1
00:00:00,060 --> 00:00:10,410
I'm going to try and just give you an idea of the story behind this, because it's been a long running story,

2
00:00:10,410 --> 00:00:18,840
certainly sort of in epidemiology, whether actually moderate alcohol consumption is actually beneficial.

3
00:00:19,260 --> 00:00:24,339
Okay. So this started off with something known as the kind of French paradox.

4
00:00:24,340 --> 00:00:29,639
So this is kind of information from an ecological study.

5
00:00:29,640 --> 00:00:38,880
And they were looking at the association between heart disease, coronary heart disease and cholesterol, and they noticed that, you know,

6
00:00:39,140 --> 00:00:49,709
for most countries, you know, there was a kind of linear relationship Finland have had sort of at that time when this was done.

7
00:00:49,710 --> 00:00:53,010
So this was done in about sort of 1980 or something like that.

8
00:00:53,280 --> 00:00:59,490
Finland had a really sort of bad sort of cardiovascular disease sort of profile.

9
00:00:59,670 --> 00:01:02,160
And they've they've since sort of modified that.

10
00:01:02,580 --> 00:01:13,590
But essentially, France, even though they had a high cholesterol intake, they actually had lower levels of coronary heart disease.

11
00:01:14,400 --> 00:01:20,040
And then a kind of follow up to this was then. So people are saying, well, what what was the explanation for this?

12
00:01:21,330 --> 00:01:31,020
And they did a plot of the per capita alcohol consumption against sort of death rates.

13
00:01:31,560 --> 00:01:36,810
And you could see that Finland were sort of down at the other end.

14
00:01:36,810 --> 00:01:44,970
And here he was, France, with low death rates, but still, you know, higher levels of alcohol consumption.

15
00:01:44,980 --> 00:01:49,139
So this could got people thinking, well, what what are the reasons behind this?

16
00:01:49,140 --> 00:01:56,490
Because it's kind of ecological data. So this kind of spawned lots of sort of hypotheses.

17
00:01:56,970 --> 00:02:01,620
And so, you know, basically based on looking at this data saying, well,

18
00:02:01,620 --> 00:02:08,759
this positive relationship with saturated fat situation in France is paradoxical simply because, you know,

19
00:02:08,760 --> 00:02:17,520
they had this kind of low mortality rate and they thought it might be due to high wine consumption because they had some,

20
00:02:17,520 --> 00:02:23,040
you know, various sort of things in wine that people believe are kind of protective.

21
00:02:23,760 --> 00:02:33,510
And I'll say more about that a bit bit later. But also there are other things, you know, in terms of hypotheses.

22
00:02:33,530 --> 00:02:38,750
Some people thought that it might be a time lag because they actually had to look at this.

23
00:02:39,020 --> 00:02:44,809
So that information and they were thinking that actually in terms of cholesterol levels,

24
00:02:44,810 --> 00:02:48,709
maybe it was taking time to kick in before you started seeing the benefits.

25
00:02:48,710 --> 00:02:53,780
So people started actually having saturated fat in France later than some other countries.

26
00:02:54,200 --> 00:02:56,089
And you see a similar thing with smoking.

27
00:02:56,090 --> 00:03:04,880
You don't see the death rate until to the 20 years later because people have to be smoking for a while before you start seeing the harm that it does.

28
00:03:06,230 --> 00:03:09,650
And then they thought, could it be diet as well?

29
00:03:09,650 --> 00:03:13,730
So the differences in diet in terms of sort of things like folate.

30
00:03:14,340 --> 00:03:19,940
Yeah, nuts and then sort of things like polyunsaturated fats as well.

31
00:03:20,360 --> 00:03:26,780
And then they also thought, well, the French, you know, they drink with males.

32
00:03:26,850 --> 00:03:28,429
So it's the patterns of drinking.

33
00:03:28,430 --> 00:03:38,240
Is that the kind of thing that might be going on as opposed to the Brits who go to the bingeing at weekends or something on an empty stomach?

34
00:03:39,440 --> 00:03:45,770
So it's that kind of thing. So that was one of the things that they thought might be of interest.

35
00:03:47,090 --> 00:03:52,370
So this led to several sort of cohort studies being done.

36
00:03:52,760 --> 00:03:59,390
And one of the things that they found, this is the data from sort of five cohort studies.

37
00:04:00,320 --> 00:04:02,630
They were the largest done at that time.

38
00:04:03,230 --> 00:04:14,570
And actually, you can see this kind of j-shaped association in all of them where actually the lower levels of alcohol consumption,

39
00:04:15,020 --> 00:04:21,710
there seemed to be some benefit. And then that starts going the way of people to the drink more and more.

40
00:04:22,220 --> 00:04:25,550
And these it is mostly done in sort of male cohorts.

41
00:04:26,360 --> 00:04:30,700
But there's a similar it's the same sort of thing with women.

42
00:04:30,710 --> 00:04:38,660
So essentially this kind of j-shaped association was always sort of related to sort of,

43
00:04:38,660 --> 00:04:43,280
you know, at the lower levels of alcohol consumption, you saw some kind of benefit.

44
00:04:43,280 --> 00:04:49,880
And then as people sort of think more and more, then you started seeing harms.

45
00:04:51,170 --> 00:04:56,270
So that was kind of, you know, evidence and kind of the analytical epidemiologic.

46
00:04:56,480 --> 00:05:04,160
But there's quite a few sort of biases in there. I'm not expecting you to read all of this, essentially.

47
00:05:05,400 --> 00:05:12,390
The kind of bias is things like confounding by the type of drink or the pattern of drinking that we mentioned,

48
00:05:12,780 --> 00:05:18,720
but things like socioeconomic status and lifestyle, those kind of things can sort of create issues.

49
00:05:19,170 --> 00:05:23,550
But also the choice of reference group makes a difference as well.

50
00:05:23,910 --> 00:05:30,150
So if you choose non-drinkers, you might see some sort of differences.

51
00:05:31,530 --> 00:05:39,660
Also things like reverse causality in terms of once people get they'll then actually they change their behaviours,

52
00:05:39,780 --> 00:05:43,049
they stop drinking or smoking, those kind of things.

53
00:05:43,050 --> 00:05:49,590
So those are the kind of things that could be going on. Some of the studies with case control.

54
00:05:49,590 --> 00:05:52,590
So there could be sort of recall or misclassification.

55
00:05:52,590 --> 00:05:55,920
You ask anyone to, you know, what they drink or what they eat.

56
00:05:55,980 --> 00:06:03,420
And just by sort of recalling, they tend to underestimate or under-report.

57
00:06:04,290 --> 00:06:08,549
There's also within person variation. So over time people will change.

58
00:06:08,550 --> 00:06:12,240
As you get older, you start sort of drinking patterns change.

59
00:06:12,750 --> 00:06:14,610
You know, maybe when you're younger, you drink more.

60
00:06:15,300 --> 00:06:22,830
And then, of course, the study design, also the publication bias that could maybe affect what's going on.

61
00:06:23,970 --> 00:06:29,670
So this was sort of a review paper that a colleague and I did several years ago now.

62
00:06:30,780 --> 00:06:39,120
But this is just to give you an idea of what the impact of the choice of the reference category is.

63
00:06:39,450 --> 00:06:47,250
So essentially this dark line is choosing individuals that are non-drinkers.

64
00:06:47,580 --> 00:06:52,890
So you see a kind of more accentuated sort of association.

65
00:06:53,280 --> 00:06:58,520
And then this dotted line is actually when you change the reference category.

66
00:06:58,530 --> 00:07:01,800
So what we did in this paper, we got hold of the individual data.

67
00:07:02,460 --> 00:07:09,240
We had a look at what they had originally when they used a reference category of non drinker and then we

68
00:07:09,240 --> 00:07:14,730
changed it to sort of like drinking and see what you would get if you sort of plotted that out again.

69
00:07:15,240 --> 00:07:28,290
So the difference that you see here is actually on how beneficial you think it might be for moderate drinking because,

70
00:07:28,530 --> 00:07:36,270
you know, you've got an idea that's in a different level for, you know, once you use a different reference category.

71
00:07:36,780 --> 00:07:41,580
And then here is showing, well, where does it become harmful?

72
00:07:41,700 --> 00:07:45,510
Well, it's actually at a different level.

73
00:07:46,950 --> 00:07:50,390
If you use one reference category as opposed to another.

74
00:07:50,460 --> 00:07:54,570
So trying to change is your kind of overall interpretation, you see.

75
00:07:54,690 --> 00:08:02,849
I mean, so this led to sort of people thinking, well, we need to have some kind of comprehensive meta analysis,

76
00:08:02,850 --> 00:08:10,620
trying to have a look at what's been done and try and really understand what's happening.

77
00:08:11,400 --> 00:08:14,820
So this paper was published in The Lancet,

78
00:08:15,690 --> 00:08:26,159
and what they did was they got 83 large perspective studies and actually got them all together to the

79
00:08:26,160 --> 00:08:32,550
information and then tried to see if they could come up with a clearer picture of what was going on.

80
00:08:34,080 --> 00:08:42,719
So the methodology of that and because this is a kind of medical sort of stats, cause I thought I'd sort of mention give you a bit of detail on this.

81
00:08:42,720 --> 00:08:50,730
So they focussed on current drinkers that three main reasons what they wanted to do is think about alcohol guidelines.

82
00:08:50,730 --> 00:08:55,260
So what can you actually say for people who are current drinkers based on that information,

83
00:08:56,580 --> 00:09:02,580
but also to try and limit some of those biases that I sort of mentioned on the previous slide.

84
00:09:02,590 --> 00:09:07,170
So things like reverse causality, residual confounding and so on.

85
00:09:08,010 --> 00:09:14,940
And also there seems to be some evidence that never drinkers kind of systematically differ from drinkers somehow.

86
00:09:15,930 --> 00:09:21,480
So that's why it's not usually a good sort of reference category to use.

87
00:09:22,080 --> 00:09:27,600
And then they tried to harmonise all the alcohol consumption across the contributing

88
00:09:27,600 --> 00:09:33,930
studies using a conversion rate of one unit being about eight grams of pure alcohol.

89
00:09:34,650 --> 00:09:38,340
And then they converted that to a kind of standard scale of grams per week.

90
00:09:38,790 --> 00:09:42,370
Then they tried to make sure they looked at sort of various confounders.

91
00:09:42,390 --> 00:09:45,030
So, you know, things like smoking,

92
00:09:45,900 --> 00:09:53,969
history of diabetes and obviously age and they all they wanted to correct for measurement error and within person variation,

93
00:09:53,970 --> 00:10:04,530
which I sort of alluded to. So within these kind of cohorts, they had repeat measures on, you know, subset of individuals.

94
00:10:05,050 --> 00:10:12,040
He showed how peoples of alcohol behaviour or drinking habits had changed over time.

95
00:10:12,040 --> 00:10:19,690
So they tried to sort of account for that change over time by using the information on these kind of cereal assessments.

96
00:10:21,580 --> 00:10:25,959
And they used a kind of method known as regression calibration.

97
00:10:25,960 --> 00:10:33,850
For those of you that are not familiar, this is kind of like a standard way of actually adjusting for measurement error,

98
00:10:33,850 --> 00:10:39,040
assuming that it's sort of a sort of random regression.

99
00:10:39,040 --> 00:10:42,940
Calibration is a kind of standard approach for dealing with that.

100
00:10:45,280 --> 00:10:49,960
And then they also looked at patterns of drinking as well.

101
00:10:50,020 --> 00:10:57,940
So they had to wine, beer and spirits, you know, frequency consumption and then episodic heavy drinking.

102
00:10:59,260 --> 00:11:02,890
So here on the left is a first.

103
00:11:03,640 --> 00:11:08,620
So this is looking at the association between all cause mortality from all the studies.

104
00:11:08,890 --> 00:11:17,440
So the usual alcohol consumption means that it's been corrected for within person variation rather than to the baseline alcohol.

105
00:11:18,700 --> 00:11:22,779
And you can see that for all cause mortality.

106
00:11:22,780 --> 00:11:30,910
I mean, it's sort of fairly flat. So the low levels and then starts picking up afterwards after about 150 or something.

107
00:11:31,210 --> 00:11:36,400
Whereas for cardiovascular disease, you still got that at the lower levels.

108
00:11:36,400 --> 00:11:44,170
You've got that inverse association that then sort of goes up afterwards.

109
00:11:44,290 --> 00:11:49,769
So in that kind of shape that they try to look at different subtypes.

110
00:11:49,770 --> 00:12:01,780
So they looked at sort of stroke and my coronary disease, excluding my heart failure and deaths and other types of sort of cardiovascular disease.

111
00:12:02,410 --> 00:12:10,149
So you can see for stroke, it looks like there is a kind of strong positive association.

112
00:12:10,150 --> 00:12:14,410
So, you know, drinking more seems to be harmful for stroke.

113
00:12:14,740 --> 00:12:26,290
Whereas for me, well, it looks like it's an inverse association there, whereas for coronary disease excluding them, I sort of fairly unclear.

114
00:12:27,130 --> 00:12:30,340
Heart failure looks like a kind of positive trend.

115
00:12:31,290 --> 00:12:39,180
And death to the cardiovascular disease is similar and they summarise the results here.

116
00:12:40,800 --> 00:12:49,770
So for all stroke, they told us about a 14% and they split into nonfatal and fatal and haemorrhagic and ischaemic.

117
00:12:50,160 --> 00:12:59,910
And sort of with that sort of subtypes. You can see those are the ones that are all sort of positively associated with alcohol consumption.

118
00:13:00,180 --> 00:13:12,060
Whereas for me, it looks like there's an inverse association and then sort of mixed results for some of these other things down there.

119
00:13:12,930 --> 00:13:21,210
And generally not too much heterogeneity, probably a little bit high for this one, but generally,

120
00:13:21,660 --> 00:13:27,420
you know, reasonably consistent results between the studies from what they reported.

121
00:13:27,870 --> 00:13:36,240
They also tried to sort of get a good public health message across that it's time to have a look at the estimated years of life lost.

122
00:13:37,110 --> 00:13:46,739
You know, if you a regular sort of low alcohol drinker, 100 to 200 grams, you know,

123
00:13:46,740 --> 00:13:53,760
sort of moderate and then sort of high amount to try and have a look at what we could do.

124
00:13:53,760 --> 00:14:00,590
And you could actually they plotted this of the years of life lost males and females and, you know,

125
00:14:00,600 --> 00:14:11,310
seeing very similar sorts of shapes for both of those two and, you know, similar amounts of years lost for drinking.

126
00:14:11,580 --> 00:14:14,880
Yeah, moderately or heavily.

127
00:14:15,240 --> 00:14:19,380
And they did try and actually sort of do a trial.

128
00:14:19,920 --> 00:14:24,720
So there was there was there was a trial that was going to be run in the US.

129
00:14:25,080 --> 00:14:30,899
At one stage they tried to get 7800 participants.

130
00:14:30,900 --> 00:14:35,220
They were going to randomise them to one drink a day to none for six years.

131
00:14:36,390 --> 00:14:45,660
And the primary outcome was going to be as those who got it, cardiovascular disease and secondary was going to be kind of vascular death and diabetes.

132
00:14:46,980 --> 00:14:53,600
But you won't be surprised to know that actually they thought, well, that's not going to work.

133
00:14:53,610 --> 00:15:00,660
And they pulled the plug on that. At the time when we were looking at, we thought, well, this is going to be sort of quite ambitious.

134
00:15:01,200 --> 00:15:09,240
And yeah, so they ended up not doing that because of, you know, the flaws in observational evidence.

135
00:15:10,020 --> 00:15:18,840
You know, there's not really much chance of getting a trial. Could genetics give us a kind of clue on what to do?

136
00:15:19,800 --> 00:15:26,610
There are a couple of genetic variants that actually sort of effect alcohol metabolism.

137
00:15:27,690 --> 00:15:31,590
So they sort of got very similar name. So it can be quite confusing.

138
00:15:31,800 --> 00:15:44,080
So 88 is alcohol, dehydrogenase and what that does, it converts to as to how old I am actually, you know, it kind of increases that.

139
00:15:44,100 --> 00:15:50,160
So basically when you feel hung over and horrible and nauseous, that that's the reason.

140
00:15:51,450 --> 00:16:01,290
And then the other one, a LDH actually sort of tunes asan aldehyde to acetate, but it slows down that process.

141
00:16:01,500 --> 00:16:06,030
It doesn't quite have a stronger effect on you in terms of how you feel,

142
00:16:06,300 --> 00:16:13,980
but essentially both of them lead to sort of can to sort of nausea and various sort of things

143
00:16:13,980 --> 00:16:17,370
and headaches and all those kind of things that you feel when you're kind of hung over.

144
00:16:18,060 --> 00:16:27,510
Essentially, the general feeling is you can use these to conduct what's known as the Mendelian randomisation study and

145
00:16:27,510 --> 00:16:36,810
essentially a study where if you think of using a genetic variant that you kind of assigned to genotype at birth,

146
00:16:37,830 --> 00:16:44,910
that can give you sort of an idea if you've got a particular genotype, whether you are tolerant for alcohol or not.

147
00:16:45,540 --> 00:16:53,310
And you could do a similar study, which is essentially what they were trying to do, randomise it to a drink or not.

148
00:16:53,910 --> 00:16:59,310
And that that would be probably the the best way of trying to sort of do that.

149
00:16:59,850 --> 00:17:04,560
So I've just put the assumptions here in one sort of figure,

150
00:17:05,430 --> 00:17:15,650
but essentially the assumption is all the genetic variant that you choose must be related to that kind of exposure that you're interested in.

151
00:17:15,670 --> 00:17:18,930
So it must be related to alcohol in this case.

152
00:17:19,560 --> 00:17:27,450
So that's known as the relevance assumption, the relationship between alcohol and the outcome of heart disease.

153
00:17:27,450 --> 00:17:30,750
In this case, it must work through.

154
00:17:31,500 --> 00:17:34,920
That pathway not to any of that pathway.

155
00:17:35,760 --> 00:17:41,880
And then genetic variant must not be associated with any sort of confounders.

156
00:17:41,890 --> 00:17:50,220
So, you know, you can kind of check some of these assumptions, but some of them are not.

157
00:17:51,450 --> 00:17:56,490
It can be difficult unless you know a lot about the kind of biology of of the variants.

158
00:17:57,180 --> 00:18:01,710
In this case, there's a lot known about these this particular snake.

159
00:18:02,770 --> 00:18:08,220
It's it's quite a good candidate to do this kind of analysis.

160
00:18:09,390 --> 00:18:17,250
This was a Mendelian randomisation study that looked at sort of individual participant data,

161
00:18:17,280 --> 00:18:21,150
tried to sort of get the information together and see what they could do.

162
00:18:22,650 --> 00:18:26,490
So what they did was they chose this particular snip.

163
00:18:27,150 --> 00:18:31,920
It is non synonymous means it actually effects that kind of protein directly.

164
00:18:32,160 --> 00:18:42,150
So basically, you know, that particular snake controlled what what's going on in this particular gene, the 88 one zombie.

165
00:18:43,110 --> 00:18:48,360
And I showed you that on the previous slide and that one of the genes that

166
00:18:48,360 --> 00:18:53,850
controls alcohol metabolism and they got an international collaboration together.

167
00:18:54,780 --> 00:19:05,040
If you only had measured this and sort of cardiovascular biomarkers and various sort of events in order to try and understand fully what was going on,

168
00:19:05,910 --> 00:19:15,239
they restricted individuals to those of European descent with data on this particular slip and,

169
00:19:15,240 --> 00:19:21,000
you know, age and sex and any of the outcomes of interest and what they did.

170
00:19:21,540 --> 00:19:33,450
I don't know how familiar you are with genetics, but this particular snap ad H-1B has to illegals.

171
00:19:33,900 --> 00:19:38,280
You know, it's the one that is the bad one and the G is the good one.

172
00:19:38,700 --> 00:19:46,260
And you can actually have three gene, two types, a G essentially.

173
00:19:46,560 --> 00:19:51,260
And what they did is they pulled the result.

174
00:19:51,270 --> 00:19:56,130
So if you had any bad illegal which is a.

175
00:19:57,400 --> 00:20:03,850
They combined individuals with the aye aye and the AG and compared them with the drug.

176
00:20:05,020 --> 00:20:12,670
And they quantified the effects of that a little on different alcohol traits and lifestyle things.

177
00:20:13,840 --> 00:20:20,049
And they also had a look to see how related it was to kind of cardiovascular biomarkers.

178
00:20:20,050 --> 00:20:24,130
So trying to sort of put together a picture of how it might sort of work.

179
00:20:25,960 --> 00:20:34,380
And then they evaluated the strokes on I'm sorry, the analysis on coronary heart disease, stroke, and they looked at type two diabetes as well.

180
00:20:35,200 --> 00:20:40,240
They did some long transformations for the continuous variables.

181
00:20:40,990 --> 00:20:46,090
They looked at the kind of mean difference and worked out the kind of percentage difference.

182
00:20:46,780 --> 00:20:53,020
And they looked at the shape of the association for the kind of biomarkers and confounders.

183
00:20:53,410 --> 00:21:04,390
And then they also compared the genetic results with some observational results as well, just to try and see how different they were, essentially.

184
00:21:05,560 --> 00:21:20,200
So here's what they found. That overall, they found that sort of marginal sort of benefit, about 10% for coronary heart disease.

185
00:21:20,530 --> 00:21:24,160
And then they had a look in drinkers and non-drinkers.

186
00:21:25,990 --> 00:21:29,080
You can see it was stronger in drinkers.

187
00:21:29,950 --> 00:21:35,500
And then they had a look on whether it was like moderate or heavy as well.

188
00:21:36,220 --> 00:21:47,980
And that didn't seem to be anything going on there in terms of types of different types of alcohol consumption there.

189
00:21:49,150 --> 00:22:01,870
They also looked at sort of CVD biomarkers, blood pressure, you know, anthropometric indices, markers of inflammation.

190
00:22:01,870 --> 00:22:10,150
So there's quite a lot in the literature about sort of alcohol being, you know, one of the ways that it might work is raising HDL,

191
00:22:10,420 --> 00:22:18,730
good cholesterol, basically, and also with that kind of markers that might be sort of important.

192
00:22:19,390 --> 00:22:26,020
And so that's what they looked at. So they looked at things like sort of interleukin six and C-reactive protein,

193
00:22:26,020 --> 00:22:33,460
which are markers of inflammation and things seem to be going roughly in the right direction for those.

194
00:22:34,480 --> 00:22:37,240
So, you know, lowering blood pressure, I mean, small amounts.

195
00:22:37,390 --> 00:22:46,840
But, you know, essentially things seem to be going in the, you know, expected direction, but were consistent with the results that they were missing.

196
00:22:48,040 --> 00:23:00,369
But there are limitations to this work. I mean, essentially that the snake that I showed you, it's actually what's known as mono morphic,

197
00:23:00,370 --> 00:23:07,360
which means that nobody with the AA in Europeans that saw that walk,

198
00:23:08,230 --> 00:23:15,490
that's why they had to use this that snap, because that is sort of prevalent in Europeans.

199
00:23:16,330 --> 00:23:22,720
But also they lacked power to identify things with markers of coagulation.

200
00:23:23,740 --> 00:23:27,820
So the type two diabetes and some of the kind of combined subtypes of stroke.

201
00:23:28,960 --> 00:23:33,940
And then there also was evidence of Pleiotropic.

202
00:23:33,950 --> 00:23:44,230
So one that I and my previous Mendelian randomisation slide, I said the pathway that works to must be my alcohol, not to match the pathways.

203
00:23:44,680 --> 00:23:51,700
And the fact is there was some association with of this particular genetic variant

204
00:23:51,700 --> 00:23:57,870
with smokers education and a few other things that you wouldn't want to see.

205
00:23:58,630 --> 00:24:03,880
You know, that sort of raises some sort of doubts about this.

206
00:24:04,780 --> 00:24:09,610
Can you do something in another population?

207
00:24:09,700 --> 00:24:18,310
I mean, one of the problems with a lot of research that's going on at the moment, it's kind of very sort of Eurocentric.

208
00:24:18,730 --> 00:24:25,930
And actually, can you use a different population? So these two genes are actually present in East Asians.

209
00:24:26,980 --> 00:24:39,120
This particular variant, the Iris 671, is actually common in the East Asian population and it really does slow down that acetaldehyde breakdown.

210
00:24:39,130 --> 00:24:45,790
So it's it's often known as the kind of Chinese flushing gene or the Asian flushing gene because

211
00:24:45,790 --> 00:24:54,490
actually people who have these two A-levels feel absolutely dreadful after just a sip of alcohol.

212
00:24:55,120 --> 00:24:56,830
So these people tend not.

213
00:24:57,230 --> 00:25:09,770
So it really does reduce alcohol intake because you feel so bad that you don't drink at all due to the variant slightly less important in East Asians,

214
00:25:10,010 --> 00:25:16,000
but it's still more prevalent in East Asians than it is in Europeans.

215
00:25:16,760 --> 00:25:27,410
And you know, it reduces alcohol intake and both of these variants are more prevalent in East Asians and other populations, not just Europeans.

216
00:25:29,390 --> 00:25:35,330
So I'm just going to introduce a study called the China Category Biobank, the eyes that work on,

217
00:25:36,350 --> 00:25:40,760
because that gives us some context to the next bit that we're going to be doing.

218
00:25:41,570 --> 00:25:50,870
This study is sort of coordinated by our department in collaboration with a group in China at Peking University,

219
00:25:51,800 --> 00:25:57,830
and we collaborate with the Chinese CDC to actually collect data.

220
00:25:58,400 --> 00:26:02,540
You know, it's a long running study. And essentially, you know, it's you know,

221
00:26:02,540 --> 00:26:09,110
it's going to go on indefinitely in terms of as long as we can sort of get money to keep the fieldwork going,

222
00:26:09,410 --> 00:26:13,790
but have collected lots of information, baseline survey and so on.

223
00:26:14,780 --> 00:26:21,859
They do regular surveys where we actually repeat usually what we've done at baseline,

224
00:26:21,860 --> 00:26:26,480
but actually add some further enhancements to kind of get new information.

225
00:26:28,430 --> 00:26:33,829
And you know, we've got sort of blood stored on sort of various sort of things,

226
00:26:33,830 --> 00:26:39,770
sort of metabolomics, proteomics and genetics and then some whole genome.

227
00:26:39,770 --> 00:26:49,490
So the sequencing data as well, because we've been able to link this to the kind of national health records in China,

228
00:26:49,790 --> 00:26:55,310
we've been able to sort of follow up people passively and actually get really good sort of complete information.

229
00:26:56,180 --> 00:27:03,260
So there's something like a 98% coverage for the health insurance within China,

230
00:27:04,100 --> 00:27:09,260
and you can actually then follow up and get lots of information on the various diseases and so on.

231
00:27:10,280 --> 00:27:15,410
But in China, men drink and women don't,

232
00:27:15,560 --> 00:27:27,020
or very rarely the meeting alcohol intake in the ten regions here you can see for the main it's sort of moderate sort of in Gansu,

233
00:27:27,740 --> 00:27:32,600
quite high in Sichuan, which is also known for kind of spicy food there.

234
00:27:33,560 --> 00:27:38,360
Whereas, you know, in the women, I mean in Sichuan it's quite high compared to others.

235
00:27:38,360 --> 00:27:43,580
But yeah, it's pretty low, you know, basically similar sort of thing.

236
00:27:44,720 --> 00:27:47,780
Just to remind you, essentially,

237
00:27:48,860 --> 00:27:56,929
these two snips are prevalent in East Asians and essentially the alcohol intake is reduced

238
00:27:56,930 --> 00:28:06,559
substantially with this particular marker and is reduced to a less moderate amount,

239
00:28:06,560 --> 00:28:10,820
but still reduced if you've got this particular variant.

240
00:28:13,970 --> 00:28:21,770
And just to check how sort of good it is inside data, we actually looked at our alcohol data and did a DUI.

241
00:28:21,780 --> 00:28:29,480
So this is known as a manhattan plot. It's posted sort of all the peaks supposed to look like a sort of Manhattan.

242
00:28:29,930 --> 00:28:35,780
But this is long tailed p values on on the y axis.

243
00:28:36,470 --> 00:28:42,550
This is a p value. So ten to the -40 or something.

244
00:28:42,560 --> 00:28:46,070
And this is almost sort of ten to the -200.

245
00:28:46,070 --> 00:28:54,110
So basically, basically, they're definitely they're in the population. They're very strong in in this in our data.

246
00:28:55,310 --> 00:29:00,590
What we did was to try and sort of work out the different combinations of the two snakes.

247
00:29:01,580 --> 00:29:04,790
Rather confusingly, they both got the same sort of alcohol.

248
00:29:04,790 --> 00:29:18,050
So if you if you think of it's always six, seven one, it's got those three genotypes and the other also has the same three genotypes.

249
00:29:19,430 --> 00:29:29,690
You've got these kind of nine different combinations. So what we did was for these was to try and put together the nine different combinations,

250
00:29:29,780 --> 00:29:33,649
but also taking into account the kind of regional differences.

251
00:29:33,650 --> 00:29:36,770
So you saw that we had massive regional differences.

252
00:29:36,770 --> 00:29:48,470
On slide couple of slides back, we took into account the kind of genotype that was, you know, combinations of these two and within the region.

253
00:29:49,100 --> 00:29:56,089
And then what we did was we use that to categorise people's alcohol intake.

254
00:29:56,090 --> 00:30:01,579
So we actually we're able to sort of work out what their predicted alcohol would be for these

255
00:30:01,580 --> 00:30:08,060
particular combinations due to sort of small numbers rather than having ten categories,

256
00:30:08,480 --> 00:30:16,430
we ended up with sort of six that we're grouped into that was defined by area and genotype in mean intake.

257
00:30:16,670 --> 00:30:20,450
Then using those six categories that I showed you on that you were able to

258
00:30:20,450 --> 00:30:28,700
try and work out what how much sort of predicted alcohol intake essentially,

259
00:30:29,060 --> 00:30:35,180
you know, if you were in Category six that worked out about sort of four drinks a day is about sort of 280.

260
00:30:36,260 --> 00:30:40,910
And you know, if you were in category one, it was less than one drink a week.

261
00:30:41,450 --> 00:30:45,500
And then we did the same sort of thing for for women as well.

262
00:30:46,400 --> 00:30:50,389
Because even though women didn't drink much data, some people did drink.

263
00:30:50,390 --> 00:30:53,480
So you could still time sort of categorise those as well.

264
00:30:54,050 --> 00:31:04,129
And then we checked to see what was the relationship with smoking, for example, and this is what we would like to see.

265
00:31:04,130 --> 00:31:13,520
So actually the, the, the kind of pattern exactly the same across the different categories for smoking, for example.

266
00:31:13,520 --> 00:31:19,009
So that would suggest that, you know, it's working through alcohol and not some other sort of pathway.

267
00:31:19,010 --> 00:31:24,980
And I'm only showing you this because it's kind of striking, but we looked at other things as well,

268
00:31:25,010 --> 00:31:32,030
you know, to make sure that it wasn't sort of related to other potential sort of confounders.

269
00:31:32,570 --> 00:31:35,840
So then we tried to have a look at genes,

270
00:31:35,990 --> 00:31:41,900
these genetic categories that predict alcohol intake to actually how does it relate to sort of various sorts of things.

271
00:31:42,170 --> 00:31:48,979
So you can see here that actually there's quite a strong positive association with blood pressure.

272
00:31:48,980 --> 00:31:55,260
So if you remember from the European one, they saw a slightly sort of lowering of blood pressure.

273
00:31:55,260 --> 00:32:05,270
It was very small, whereas this is suggesting that actually, you know, drinking more actually raises your blood pressure by quite a lot.

274
00:32:06,140 --> 00:32:08,450
And, you know, it does raise HDL,

275
00:32:08,450 --> 00:32:17,329
but only by a small amount and gets kind of like a good marker of the liver function that shows whether you've been drinking or not as well.

276
00:32:17,330 --> 00:32:21,889
So again, you can see that it's related to that.

277
00:32:21,890 --> 00:32:25,219
So one of the reasons we were doing this was to try and say, well,

278
00:32:25,220 --> 00:32:31,130
is it relates to things that we know sort of biologically and physiologically that have been shown before.

279
00:32:31,580 --> 00:32:36,380
So here's what we're looking at for alcohol in this scheme stroke.

280
00:32:37,220 --> 00:32:46,430
One of the reasons for showing these side by side was for you saw these kind of j-shaped associations in the kind of prospective cohort.

281
00:32:46,430 --> 00:32:50,660
So you still see that in our data if we look at the observational data.

282
00:32:51,050 --> 00:32:58,850
But when you look at the genetic, it seems to be, you know, clearer in terms of,

283
00:32:58,880 --> 00:33:04,940
you know, that there's that definite sort of harm from, you know, sort of drinking.

284
00:33:05,750 --> 00:33:12,890
Similarly, for haemorrhagic stroke, again, you see the same sort of j-shaped.

285
00:33:13,760 --> 00:33:17,840
At lower levels. But again, for the genetics, you don't see that.

286
00:33:17,850 --> 00:33:25,640
You see it's sort of fairly sort of clear positive association for MRI.

287
00:33:26,750 --> 00:33:31,100
So this was one that was sort of looked like it was beneficial.

288
00:33:32,180 --> 00:33:36,640
It's inconclusive. Some genetic here, I would say.

289
00:33:36,650 --> 00:33:40,810
I mean, you know, the confidence interval sort of overlaps.

290
00:33:40,820 --> 00:33:49,900
One could be something going on. You know, it could be a lack of power, but it doesn't look like there's any benefit.

291
00:33:49,910 --> 00:33:58,670
If you remember, what we saw from that previous meta analysis in Europeans is there's a very strong inverse relationship with MRI.

292
00:34:00,080 --> 00:34:03,950
We didn't see that here. Do you?

293
00:34:03,950 --> 00:34:08,990
The thing was looking it women. Women act almost like a negative control.

294
00:34:09,620 --> 00:34:14,300
So essentially, women rarely drink and or smoke as well.

295
00:34:14,780 --> 00:34:22,190
In fact, you know, looking at the analysis in, women can actually say, well, eat due to alcohol intake.

296
00:34:22,400 --> 00:34:32,810
And there were no positive associations with the biomarkers, you know, blood pressure, HDL, GTI and stroke or am I in in women?

297
00:34:33,860 --> 00:34:38,690
You know, that would suggest it's sort of alcohol that's, you know, sort of driving it.

298
00:34:39,980 --> 00:34:45,950
We then just tried to look at some other things. So things like sort of BMI and glucose.

299
00:34:46,700 --> 00:34:57,200
And again, it's all positive associations of BMI and glucose with, you know, genetically predicted alcohol.

300
00:34:58,260 --> 00:35:04,050
For diabetes, we saw some kind of positive association.

301
00:35:04,350 --> 00:35:07,950
Again, it's sort of modest.

302
00:35:09,210 --> 00:35:13,820
And actually, when we adjusted for BMI, it's it's all gone away.

303
00:35:14,190 --> 00:35:22,950
And actually, that that kind of makes sense, really. So it's mostly driven by the adiposity rather than sort of alcohol.

304
00:35:23,910 --> 00:35:26,940
And then we also had to look at that cancer.

305
00:35:26,950 --> 00:35:34,889
So, I mean, there's all that benefit for, you know, potentially people that thought there were benefits for cardiovascular disease.

306
00:35:34,890 --> 00:35:47,220
But essentially, you can see there's quite strong associations for different types of cancer that related to sort of alcohol consumption.

307
00:35:49,090 --> 00:35:54,310
And you know, for lung cancer again just looking in evening never smokers.

308
00:35:54,460 --> 00:36:00,880
You know that's that's probably the one to sort of look at you can see I mean, the numbers are small, but,

309
00:36:00,880 --> 00:36:08,740
you know, there's a positive association in sort of never smokers with, you know, alcohol and lung cancer.

310
00:36:09,880 --> 00:36:14,130
People are talking about things sort of like so the dietary factors, the various,

311
00:36:14,200 --> 00:36:20,230
the things we looked at, those are things like linoleic acid and omega three fatty acids.

312
00:36:20,650 --> 00:36:25,930
It looks like. Yeah, it raises those things sort of moderately.

313
00:36:26,480 --> 00:36:32,139
You know, so there could be, you know, if there are any benefits, it could be sort of working through these things.

314
00:36:32,140 --> 00:36:39,430
But, you know, the effects are sort of fairly small. The global burden of disease, you know, says, well, you know,

315
00:36:39,460 --> 00:36:46,780
actually there's quite a few deaths related to sort of alcohol and disability life years and all those kind of things.

316
00:36:47,140 --> 00:36:51,190
And obviously, it's not just sort of CVD.

317
00:36:51,190 --> 00:36:59,530
There's things like injuries and digestive disease and various of the things that can be caused by sort of alcohol.

318
00:36:59,530 --> 00:37:02,530
And it's sort of violence and various other things too.

319
00:37:03,580 --> 00:37:12,610
So those apparent protective effects that they are showing and saying that, you know, benefits and that it's all based on conventional epidemiology.

320
00:37:12,610 --> 00:37:25,480
So, I mean, is that the right thing to be working with if if there is clear evidence that this J-shaped association that seen could be due to sort of,

321
00:37:25,540 --> 00:37:29,320
you know, sort of biases some of the genetic evidence.

322
00:37:29,380 --> 00:37:33,820
Essentially, stroke increases by about a third for every four drinks a day.

323
00:37:34,660 --> 00:37:38,480
You know, moderate drinking doesn't appear to protect against stroke.

324
00:37:39,550 --> 00:37:42,430
Didn't appear to be any clear benefits for me.

325
00:37:43,600 --> 00:37:52,569
We showed that, you know, genetically predicted alcohol based blood pressure, glucose adiposity, those kind of things.

326
00:37:52,570 --> 00:37:58,570
So, yeah, there's a lot of other things going on. And of course, it's associated with several cancers.

327
00:37:59,410 --> 00:38:10,080
Just to conclude. It would say that, you know, that essentially genetic epidemiology suggests that, you know,

328
00:38:10,080 --> 00:38:19,620
the observational work is non causal, you know, so essentially the protective effect of moderate alcohol is not causal.

329
00:38:20,700 --> 00:38:26,160
You know, it does raise blood pressure. And that's that's that's not a good thing.

330
00:38:26,340 --> 00:38:31,620
The it looks like it raises blood pressure to a greater extent than it increases

331
00:38:31,620 --> 00:38:35,970
things like HDL and some of the other things that are considered sort of good.

332
00:38:37,200 --> 00:38:40,750
But it could be that for me, you know,

333
00:38:40,770 --> 00:38:45,629
some of the kind of benefits that are going on could have cardioprotective effects

334
00:38:45,630 --> 00:38:49,950
that actually would being balanced out by sort of raising that pressure example,

335
00:38:50,130 --> 00:38:54,180
which is why you didn't see the line definitely going up for MRI.

336
00:38:54,420 --> 00:39:01,440
It was, you know, sort of fairly flat. So there could be a couple of things that are counterbalancing what's going on there.

337
00:39:02,490 --> 00:39:11,070
To finish, taken together, I would suggest that the harms outweigh the benefits.

338
00:39:11,790 --> 00:39:15,119
So that was my final point. So thanks.

339
00:39:15,120 --> 00:39:22,620
And the collaborators from Google Biobank and the various sort of funders and that's it.