800 --> 00:34:38,000
which is basically a condition with newborn.

293
00:34:38,540 --> 00:34:44,150
Sorry, preterm infants basically go blind. That's that's the simple way to explain it.

294
00:34:44,690 --> 00:34:50,150
And they found that there was an effect so that, yeah, vitamin E did prevent this.

295
00:34:50,840 --> 00:34:58,219
However, looking more closely what was happening is that in order to give them vitamin D,

296
00:34:58,220 --> 00:35:05,900
they had to be removed from the 100% oxygen chamber, and they were given that vitamin D and then put back in.

297
00:35:06,410 --> 00:35:10,010
And it was actually that wasn't done to the control group.

298
00:35:11,180 --> 00:35:13,280
Okay. So it was that aspect.

299
00:35:13,700 --> 00:35:22,130
Rather than giving them or not giving them a vitamin B, vitamin D that was having the impact on the outcome that was preventing them going blind.

300
00:35:22,490 --> 00:35:31,460
Okay. So whenever we do a trial, we need to make sure that apart from the intervention, everything else is done in the same way.

301
00:35:31,550 --> 00:35:37,190
Otherwise it might be whichever other thing is not being done in the same way for for both groups,

302
00:35:37,460 --> 00:35:41,000
that might be the one responsible for the effect that we have.

303
00:35:41,180 --> 00:35:45,920
We actually getting rid of the, of the, of the main benefit of doing a randomised controlled trial.

304
00:35:46,250 --> 00:35:48,110
So in the case of for example,

305
00:35:48,110 --> 00:35:57,110
a drug trial is relatively straightforward because we have a drug trial and generally a placebo and everything else is done in the same way.

306
00:35:57,110 --> 00:36:04,040
So you give, you give them medication, you tell them to, you tell whoever it is to take them two or three times a day.

307
00:36:04,190 --> 00:36:07,489
You follow it up for however long, and then you find what the outcome is.

308
00:36:07,490 --> 00:36:11,870
So everything is exactly the same, and then you can be sure if there's a difference,

309
00:36:12,080 --> 00:36:16,490
then the active ingredient in the drug, in the medication is what's actually having the effect.

310
00:36:17,060 --> 00:36:21,170
Okay. In some of the interventions like for example, this one.

311
00:36:22,130 --> 00:36:27,709
Yeah, stockings it's a bit more different were even more difficult to identify that all the

312
00:36:27,710 --> 00:36:32,300
participants did exactly the same thing or had slightly the same type of intervention.

313
00:36:32,930 --> 00:36:38,480
But they do mention something and they talk about whether there was equal treatment in the two groups.

314
00:36:38,900 --> 00:36:45,320
Okay. In order to do that, they they, for example, present something in table three.

315
00:36:45,560 --> 00:36:54,320
And they they mentioned that looking at particularly had drugs that might affect the propensity of an individual to have deep vein thrombosis,

316
00:36:54,500 --> 00:37:00,860
that both groups were actually roughly in the same types of medications aspirin, hormone replacement therapy, etc.

317
00:37:01,400 --> 00:37:08,840
Whether it's anything else in terms of how these two groups were treated, do they mentioned that they were treated equally or differently?

318
00:37:09,200 --> 00:37:12,290
Can you have a look? I'm going to wait 10 seconds for you to.

319
00:37:18,460 --> 00:37:22,480
To identify anything else that was done differently or where they both treat in the same way.

320
00:37:32,720 --> 00:37:42,710
No. You have to turn to donations for the research volunteers that have to travel to time ships before their travel.

321
00:37:43,020 --> 00:37:46,580
Okay, I decided that it is just not practical to do that.

322
00:37:46,700 --> 00:37:55,140
Okay. The rest of the volunteers will have. And it doesn't actually say which group which you belong to.

323
00:37:55,500 --> 00:37:59,960
Okay. So it potentially could have made a slight difference.

324
00:37:59,970 --> 00:38:05,360
And they also don't don't say whether they happen to be the ones that had the vein thrombosis or not.

325
00:38:05,370 --> 00:38:13,380
But apart from that, I think that they do specify that they try to treat both groups in exactly the same way they gave the same type of advice.

326
00:38:14,400 --> 00:38:18,080
Okay. What about the follow up?

327
00:38:18,080 --> 00:38:23,389
And this is in terms, again, of maintenance, deciding whether they looked at it.

328
00:38:23,390 --> 00:38:29,660
They follow all the individuals they try to ascertain everyone in and they analysed everyone that got into into the trial.

329
00:38:30,140 --> 00:38:35,360
Now, what do you have on the right hand side? Is some sort of a flow diagram or flow diagram.

330
00:38:35,480 --> 00:38:39,680
So everyone familiar with them who's not familiar with load diagrams?

331
00:38:41,120 --> 00:38:50,449
Okay. Okay. So we have a total population that were considered and then how they, they, they go so how they excluded some, how many were excluded,

332
00:38:50,450 --> 00:38:56,509
how many were randomised to each one of the groups and then what happened to them and if they were excluded,

333
00:38:56,510 --> 00:39:02,900
the reasons why they were excluded and finally how many of them were analysed and we have hundred in each one of the groups.

334
00:39:04,940 --> 00:39:12,830
This is important because. But particularly the second part of it after randomisation.

335
00:39:14,270 --> 00:39:17,390
If we lose a lot of people if this.

336
00:39:20,550 --> 00:39:24,090
A great loss of follow up after randomisation.

337
00:39:24,750 --> 00:39:32,520
It might have a very important impact in how we interpret and analyse our findings.

338
00:39:34,290 --> 00:39:39,650
So for starters, it could be that this type of intervention cannot be applied to everyone.

339
00:39:39,870 --> 00:39:44,150
It could talk about it could be a proxy for the quality of the trial itself.

340
00:39:44,160 --> 00:39:53,700
If they lose 50% of their lives, 60% of the ones that were randomised, maybe the overall quality of what they actually find might be put into doubt.

341
00:39:54,090 --> 00:39:56,690
So this is this obviously a very important issue.

342
00:39:56,700 --> 00:40:04,320
And not only that, but it might be that the actual loss to follow up might be linked to something related to the intervention.

343
00:40:04,350 --> 00:40:05,909
So if we have differential attrition,

344
00:40:05,910 --> 00:40:14,580
which I'll go back for a second that we lose more in the intervention group than in control group that might start ringing some alarm bells.

345
00:40:14,580 --> 00:40:18,780
So there might be something wrong or interesting happening there.

346
00:40:21,450 --> 00:40:24,660
In this particular trial, there were 231 included.

347
00:40:24,690 --> 00:40:28,740
200 of them were analysed. So that was 87% of them.

348
00:40:28,950 --> 00:40:38,640
And they do tell us why they lost the other 3127 one were unable to attend so they just couldn't be bothered or

349
00:40:38,970 --> 00:40:44,580
they just didn't even come back to where it's clear from analysis because they were upgraded to business class.

350
00:40:44,730 --> 00:40:51,389
We just mentioned about that and to exclude because they were taking anticoagulants medication so that they

351
00:40:51,390 --> 00:40:58,080
were already likely to have probably likely to have from doses and therefore they were at higher risk.

352
00:40:59,460 --> 00:41:04,440
Okay. Any questions about that, too? Are we happy? Yes, absolutely.

353
00:41:07,910 --> 00:41:11,050
Okay. We'll talk that in a second. Any other witness?

354
00:41:11,660 --> 00:41:15,010
There's just a number of accounts of how to be secure.

355
00:41:15,050 --> 00:41:18,430
Number three shuts down. Okay.

356
00:41:18,690 --> 00:41:27,220
They do? Yeah, they do mention in in the methods that they were able to work out a sample size,

357
00:41:27,230 --> 00:41:31,870
a proper sample size calculation because they didn't have enough pilot studies.

358
00:41:31,880 --> 00:41:35,660
And, in fact, they they say that this can be treated as a pilot study.

359
00:41:36,110 --> 00:41:42,950
And what they had proposed is to keep doing it until they got 100 people recruiting into each other.

360
00:41:43,980 --> 00:41:48,680
I think that's that's why you get those very convenient numbers.

361
00:41:48,680 --> 00:41:58,610
But I think that's that's how they they set up to keep in terms of how many large groups they basically kept on recruiting.

362
00:41:58,640 --> 00:42:01,340
Oh, sorry, 4/100 for the proposed sample size.

363
00:42:04,370 --> 00:42:13,970
I can't remember if they mentioned specifically why one of the main objectives was to determine the prevalence of deep vein thrombosis.

364
00:42:14,870 --> 00:42:22,969
And I'm not sure if that's one of the the reasons why they said, well, with 100, we should be able to at least get get a rough.

365
00:42:22,970 --> 00:42:29,090
I think they do mention that if they if the vein thrombosis, the prevalence was fairly low about 2%,

366
00:42:29,360 --> 00:42:33,830
that would give them fairly tight confidence intervals with 100. And that's why they came up with that number.

367
00:42:34,700 --> 00:42:40,640
Remember that? Yes. That's how the study.

368
00:42:40,670 --> 00:42:45,980
Yep. Is that in or out, though? No.

369
00:42:46,430 --> 00:42:50,360
Conclusive. How? Yes. Well, there was a power calculation.

370
00:42:50,660 --> 00:42:56,730
How about the study? Okay. Yeah.

371
00:42:56,960 --> 00:43:01,700
The power of the study, in a way, tells you how likely you are to find a difference if one exists.

372
00:43:01,970 --> 00:43:10,490
So find an effect, even if it's actually present. If you if your trial is very small and there is an effect, you're probably not likely to find it.

373
00:43:10,590 --> 00:43:14,940
You could try this very, very large, even if the effect is small. Probably likely to find an effect.

374
00:43:15,440 --> 00:43:19,620
And we'll go more into that when we look at how to deal with uncertainty.

375
00:43:19,640 --> 00:43:23,120
So P values and confidence intervals. We'll talk about that in a second.

376
00:43:23,420 --> 00:43:34,340
But going back to how many can we lose if really is actually dependent on the type of condition you're looking at?

377
00:43:34,760 --> 00:43:42,800
So for some things, you want to have pretty much every one. For other things, you would be very, very happy if you have more than 20% followed up.

378
00:43:43,070 --> 00:43:49,460
So it's actually just tying yourself to one number is actually not helpful.

379
00:43:49,730 --> 00:43:54,050
But generally and as a rule of thumb, you can talk of the 5020 rule of thumb.

380
00:43:55,010 --> 00:44:02,780
And that basically means that if you lose about 5% of your individuals in terms of follow up does slightly to to produce very little bias.

381
00:44:03,200 --> 00:44:09,350
If you lose more than 20%, then actually have a strong implication in terms of the validity of the study.

382
00:44:09,710 --> 00:44:15,050
You're probably losing too many generally. Again, this has to be taken in context.

383
00:44:16,100 --> 00:44:24,790
Importantly, this depends, as I mentioned, in the outcome event rate and and also on comparative loss rates in the two groups.

384
00:44:24,800 --> 00:44:30,860
So if we have that the loss of follow up rate exceeds the outcome event.

385
00:44:31,240 --> 00:44:41,300
It's that we have a lot more that were lost to follow up than the number of people that benefited or or that did have or some believe vein thrombosis.

386
00:44:41,500 --> 00:44:47,600
However important because all of those that we lost may have had be vein thrombosis and we don't know about them.

387
00:44:48,140 --> 00:44:54,500
And probably more importantly, if the the attrition, the loss of follow up is differential,

388
00:44:54,500 --> 00:44:59,830
we have more in one group than another that might have, again, implications as to whether the treatment itself,

389
00:44:59,960 --> 00:45:09,770
the effect that we find is is is is real or it might be just being produced by the fact that we have differential follow ups.

390
00:45:09,980 --> 00:45:17,330
So, for example, let's say that we have a medication that is very that reduces people to feel very nauseous and people stop taking it.

391
00:45:17,510 --> 00:45:20,899
Therefore, you lose them from your analysis maybe,

392
00:45:20,900 --> 00:45:24,440
and you find that the medication is brilliant because you have you haven't actually taken

393
00:45:24,440 --> 00:45:28,250
into account of other people that actually couldn't take the medication in the first place.

394
00:45:28,730 --> 00:45:32,300
Okay. So that's might be an extreme example of differential follow up, right?

395
00:45:35,230 --> 00:45:43,780
And in this particular case, they do mention that there were losses, but these losses were equally distributed in the stock group.

396
00:45:43,990 --> 00:45:47,830
They lost six men and women, roughly 15 in the non-smoking group.

397
00:45:47,830 --> 00:45:53,320
So men and women with 16, do they have similar characteristics to those that were included?

398
00:45:54,790 --> 00:46:02,379
Unfortunately, they didn't actually provide information about them and say it could have been just a sentence in the article saying the people

399
00:46:02,380 --> 00:46:11,260
that were lost to follow up were roughly of the same age and and the medications taken were roughly the same as the other group.

400
00:46:11,290 --> 00:46:14,320
They don't provide that type of information, but it would have been important to know about them.

401
00:46:15,560 --> 00:46:20,800
And related to this is what I'm sure you've all heard of, is intention to treat principle,

402
00:46:21,190 --> 00:46:28,630
which maintains that once a patient is randomised to one of the groups they should be analysed in the group they were randomised to.

403
00:46:28,900 --> 00:46:39,850
So what Mike just did in terms of switching the wine gums and with pastilles would have still kept them into the gums initially.

404
00:46:40,270 --> 00:46:45,820
I think, okay, he should have he would have still be kept in the commonest group, okay.

405
00:46:46,000 --> 00:46:55,270
Because otherwise randomisation allows you randomisation means that if it's done properly, you get equivalent groups.

406
00:46:55,540 --> 00:47:03,280
But if you have some kind of crossover because you don't like gums or because you don't like stockings, let's say,

407
00:47:03,550 --> 00:47:10,810
then your findings might represent a completely different thing than the actually the actual effect of the intervention.

408
00:47:11,050 --> 00:47:21,580
Okay. One exception to the intention to treat principle is that if the patient is found to have on blind reassessment to have been ineligible,

409
00:47:22,660 --> 00:47:24,580
they should be excluded. And that's, for example,

410
00:47:24,580 --> 00:47:30,100
what was mentioned earlier on in terms of the people that were flying in business class were upgraded

411
00:47:30,100 --> 00:47:37,750
to business class because they were not they shouldn't be included as part of your population group.

412
00:47:37,990 --> 00:47:41,270
They were part of the group that was part one of the exclusion criteria. Okay.

413
00:47:41,350 --> 00:47:50,270
After after reassessment. So how do you assess the outcome for this trial or general belief?

414
00:47:51,910 --> 00:47:55,690
Okay, that that's a very good question.

415
00:47:55,690 --> 00:48:04,059
Intention to treat is very much seen as a gold gold standard way of doing the analysis.

416
00:48:04,060 --> 00:48:09,250
And it might be very difficult to implement depending on what type of outcome you have.

417
00:48:09,610 --> 00:48:17,259
If you have binary outcomes, where there's a very clear definition as to what with a positive outcome or what's a negative outcome,

418
00:48:17,260 --> 00:48:21,850
let's say the vein thrombosis, if you were able to.

419
00:48:22,900 --> 00:48:26,559
It's a simple scenario if you collected outcomes from everyone,

420
00:48:26,560 --> 00:48:33,160
but you just have things like crossovers and people just not doing what they were told, then that's fine.

421
00:48:33,160 --> 00:48:37,580
You just can keep them in the same group if you have not only people crossing over.

422
00:48:37,580 --> 00:48:45,070
We're also missing values. For example, you don't know if they had the brain thrombosis or if they didn't have thrombosis.

423
00:48:45,310 --> 00:48:51,250
You need to make some kind of decision as to what type of value you're going to give to that particular outcome,

424
00:48:51,490 --> 00:48:56,920
and then try to analyse with that assumption and without that assumption to see if you

425
00:48:57,190 --> 00:49:02,800
are actually increasing the or producing some kind of bias in your in your analysis.

426
00:49:03,070 --> 00:49:08,020
But if you looking at other types of outcomes like continuous data and it becomes slightly more complicated,

427
00:49:08,020 --> 00:49:19,510
you might need to use more sort of different statistical approaches to impute that missing information.

428
00:49:20,860 --> 00:49:28,840
Okay. Finally measurement, the last part of a run when we do measurements.

429
00:49:29,440 --> 00:49:37,840
What we're interested in is finding out whether the outcomes that we get are actually a true measure of the effect, if there is any.

430
00:49:38,290 --> 00:49:38,889
And for that,

431
00:49:38,890 --> 00:49:44,830
we're looking at things like whether the patients and clinicians were applying that or whether the measurements objective or not objective.

432
00:49:46,090 --> 00:49:53,560
So in this particular study, if you looked at your paper, it says that blood was taken from all participants before travel.

433
00:49:54,530 --> 00:49:58,750
Do they talk about the timer? Can anyone explain that to me?

434
00:50:00,610 --> 00:50:09,620
No. It's just one of the focus that comes out of what you just talked about, the process.

435
00:50:09,800 --> 00:50:16,690
Okay. And that that should be a fairly objective measure of the brain in the patients.

436
00:50:17,740 --> 00:50:22,750
Okay. And the other type of of outcome or measurement of the outcome that they use was ultrasonography.

437
00:50:22,880 --> 00:50:29,230
Okay. And they mentioned that all patients had ultrasonography once before travel, and then afterwards 30 of them had to.

438
00:50:30,520 --> 00:50:36,000
When you consider ultrasonography to be objective as well. Yeah.

439
00:50:36,250 --> 00:50:41,860
So whoever is doing the assessment, that's the comparable thing.

440
00:50:43,730 --> 00:50:48,590
Temptation. Okay. The skill of whoever is doing the assessment.

441
00:50:48,800 --> 00:50:57,050
Okay, so this so final, you see, we say we find that the outcome of all was measured using ultrasonography 24,

442
00:50:57,170 --> 00:51:09,340
48 hours after the return of the flight. Now relating ultrasonography to to the assessment of the outcome,

443
00:51:09,350 --> 00:51:20,149
here we have this one thing that we want to ask is whether blinding was done and when when for example,

444
00:51:20,150 --> 00:51:23,420
we read this was a randomised double blind study.

445
00:51:24,650 --> 00:51:34,010
Was that mean? So here we have an example where the we have the author being double blind versus single blinded.

446
00:51:34,250 --> 00:51:39,140
Okay. So double blinded might mean nothing at all.

447
00:51:39,650 --> 00:51:42,270
Okay. We need to identify who was actually blind.

448
00:51:42,380 --> 00:51:47,630
Were the participants blinded with the investigators planet with the outcomes of blind or the analyst blinded?

449
00:51:50,010 --> 00:51:59,389
That makes sense. Okay. Importantly, particularly for this type of study, we might need to identify if the person that was blind,

450
00:51:59,390 --> 00:52:04,460
what the outcome assesses, if the outcome was not objective.

451
00:52:05,270 --> 00:52:08,009
Why? Because in the worst case scenario,

452
00:52:08,010 --> 00:52:15,830
if the if the whoever is doing the assessment for that particular outcome has a belief that stockings work or they don't work,

453
00:52:16,580 --> 00:52:23,270
yeah, they might, for whatever reason, find more deep vein thrombosis in one group than another group.

454
00:52:24,470 --> 00:52:29,960
Okay. And that's why it's crucial to identify when reading an article who was blind and who wasn't lining.

455
00:52:34,030 --> 00:52:41,440
Do they mention who was blinded? I'll just want to get this from ultrasonography as we're blinded.

456
00:52:45,410 --> 00:52:49,850
They give this, most passengers remove their stockings on completion of the journey,

457
00:52:51,230 --> 00:52:55,340
and they don't say the nurse removes the stockings of those passengers who have continued to wear them.

458
00:52:56,960 --> 00:53:00,620
A further duplex combination was undertaken with a technician unaware of the group to

459
00:53:00,620 --> 00:53:04,220
which the volunteer has been randomised to generally want to take your stockings off.

460
00:53:09,090 --> 00:53:14,620
I'm sure. How did you find them?

461
00:53:15,490 --> 00:53:24,670
On pins and needles. Okay. So if someone were to analyse, to look at your legs right now, would they be able to tell if.

462
00:53:28,570 --> 00:53:36,900
If you were wearing stockings or not. How you reckon?

463
00:53:41,270 --> 00:53:46,799
You could describe. Yeah.

464
00:53:46,800 --> 00:53:50,370
That sounds, you know, similar there.

465
00:53:51,240 --> 00:53:54,430
Oh, yes. Okay. Thank you very much.

466
00:53:55,410 --> 00:53:58,440
Thank God. Thank you. Okay.

467
00:53:58,890 --> 00:54:08,340
So assessment was done in some cases when people were remove their stockings.

468
00:54:09,120 --> 00:54:13,310
Not that long ago, not many of them.

469
00:54:13,320 --> 00:54:19,780
Hopefully they don't specify how many of them. Okay. They don't mention any of this apart from that.

470
00:54:19,860 --> 00:54:27,630
Just a remark in passing. What do you think? It is important to continue awareness.

471
00:54:29,960 --> 00:54:33,620
Is there a benefit? Potentially maybe an issue.

472
00:54:34,190 --> 00:54:41,600
So it's not only the flight, but wearing them after the flight. We don't know the exact examination subject to the question.

473
00:54:42,110 --> 00:54:45,329
Which one do they. The ultrasound is subjective.

474
00:54:45,330 --> 00:54:48,370
The best is subjective. Okay.

475
00:54:50,460 --> 00:54:54,070
So if it's if it's objective, then we're not worried about the blinding.

476
00:54:54,590 --> 00:55:00,110
So the more objective an outcome is, the less we worry about the blinding, the less objective.

477
00:55:00,350 --> 00:55:09,260
So sample if it was a questionnaire that they had to fill in and it was a completely subjective assessment,

478
00:55:09,560 --> 00:55:13,700
in a way, we might be more worried if there was no blinding done. Okay.

479
00:55:16,840 --> 00:55:23,560
I'm going to go to talk finally about the results and when we have the results.

480
00:55:23,740 --> 00:55:28,060
We've done our analysis, we've done around, but we've decided that this is worthwhile looking into.

481
00:55:28,360 --> 00:55:32,490
Then there's the three basic when we look at the results of three basic explanations.

482
00:55:32,500 --> 00:55:38,800
One is that the findings that we have could be explained through placebo, through chance, or through a real effect placebo.

483
00:55:38,810 --> 00:55:44,170
So it's a very important explanation.

484
00:55:44,200 --> 00:55:51,400
Here we have a trial in patients with chronic severe itching of two types of medications we have.

485
00:55:51,760 --> 00:56:00,220
So perhaps adding and trimethoprim a pressing and we have a no treatment group and we have a score of an eating score.

486
00:56:01,030 --> 00:56:02,679
Okay. So looking at this,

487
00:56:02,680 --> 00:56:12,500
would you try one of those of these medications if you had a chronic severe itching or you would recommend to someone some say yes.

488
00:56:12,520 --> 00:56:19,330
I'm saying now, if I were to look a bit further and I like to present this, that was actually the same.

489
00:56:21,800 --> 00:56:32,740
If I am placebo group, then what do you would you do? Okay, so placebo is actually a very important explanatory intervention.

490
00:56:33,100 --> 00:56:38,990
Okay. It shouldn't be taken lightly. Chance.

491
00:56:39,440 --> 00:56:43,700
What about chance? And it's related to what John was mentioning in terms of sample sites, etc., etc.

492
00:56:44,060 --> 00:56:48,080
What about if it's just chance that we find whatever we found with the effect that we found?

493
00:56:48,110 --> 00:56:54,620
Well, there's two ways of trying to address that. That particular difference was done by chance, was happened to be there by chance.

494
00:56:54,950 --> 00:56:59,930
And the two ways that we deal with it is to use hypothesis testing basically p values,

495
00:57:00,350 --> 00:57:04,790
an estimation which is confidence intervals and many times they're given in conjunction.

496
00:57:04,790 --> 00:57:14,270
So you give both a p value and a confidence interval. A pivot is just a way of estimating how likely it is that the null hypothesis is true.

497
00:57:14,510 --> 00:57:22,160
So if the p value is very small and we would say the null hypothesis unlikely to be true generally a null hypothesis being there is no effect.

498
00:57:22,610 --> 00:57:27,770
So you have a p value that's very, very small. We say, okay, there's some evidence that there is an effect.

499
00:57:28,190 --> 00:57:33,860
The intervention in this case, stockings appear to work. On the second hand, we can have confidence intervals.

500
00:57:34,520 --> 00:57:44,570
When confidence intervals, well, they do very, very, very simple terms in estimates of range of values where the true effect probably will lie.

501
00:57:44,930 --> 00:57:51,020
Okay. So it gives you a number of a range of possible numbers where the effect is likely to be.

502
00:57:51,290 --> 00:57:58,250
And in the case of that, what we're looking at is a reduction in deep vein thrombosis.

503
00:57:58,460 --> 00:58:03,230
So how how are these two things integrated and presented, sorry, in this trial?

504
00:58:03,680 --> 00:58:09,469
So in the case of looking at hypothesis testing, they looked at incidence of big vein thrombosis.

505
00:58:09,470 --> 00:58:11,230
In the two groups, we have the stalking groups.

506
00:58:11,230 --> 00:58:18,350
So they have none of them have deep vein thrombosis, non symptomatic bleeding from doses in the no stocking group we have 12%.

507
00:58:18,950 --> 00:58:23,330
So that tells us there's a risk difference difference between the known stockings and the stockings.

508
00:58:24,080 --> 00:58:28,790
So the stock. Yeah. They're not talking about stockings of 12% that has associated a p value.

509
00:58:28,790 --> 00:58:38,060
That P value basically means if there was no effect, the chances of us seeing this difference is one in a thousand.

510
00:58:39,020 --> 00:58:42,470
Okay. If there's no effect, the chance of seeing this is one in a thousand.

511
00:58:42,770 --> 00:58:46,190
Therefore, it's very likely there is an effect. That's what that means.

512
00:58:47,660 --> 00:58:54,500
What about confidence intervals? Again, we have the values in the stock in group zero, non-smoking group 12%.

513
00:58:54,650 --> 00:58:56,360
The risk difference is again, 12%.

514
00:58:56,390 --> 00:59:05,330
We can calculate a confidence interval, interval or a range of values where the true difference should lie, and it goes from about 6% to 20%.

515
00:59:05,750 --> 00:59:12,440
That should be the difference that compression stockings probably would give you would give you an.

516
00:59:15,970 --> 00:59:22,960
What does that mean? Well, because we're actually interested in comparing this to what would happen if there was no effect.

517
00:59:23,590 --> 00:59:31,550
If there was no effect, then we would expect that the range of values that are possible to contain zero.

518
00:59:32,530 --> 00:59:39,250
Okay, zero would mean that it's equally likely that this could be explained by chance.

519
00:59:39,910 --> 00:59:45,520
And therefore, if series containing the confidence interval will say this is probably not statistically significant,

520
00:59:45,520 --> 00:59:52,149
just that it's just as explainable to say that whatever you found in this particular test,

521
00:59:52,150 --> 00:59:55,690
in this particular study could be due to just random variation.

522
00:59:57,220 --> 01:00:00,850
So we decided there is potentially a real effect.

523
01:00:01,960 --> 01:00:10,180
Okay. Given the two things that I just show you, the p value of one in a thousand and a confidence interval that does not obtain,

524
01:00:10,990 --> 01:00:13,090
does not contain C or there appears to be a real effect.

525
01:00:13,510 --> 01:00:18,190
So the next question I want to ask is now, would you consider wearing stockings on a long haul flight?

526
01:00:18,790 --> 01:00:23,240
How many of you would? No.

527
01:00:23,270 --> 01:00:29,240
No, many. You're still not completely convinced. If I were to tell you that there was a systematic review.

528
01:00:29,990 --> 01:00:35,510
Okay, now there's a systematic review. I looked at a lot of it from the ten randomised controlled trials.

529
01:00:36,320 --> 01:00:42,650
They did a critical appraisal of them. Nine of them looked at stockings in both legs versus not wearing then.

530
01:00:43,340 --> 01:00:48,010
And they actually there's a lot of extra information of these trials.

531
01:00:48,020 --> 01:00:53,719
Some of them were in low and medium risk, all of the others were in high risk participants getting away with this particular problem

532
01:00:53,720 --> 01:00:58,700
of these are for this particular other particular trial they were all over 50 with no,

533
01:00:59,810 --> 01:01:02,210
they were not at high risk for they were medium or low risk.

534
01:01:02,570 --> 01:01:12,170
So some information on high risk participants and the results were that 50 out of 2637 participants had symptomless DVT,

535
01:01:12,410 --> 01:01:16,250
three of them with stockings, and 47 with not with stockings.

536
01:01:17,720 --> 01:01:21,740
There were no that no pulmonary emboli or symptomatic DVT reported.

537
01:01:22,520 --> 01:01:29,390
Wearing stockings did have a significant impact on reducing oedema and no significant adverse events were reported.

538
01:01:30,040 --> 01:01:34,040
I mean, if you were to look at the forest plot, this is the type of information that you would get.

539
01:01:34,250 --> 01:01:38,000
We'll talk more about for plus later on. This is not really part of the main focus.

540
01:01:38,000 --> 01:01:41,090
They ask again, how many of you will consider wearing stockings?

541
01:01:42,920 --> 01:01:47,719
Half human. Okay. Thank you very much.

542
01:01:47,720 --> 01:01:48,980
Now go back to the small.