1
00:00:00,270 --> 00:00:06,390
So without further ado, Professor Geoffrey Watson, professor emeritus of clinical pharmacology.

2
00:00:07,950 --> 00:00:11,099
Would you like to tell us whether antibiotics will make us better? Yeah.

3
00:00:11,100 --> 00:00:16,240
Well, who would like antibiotics to make them pregnant?

4
00:00:16,260 --> 00:00:20,370
Any volunteers here? Anybody and anybody want it?

5
00:00:21,300 --> 00:00:26,190
There's the title and there's Robyn, who is a co-author on this work.

6
00:00:26,880 --> 00:00:31,990
And actually, of course, the title should really be that,

7
00:00:34,560 --> 00:00:44,520
because I'm going to talk about a putative interaction of antibiotics with oral contraceptives primarily,

8
00:00:46,800 --> 00:00:58,320
and this starts in 1971 with a paper in the German literature from Reimers and a later paper by Reimers again in 73,

9
00:00:58,320 --> 00:01:07,979
and then somebody else's paper in 74, all in the German literature saying that this drug rifampicin you can see at the top,

10
00:01:07,980 --> 00:01:16,350
which is used to treat tuberculosis, incidentally, named after the French film Rififi.

11
00:01:16,380 --> 00:01:28,890
Anybody seen Rififi? Do you rififi socialism or Shazaam is a film by Zhu Das, which is about a robbery in Paris.

12
00:01:29,670 --> 00:01:40,379
It won the jury prize at Cannes, and it's how Dessa met Melina mercouri and made his career and the people who first isolated

13
00:01:40,380 --> 00:01:47,100
the rifampicin antibiotics from a particular organism called the culture after rififi.

14
00:01:48,090 --> 00:01:53,999
And they called them River My Sins. And so rifampicin is an MP derivative of the Reformation.

15
00:01:54,000 --> 00:02:03,930
So that's why it's called rifampicin. So what, what they showed was that there is an interaction of rifampicin with the oral contraceptive.

16
00:02:05,940 --> 00:02:11,610
And this made news in the Journal of the American Medical Association in 1974.

17
00:02:11,850 --> 00:02:13,380
There's a picture of Reimers,

18
00:02:13,980 --> 00:02:22,710
and this annotation tells you that if you give rifampicin to people who are taken to women who are taking the oral contraceptive,

19
00:02:22,980 --> 00:02:25,920
you get breakthrough bleeding or spotting.

20
00:02:27,570 --> 00:02:35,580
And some women had absence of bleeding after discontinuation of the pill, but had breakthrough bleeding in the following cycle, which is a bit odd.

21
00:02:37,020 --> 00:02:42,570
And even some women became pregnant despite taking the oral contraceptive.

22
00:02:43,560 --> 00:02:50,310
Now the mechanism of this interaction is well-described and totally agreed upon.

23
00:02:50,310 --> 00:02:55,170
Nobody has any difficulty about accepting that this is a true interaction.

24
00:02:55,800 --> 00:03:03,150
My old professor, David Graham Smith, gave a public lecture some years ago talking about this kind of interaction.

25
00:03:03,150 --> 00:03:06,360
He said there are certain drugs which are enzyme inducers.

26
00:03:06,360 --> 00:03:11,700
I'll come back to that and they will reduce the amount of contraceptive you have in your

27
00:03:11,700 --> 00:03:15,960
body and you may therefore become pregnant and a woman in the front row and oh my God.

28
00:03:17,090 --> 00:03:22,020
But it does happen. And here's the mechanism.

29
00:03:22,030 --> 00:03:27,089
This is from a much later paper, but it shows you that rifampicin or rifampin,

30
00:03:27,090 --> 00:03:35,050
as they Americans call it, reduces the amount of oestrogen oestrogen, the IEEE.

31
00:03:35,130 --> 00:03:40,620
On the slide there is ethanol oestradiol one of the common oestrogens used in oral contraceptives.

32
00:03:41,370 --> 00:03:54,000
You can see that when you add ampicillin after 14 days, the concentrations of the ethanol oestradiol reduce by about 50%.

33
00:03:55,140 --> 00:04:00,420
And the same is true of another member of that group refer Biotin, which is another rifampicin.

34
00:04:00,660 --> 00:04:08,670
The effect is less, but nonetheless it occurs. And rifampicin is probably one of the most potent enzyme inducers.

35
00:04:08,670 --> 00:04:19,200
It increases the metabolism of drugs in the liver by increasing the amount of enzyme, and there are many enzyme inducers which do exactly the same,

36
00:04:19,410 --> 00:04:29,100
although two different degrees, and they all tend to reduce the efficacy of the oral contraceptive.

37
00:04:29,670 --> 00:04:32,309
And the first sign is mid-cycle spotting.

38
00:04:32,310 --> 00:04:39,600
You're in the middle of your cycle, you've had a period here, you're expecting a period there, and in the middle you get a little bit of bleeding.

39
00:04:40,080 --> 00:04:43,560
That's the first sign that your contraceptive is failing.

40
00:04:44,610 --> 00:04:49,200
And of course, pregnancy is the most dramatic sign that your contraceptive has failed.

41
00:04:50,700 --> 00:04:59,520
Now, in 1975, this GP, Jill Dorset at the top there wrote a letter to the BMJ and the BMJ.

42
00:04:59,950 --> 00:05:06,520
Hated it. Drug interactions with oral contraceptives. And what she said was, it's well known.

43
00:05:07,090 --> 00:05:15,280
See if I can get this. It's well known that there are drug interactions with these drugs barbiturates, rifampicin, phenol, the Arizona phenytoin.

44
00:05:15,280 --> 00:05:18,609
They all can act as enzyme inducers.

45
00:05:18,610 --> 00:05:24,370
I'm not sure about female beauties being there, but the others are certainly enzyme inducers.

46
00:05:25,180 --> 00:05:35,500
And then she says, But I have seen three cases, three patients who became pregnant when they were given ampicillin.

47
00:05:36,580 --> 00:05:43,210
And she wonders if there might be an interaction with antibiotics that are not enzyme inducers.

48
00:05:45,610 --> 00:05:53,649
The BMJ clearly had pregnancy on its mind because it followed that with another letter about the Dow concealed,

49
00:05:53,650 --> 00:05:56,590
which is also problematic but in a different way.

50
00:05:57,460 --> 00:06:07,660
So at about the same time, in fact, in the same month, November 1975, as Jill Nossiter wrote her letter, a group in Finland,

51
00:06:08,560 --> 00:06:17,400
Adler, CREUTZ and his colleagues had published data on the excretion of oestrogens in the urine of women who are pregnant.

52
00:06:18,190 --> 00:06:24,520
And for some reason they had studied the effects of ampicillin, and this is what they found.

53
00:06:26,830 --> 00:06:31,870
Here you have the pregnant women excreting all these different oestrogens.

54
00:06:31,870 --> 00:06:35,800
These three are probably the most important, and here they are.

55
00:06:35,950 --> 00:06:38,950
These are conjugated oestrogens. I'll come back to that in a minute.

56
00:06:39,730 --> 00:06:49,300
Here they are after ampicillin, huge increases in the excretion in the faeces of conjugated oestrogen.

57
00:06:50,260 --> 00:06:51,790
Now, why is that important?

58
00:06:51,820 --> 00:07:02,200
Well, there is a mechanism that I'll come back to, but remember that ampicillin increases the excretion of conjugated oestrogens in the faeces.

59
00:07:05,650 --> 00:07:09,040
This seems to be a highly variable effect.

60
00:07:09,790 --> 00:07:16,569
These are three patients from another of their papers were actually reproduced in the contraception paper,

61
00:07:16,570 --> 00:07:22,930
but taken from another paper showing that there is huge variability in this effect.

62
00:07:23,020 --> 00:07:25,870
And they've demonstrated here three individuals,

63
00:07:26,170 --> 00:07:37,060
one of whom has an increase in this time plasma and conjugated oestradiol and conjugated one who has no change and one who has a fall.

64
00:07:38,290 --> 00:07:48,100
So there is a huge variability, one assumes, in the extent to which conjugation and conjugation or conjugation occurs in different people.

65
00:07:50,140 --> 00:07:54,190
How this translates into the general population, I don't think anyone knows.

66
00:07:54,550 --> 00:07:56,590
These are just three cases that they showed,

67
00:07:57,250 --> 00:08:05,350
but it does suggest that there's huge variability and that would not be surprising if this is a bacterial effect,

68
00:08:05,350 --> 00:08:08,530
if ampicillin is in some way affecting bacteria in the gut.

69
00:08:08,980 --> 00:08:14,740
And we all have very different microbiomes, the families of bugs that live in our gut.

70
00:08:15,040 --> 00:08:21,910
So variability would not be surprising. So here's the mechanisms we have.

71
00:08:22,210 --> 00:08:28,690
On the left, we have the normal events for oestrogens when you take a dose of the pill.

72
00:08:29,650 --> 00:08:36,670
First of all, the ethanol oestradiol, which is the usual oestrogen in the pill, is absorbed from small.

73
00:08:37,750 --> 00:08:44,500
The people who drew this diagram missed out intestine, not pill, but intestine.

74
00:08:45,460 --> 00:08:52,150
So the ethanol oestradiol is absorbed. You swallow it, it's absorbed from the gut into the systemic circulation.

75
00:08:52,750 --> 00:08:58,470
It's then metabolised in the liver where it is conjugated, joined to glucose,

76
00:08:58,560 --> 00:09:04,750
chronic acid and sulphate, primarily in the liver, also in the intestine.

77
00:09:07,570 --> 00:09:14,950
Drug that isn't conjugated goes to the organs where ovulation occurs and prevents ovulation,

78
00:09:15,400 --> 00:09:24,100
but drug that is conjugated is excreted via the bile back into the gut where bacteria d conjugate

79
00:09:24,100 --> 00:09:30,070
it so that it can be reabsorbed topping up the amount of oestrogen you have in your body.

80
00:09:32,850 --> 00:09:39,270
So the two mechanisms whereby this could potentially be altered is enzyme induction.

81
00:09:39,270 --> 00:09:44,850
You increase the amount of metabolism and so you get less of the oestrogen in the body as a whole.

82
00:09:48,430 --> 00:09:54,010
It's excreted as inactive metabolites or if that doesn't happen.

83
00:09:54,880 --> 00:10:06,850
An antibiotic in the gut can inhibit the bacteria that did conjugate the oestrogen after it's been put from the bile into the gut and reabsorbed.

84
00:10:07,450 --> 00:10:14,560
The de conjugation process necessary for reabsorption is inhibited by removing the D conjugate in bacteria.

85
00:10:15,490 --> 00:10:18,640
And so again, you get less drug in the body.

86
00:10:20,020 --> 00:10:30,910
And that's the importance of the enormous increase in conjugated oestrogens in the gut in the faeces after the administration of ampicillin.

87
00:10:31,680 --> 00:10:39,790
The implication is that ampicillin has inhibited the bacteria which would normally de conjugate the oestrogen,

88
00:10:40,150 --> 00:10:43,360
and so the conjugated oestrogen rises enormously.

89
00:10:43,600 --> 00:10:47,350
The oestrogen is not reabsorbed and you lose the effect of the oestrogen.

90
00:10:48,130 --> 00:10:52,920
So that is the supposed mechanism. Okay.

91
00:10:53,920 --> 00:10:59,020
So now this is a picture of the first issue of the British National Formulary,

92
00:10:59,020 --> 00:11:07,660
which is now the standard text that all prescribers use as their first go to to determine how they should prescribe a medicine.

93
00:11:08,440 --> 00:11:11,009
You will find it on the desk of every general practitioner,

94
00:11:11,010 --> 00:11:17,770
or at least online on the screen when they're prescribing and they'll look up the British National Formulary.

95
00:11:18,010 --> 00:11:22,480
If you go to a coroner's inquest, the coroner will say, what does it say in the BNF?

96
00:11:23,620 --> 00:11:26,500
Or you can go to a trial. They'll say, What does it say in the BNF?

97
00:11:26,510 --> 00:11:32,170
There are other sources of information, of course, important sources, but this is one of the major sources of information.

98
00:11:32,650 --> 00:11:38,680
And here's the first issue from 1981 and all the extracts.

99
00:11:38,680 --> 00:11:46,959
So I'll show you from the various issues of the BNF are taken from the actual hard copies that I've collected over the years.

100
00:11:46,960 --> 00:11:52,870
That's why the quality is not awfully good, because the quality of the original print is not very good.

101
00:11:53,470 --> 00:12:02,410
So here's the extract from the 1981 number one issue on interactions of drugs with oral contraceptives.

102
00:12:02,740 --> 00:12:11,290
And you can see it gives you two sets. One is the drugs that I've mentioned already barbiturates, phenytoin, rifampicin, and so on.

103
00:12:11,590 --> 00:12:21,520
All these drugs are enzyme inducers, and there is total agreement that that reduces the effect of the pill without doubt.

104
00:12:23,050 --> 00:12:27,820
Notice this drug grabs you full, then I'll mention it briefly later.

105
00:12:27,820 --> 00:12:32,650
It's an antifungal drug, not an antibiotic, not an antibacterial drug.

106
00:12:34,930 --> 00:12:42,310
Rifampicin is the only antibacterial drug on this list that's antifungal and the rest are used for other purposes.

107
00:12:43,330 --> 00:12:46,510
But then it says oral antibiotics.

108
00:12:46,930 --> 00:12:51,759
Black It's not enzyme inducers such as ampicillin and tetracycline.

109
00:12:51,760 --> 00:12:58,960
Reduced effect risk, probably small. And the risk was thought to be is not clear.

110
00:12:59,470 --> 00:13:05,680
I don't think anybody really knew and I think nobody still knows what the risk is if there is a risk at all.

111
00:13:09,910 --> 00:13:22,629
So that's 1981. By 1988, when this paper was published, the MHRA, the Medicines and Healthcare Products Regulatory Agency, actually its predecessor,

112
00:13:22,630 --> 00:13:32,110
which was called the Medicines Control Agency, has gathered a whole lot of isolated reports coming in from here, there and everywhere on yellow cards.

113
00:13:32,260 --> 00:13:36,760
You can fill out a yellow card. They used to be actually physical yellow cards available.

114
00:13:37,030 --> 00:13:43,990
GP's would have them and they'd fill out a report saying I suspect that there has been an adverse effect

115
00:13:44,200 --> 00:13:49,749
in this case or an adverse interaction in this case and the yellow card would go into the emkay,

116
00:13:49,750 --> 00:13:55,690
the Medicines Control Agency, and a whole set of reports would be gathered together.

117
00:13:55,960 --> 00:14:01,330
And Alastair Brackenridge and his colleagues in Liverpool looked at all of these in 1988 and they

118
00:14:01,330 --> 00:14:08,350
found that there were 63 reports of alleged interactions with antibiotics and oral contraceptives,

119
00:14:08,710 --> 00:14:11,950
compared with only 43 for antiepileptic drugs.

120
00:14:12,640 --> 00:14:16,990
Now of course there's reporting bias in this and why people should want to

121
00:14:16,990 --> 00:14:21,460
send in an individual report or not send in an individual report is not clear.

122
00:14:22,000 --> 00:14:26,980
And probably less than 10% of all incidents are actually reported in this way.

123
00:14:27,370 --> 00:14:30,519
So it's it's difficult to assess the quality of the data.

124
00:14:30,520 --> 00:14:31,390
But nonetheless,

125
00:14:31,810 --> 00:14:40,810
you can see there were more reports of putative interactions with antibiotics than there were with the anti-epileptic enzyme inducing drugs,

126
00:14:41,740 --> 00:14:47,410
which is suggestive. So now we come to the BNF of 1990.

127
00:14:47,580 --> 00:14:55,410
2/24 edition is published twice yearly, and now for the first time,

128
00:14:55,890 --> 00:15:01,980
the information is increased from the minimal information that you saw in the first

129
00:15:01,980 --> 00:15:06,840
issue took a long time and it says in the case of combined oral contraceptives,

130
00:15:07,050 --> 00:15:11,700
some broad spectrum means they cover a wide range of organisms.

131
00:15:12,060 --> 00:15:17,220
Antibiotics, for example, ampicillin may interfere with oestrogen absorption is not quite right.

132
00:15:17,550 --> 00:15:21,240
It's oestrogen reabsorption after conjugation.

133
00:15:21,240 --> 00:15:28,620
But get the point and it doesn't tell you what the mechanism is supposed to be, just says interferes with absorption.

134
00:15:29,700 --> 00:15:35,790
And it says if you're taking a short course for seven days after and for seven days after stopping,

135
00:15:36,090 --> 00:15:40,320
take extra precautions because your pill may not be as effective.

136
00:15:42,450 --> 00:15:45,540
If the course exceeds two weeks, resistance develops.

137
00:15:45,540 --> 00:15:58,350
And again, they don't say what the mechanism is, but that's the advice being given in 19 can remember 1990 to 1992.

138
00:15:59,160 --> 00:16:10,650
Okay. And this this information persists until 2010 when the information is expanded for the first time.

139
00:16:11,070 --> 00:16:13,860
And here's the information given in 2010,

140
00:16:16,500 --> 00:16:25,140
and it tells tells you again that there is a possible interaction while taking a short course or four and for seven days after stopping.

141
00:16:26,340 --> 00:16:32,160
And that if the course in this case exceeds three weeks. No, not to I don't know why they've changed that.

142
00:16:32,790 --> 00:16:37,619
They say that the bacterial flora develop antibacterial resistance.

143
00:16:37,620 --> 00:16:41,819
So the problem that was started by ampicillin, say,

144
00:16:41,820 --> 00:16:49,170
or tetracycline inhibiting the bacteria is no longer a problem because the bacteria are no longer inhibited and grow back.

145
00:16:49,650 --> 00:17:01,660
And so the status quo is restored. And they say, in addition, that this may also apply to patches and vaginal rings,

146
00:17:02,380 --> 00:17:05,950
although there's not a lot of information about that actually if you look for it.

147
00:17:09,370 --> 00:17:19,840
So that's 2010. But already in 2009, just before the publication of that BNF,

148
00:17:20,350 --> 00:17:27,430
the W.H.O. had looked at the data and had decided that there was no evidence of an interaction.

149
00:17:29,650 --> 00:17:34,870
And this is their fourth edition of the medical eligibility criteria for contraceptive use.

150
00:17:35,440 --> 00:17:38,530
And you can see that at the top broad spectrum antibiotics.

151
00:17:38,800 --> 00:17:44,260
Most broad spectrum antibiotics do not affect the contraceptive effectiveness of combined oral contraceptives.

152
00:17:45,160 --> 00:17:53,830
This is a curious statement because they say most antibiotics, which suggests that maybe some broad spectrum antibiotics do,

153
00:17:53,830 --> 00:17:58,180
but they don't say which, even if they think that some might.

154
00:17:59,200 --> 00:18:05,680
So it's hard to know what that means. I also draw your attention to the second entry on antifungals.

155
00:18:06,580 --> 00:18:14,800
No clinically significant directions, but you remember that I told you that crazy, awful thing, which is an enzyme inducer, is an antifungal agent.

156
00:18:15,670 --> 00:18:24,220
So there is evidence that at least one antifungal agent that has an interaction which they seem to have ignored in this survey,

157
00:18:25,030 --> 00:18:35,200
I won't go into the others, but they do mention rifampicin a bit in the the metabolism and so on.

158
00:18:36,280 --> 00:18:45,910
Here's let's 2009 too early or too late rather for the 2010 issue of the BNF to pick that up.

159
00:18:47,020 --> 00:18:56,830
But the following year, in 2011, this group from the Faculty of Sexual and Reproductive Health Care of the Royal College of

160
00:18:57,070 --> 00:19:05,110
Obstetricians and Gynaecologists published a report in which they echoed the W.H.O. advice.

161
00:19:06,190 --> 00:19:11,229
And this is what they say. They say rifampicin like drugs,

162
00:19:11,230 --> 00:19:17,110
enzyme inducers that are the only antibiotics that are enzyme inducers have consistently been

163
00:19:17,110 --> 00:19:23,290
shown to reduce serum concentrations of ethanol oestradiol with the other antimicrobial agents,

164
00:19:23,290 --> 00:19:28,809
penicillins, tetracyclines, macrolides, fluoroquinolones and midazolam.

165
00:19:28,810 --> 00:19:34,720
Antifungal drugs, which I'm not going to talk about, which are not enzyme inducers.

166
00:19:35,320 --> 00:19:44,320
The mechanism is supposed to be through inhibiting colonic bacteria, but they say there is no evidence to prove such an interaction.

167
00:19:44,890 --> 00:19:52,270
What they mean is there's no evidence to prove such a mechanism for such an interaction that

168
00:19:52,270 --> 00:19:58,120
they've made that kind of leap without really considering the evidence properly in our review.

169
00:19:58,690 --> 00:20:06,879
Anyway, that's what they said. There's no evidence. But then they go on to say that previously, well, we weren't sure.

170
00:20:06,880 --> 00:20:11,320
And so we were precautionary. It's a big event.

171
00:20:11,650 --> 00:20:22,030
Becoming pregnant. If you hadn't intended to be life changing, so why wouldn't you take precautions in case for a two week course of antibiotics?

172
00:20:22,030 --> 00:20:26,080
It might change your life. Wouldn't you take precautions?

173
00:20:27,490 --> 00:20:34,060
But no. They say we don't think it's necessary. That alone to me is worrying.

174
00:20:34,330 --> 00:20:40,030
Even if I even if I agreed with their assessment of the evidence, I'd be worried about that,

175
00:20:40,270 --> 00:20:46,380
because as long as there is a smallest doubt, proving a negative is not possible.

176
00:20:46,390 --> 00:20:51,250
And that's what they are assuming. They're assuming that the negative has been proven and the precautions are not needed.

177
00:20:51,550 --> 00:21:00,270
That alone gives me pause. And they then mentioned that the shows already said this, as I've shown.

178
00:21:00,480 --> 00:21:07,320
And the US medical eligibility criteria for contraceptive use have agreed they've adopted the recommendations.

179
00:21:07,620 --> 00:21:14,970
So they feel that they should do so too. And later that year, in September 2011,

180
00:21:15,210 --> 00:21:22,020
the BNF picks this up and says latest recommendations are that no additional contraceptive precautions are required.

181
00:21:24,060 --> 00:21:27,780
I was on the British National Formulary Committee at that time.

182
00:21:28,500 --> 00:21:36,370
Normally what would happen is that the pharmacists on that committee would bring suggestions to the committee for ratification.

183
00:21:36,390 --> 00:21:40,950
That didn't happen. They just put this in straight from the college.

184
00:21:41,460 --> 00:21:45,690
Well, College of Obstetricians guidance without consultation as far as I can see.

185
00:21:45,720 --> 00:21:50,250
I may be wrong about that. And I was very upset about this when I heard I said this is this.

186
00:21:50,760 --> 00:21:57,380
Even if even if the evidence is the way it is, I would still recommend taking precautions.

187
00:21:57,390 --> 00:22:00,920
Why not? But there you are.

188
00:22:01,200 --> 00:22:09,120
And that is, I think, in the current guidance. So what was the evidence?

189
00:22:10,050 --> 00:22:16,500
Well, the college divided the evidence into two types, direct and indirect.

190
00:22:17,550 --> 00:22:27,330
This is the direct evidence. Several studies looked at combined hormonal methods have not demonstrated a decrease in levels concentrations of ethanol.

191
00:22:27,390 --> 00:22:34,920
Stradale I'll come back to that. It's an important point. Small Nonrandomized studies no effects on pharmacokinetics.

192
00:22:35,670 --> 00:22:41,180
Small prospective Nonrandomized studies failed to show any effect on gonadotropin.

193
00:22:41,190 --> 00:22:51,630
The hormones that control certain aspects of pregnancy and three small randomised trials suggest that those two drugs may not affect.

194
00:22:52,290 --> 00:22:57,780
And when we look at those trials, this is the number of individuals in each of the trials.

195
00:22:58,380 --> 00:23:02,280
The study number is you see the top one is 57 to 59.

196
00:23:02,280 --> 00:23:11,729
So that's those three references and so on. You can see the three randomised trials, a 64 to 65, they have a total of 54 subjects in all.

197
00:23:11,730 --> 00:23:16,440
That's 27 in each group. Tiny study.

198
00:23:17,130 --> 00:23:22,470
You could not conclude from this that there was no interaction,

199
00:23:23,080 --> 00:23:27,060
but I don't know what the power of the studies would be, but it ain't going to be very high.

200
00:23:29,340 --> 00:23:31,800
So the direct evidence is not very good.

201
00:23:33,750 --> 00:23:42,780
I want to have now an interlude to discuss the question of whether all women would be affected by this interaction if it occurred.

202
00:23:43,080 --> 00:23:48,209
Because the assumption has been looking at the evidence that the W.H.O. and the

203
00:23:48,210 --> 00:23:52,980
Royal College of Obstetricians has quoted that every woman would be affected.

204
00:23:52,980 --> 00:23:55,680
And that is generally the case with interactions.

205
00:23:57,840 --> 00:24:04,920
If I give you a drug that induces liver enzymes, then there will be variability from individual to individual.

206
00:24:05,100 --> 00:24:14,429
But everybody will have some enzyme induction to some extent, and everybody will therefore be at risk of some sort.

207
00:24:14,430 --> 00:24:22,440
Some will be more effective than others, of course, and the distribution of risk will be a normal distribution, generally speaking.

208
00:24:24,540 --> 00:24:28,320
And the general assumption is that adverse interactions behave in that way,

209
00:24:28,920 --> 00:24:34,500
that although there is variability, everybody is at risk of some degree or other.

210
00:24:35,490 --> 00:24:38,520
That doesn't necessarily happen with some interactions.

211
00:24:38,520 --> 00:24:48,030
And it is particularly not the case when bacteria are involved because of the huge variability in bacterial flora in our guts.

212
00:24:48,810 --> 00:24:51,660
And this is an example that demonstrates this.

213
00:24:53,220 --> 00:25:03,810
So it's an interlude from oral contraceptives and this is a report of an interaction of digoxin, which is we don't use nowadays very much, if at all.

214
00:25:03,930 --> 00:25:07,770
It's a cardiac drug with an antibiotic erythromycin.

215
00:25:08,160 --> 00:25:10,830
And in this case, it caused toxicity.

216
00:25:12,660 --> 00:25:20,370
This is the first case report that this is so and this has proved to be a correct assumption that this drug erythromycin can,

217
00:25:20,370 --> 00:25:27,449
in some women cause some men and indeed in some people cause digoxin toxicity.

218
00:25:27,450 --> 00:25:38,520
And it does it by changing the metabolism of d Jackson in the gut mediated by a bacterium called you bacterium Lenton.

219
00:25:42,170 --> 00:25:50,420
And this bacterium is responsible for changing the structure of digoxin to inactive compounds.

220
00:25:52,580 --> 00:25:57,590
So here's the structure of the Jackson. It looks like a steroid.

221
00:25:58,580 --> 00:26:05,040
This bit here is a steroid. So it's not unrelated to the oestrogen and progesterone.

222
00:26:05,630 --> 00:26:08,840
Although the three dimensional structure is different.

223
00:26:12,140 --> 00:26:17,290
And this bit of the molecule is called the genuine.

224
00:26:17,300 --> 00:26:25,280
So the whole molecule is digoxin. This bit is called digoxin genuine because it gives rise to the whole molecule.

225
00:26:25,730 --> 00:26:35,570
These things each are glucose like molecules, and so the whole structure is known as a cardiac glycosides.

226
00:26:36,110 --> 00:26:40,310
It's a drug that acts on the heart and has glucose like molecules in it.

227
00:26:42,260 --> 00:26:46,100
And this group of compounds that are many of them are called cardiac glycosides.

228
00:26:46,370 --> 00:26:54,500
So digoxin is a cardiac glycosides. And as it's metabolised in the body, each one of these glucose molecules is snipped off.

229
00:26:56,690 --> 00:27:00,700
Now, usually when you metabolise a chemical, it makes it less potent.

230
00:27:00,710 --> 00:27:09,050
But in the case of cardiac glycosides, removing each of the molecules of glucose like molecules increases the activity of the drug.

231
00:27:09,440 --> 00:27:20,240
They become more potent. But if you change this bit here by reducing the double bond, putting hydrogen molecules on you, reduce the activity.

232
00:27:20,270 --> 00:27:26,120
So there are varying changes in the activity of the molecule, depending on which bit of it is metabolised.

233
00:27:26,810 --> 00:27:36,200
And when you add that, when you put it in the presence of the you bacterium, you get a whole load of active of inactive compounds.

234
00:27:36,980 --> 00:27:46,580
Sorry. Here's the normal metabolism to a variant of a variety of active compounds you put in the you bacterium and you get inactive compounds.

235
00:27:46,850 --> 00:27:52,220
If you inhibit the you bacterium, you get more active compounds and therefore toxicity.

236
00:27:55,430 --> 00:28:04,760
Only 10% of people are subject to this interaction, but 20 or 30% have the bacterium.

237
00:28:05,300 --> 00:28:09,890
And with the variability, about 10% actually get the interaction.

238
00:28:10,880 --> 00:28:19,110
So this is a very good example of how interactions with bacteria where an antibiotic changes.

239
00:28:19,640 --> 00:28:24,530
In this case, the metabolism of a drug only affects a small number of individuals.

240
00:28:24,530 --> 00:28:27,770
In this case, about 10%. Does not affect everybody.

241
00:28:28,970 --> 00:28:38,090
And even of those whom it affects, not everybody is going to be at risk of toxicity because of the variability in those individuals.

242
00:28:38,720 --> 00:28:46,970
So what I one wondered in relation to the numbers of individuals who had been studied in those.

243
00:28:48,490 --> 00:28:58,200
Study as I show you before. How many such individuals would you need if you wanted to demonstrate that there was an interaction?

244
00:28:58,800 --> 00:29:02,670
Most interaction studies only have a dozen or so people.

245
00:29:03,450 --> 00:29:12,180
24 was the most in those studies I showed you. Is that enough to show an effect if it affects, say, only 10% of the population?

246
00:29:13,470 --> 00:29:17,550
Now you'll know. The answer to that straight away is obviously far too few.

247
00:29:18,270 --> 00:29:21,480
But I thought I'd do the sums just to convince myself.

248
00:29:22,740 --> 00:29:28,500
And what I've done here is to set up the putative results of a randomised experiment.

249
00:29:30,510 --> 00:29:34,020
And I've said, right, we've got a group of women who are taking the oral contraceptive.

250
00:29:34,770 --> 00:29:40,800
I'm going to give half of them an antibiotic and the other half a placebo to be double blind.

251
00:29:41,490 --> 00:29:47,670
Nobody will know what's going on. And then I'm going to watch and see how many of them become pregnant and how many of them don't.

252
00:29:49,170 --> 00:29:53,220
Not an ethical experiment, but just a thought experiment.

253
00:29:54,690 --> 00:29:57,120
And I want to know, how big is X?

254
00:29:59,520 --> 00:30:08,280
If one of the women in the placebo group becomes pregnant because people do become pregnant despite taking oral contraceptives.

255
00:30:08,820 --> 00:30:15,750
And if 1/10 of the women in the antibiotic group become pregnant because that's the risk, 10%.

256
00:30:16,470 --> 00:30:21,750
So here's some numbers I've said. Well, suppose I studied 30 women.

257
00:30:22,620 --> 00:30:26,280
1530 in each group bigger than any of the other studies.

258
00:30:27,480 --> 00:30:30,720
And one of them in the placebo group became pregnant.

259
00:30:31,860 --> 00:30:35,250
Three of them, 10% in this group became pregnant.

260
00:30:36,240 --> 00:30:41,600
You might say that should before because there's one there and one should be there.

261
00:30:41,610 --> 00:30:46,600
But roughly what do you get?

262
00:30:46,620 --> 00:30:52,020
Well, there's the chi square, nowhere near significant, as you all doubtless expected.

263
00:30:52,860 --> 00:31:04,260
So my question to you is how many people should I study in this paradigm if I want to find a P value less than 0.05?

264
00:31:04,890 --> 00:31:08,850
Anybody got any ideas? If they want to make a suggestion?

265
00:31:10,710 --> 00:31:17,220
500 500 in each group are to 50 in total to 50 in each group.

266
00:31:17,330 --> 00:31:20,880
Yeah. Any advance on that? Is that.

267
00:31:20,910 --> 00:31:27,180
Does that sound right? Sounds good. Susanna says here roughly.

268
00:31:27,570 --> 00:31:29,330
Yeah. It's not bad, actually. It's not a bad.

269
00:31:29,340 --> 00:31:38,700
I've doubled the numbers here, 60 in each group and it just almost gets to be significant and I really want better than that.

270
00:31:38,700 --> 00:31:41,910
So 250 in each group. Yeah. It might be. Could be.

271
00:31:42,150 --> 00:31:54,060
There's that sort of order. We need 100, 200, 300 in each group before I'm going to expect to detect this effect on the assumption of a 10% risk.

272
00:31:54,240 --> 00:32:00,040
It might be less than that. The risk might be less. Might be one one in 100.

273
00:32:00,540 --> 00:32:03,060
Still be important to that one individual.

274
00:32:03,070 --> 00:32:10,200
And how many millions of women take contraceptives and some of them come through the course of antibiotics because of a urinary tract infection.

275
00:32:12,390 --> 00:32:21,090
So you need quite a lot. One of my colleagues in the department in the statistics group said, well, she said,

276
00:32:21,090 --> 00:32:34,530
if you wanted to do a study in which you gave advice to women about what they should do when you gave them an antibiotic,

277
00:32:35,010 --> 00:32:40,870
you would need a huge study to determine whether there was an interaction in that scenario.

278
00:32:40,890 --> 00:32:46,650
This is a much simpler scenario. I'm actually counting. I'm watching them day by day and counting the pregnancies.

279
00:32:46,860 --> 00:32:55,410
If I just gave them advice to take extra precautions, you'd need a much bigger study to determine whether your advice was helpful or not.

280
00:32:55,740 --> 00:33:00,840
And that's the practical problem, actually, in a way. How do you advise women what to do?

281
00:33:03,180 --> 00:33:08,430
So not very good. The direct so-called direct evidence.

282
00:33:08,850 --> 00:33:15,929
And then there's some indirect evidence. I'm not going to go through it in detail, but they have seven pieces of indirect evidence.

283
00:33:15,930 --> 00:33:19,110
And we've analysed these pieces and none of them is very strong.

284
00:33:19,860 --> 00:33:24,390
They're all circumstantial or illogical in one way or another.

285
00:33:24,570 --> 00:33:28,020
I just want to put up the first one of these.

286
00:33:28,260 --> 00:33:36,300
It's the statement that women who have had a colectomy and ileostomy have no intrahepatic circulation because there's no de conjugation,

287
00:33:37,260 --> 00:33:40,260
no no excretion or whatever.

288
00:33:40,860 --> 00:33:47,280
That's supposedly the the idea. So you get drug.

289
00:33:47,620 --> 00:33:52,929
Conjugated drug excreted and not D conjugated because there's no de conjugation going on.

290
00:33:52,930 --> 00:33:57,219
There's no gut to do it, and yet the efficacy does not appear to be reduced.

291
00:33:57,220 --> 00:33:59,200
They say, well, actually that's not true.

292
00:33:59,200 --> 00:34:12,640
There's a a letter here and a study later showing that this does, in fact, reduce the amount of the amount of oestrogen that you have in your body.

293
00:34:14,440 --> 00:34:21,280
Equally important, and the reason I put this up is because it makes a statement that nobody, I think has considered,

294
00:34:22,030 --> 00:34:30,310
and that is that in these patients they studied who were given antibiotics, there was a change in sex, steroid binding globulin.

295
00:34:32,020 --> 00:34:43,000
Now, all the studies of oestrogens in these all the work that's been done in these studies has been on total oestrogen concentrations,

296
00:34:44,140 --> 00:34:48,640
not the unbound concentration. It's the unbound concentration that's active.

297
00:34:49,570 --> 00:34:54,070
So it's possible that you might have no change in total oestrogen concentration,

298
00:34:54,490 --> 00:34:59,080
but a change in a small fraction of unbound concentration, which would be important.

299
00:34:59,140 --> 00:35:02,200
Nobody has ever considered that as far as we can see.

300
00:35:03,520 --> 00:35:08,259
Incidentally, it's these oestrogens are not banned ethanol.

301
00:35:08,260 --> 00:35:13,360
These $2 are not bound to globulin, they're bound to see them albumin, but about 98%.

302
00:35:15,400 --> 00:35:21,220
So if I, I usually when I'm teaching students about this, I talk about my pocket money problem.

303
00:35:22,390 --> 00:35:26,140
My father comes home with £100 in his pay packet.

304
00:35:26,650 --> 00:35:31,330
Well, nowadays it's a thousand or whatever, £100 in his pay packet.

305
00:35:31,330 --> 00:35:37,150
He gives my mother £98 to run the house and he gives me £2 pocket money.

306
00:35:38,740 --> 00:35:42,879
One week he comes home, he says, Geoff, I'm sorry, we've got extra expenses.

307
00:35:42,880 --> 00:35:46,180
It's not very much. Your mother needs £99 this week.

308
00:35:47,290 --> 00:35:51,790
And I say, you what? You've just reduced my pocket money from £2 to £1.

309
00:35:52,680 --> 00:36:00,490
A 50% reduction in my pocket money for your £1 extra expense and the over 98.

310
00:36:01,270 --> 00:36:04,959
So that's devastating. Same is true of bound and unbound.

311
00:36:04,960 --> 00:36:15,020
If your drug is 98% pound, then a small change in the unbound can produce an enormous change in effect, and that could easily happen with ethanol.

312
00:36:15,020 --> 00:36:18,410
The stradale which is 98% band to see them out them.

313
00:36:18,430 --> 00:36:31,720
That has not, to our knowledge, been studied. Another important point about all these studies is that they have chosen to look at surrogate markers.

314
00:36:32,740 --> 00:36:38,740
They've looked at total ethanol oestradiol concentrations as a surrogate marker,

315
00:36:38,740 --> 00:36:43,090
although they don't seem to appreciate it for unbound ethanol oestradiol.

316
00:36:43,810 --> 00:36:51,430
And that in itself is a surrogate marker for the action of the oral contraceptive and for pregnancy, which is the important outcome.

317
00:36:51,700 --> 00:36:59,830
So nobody has done a study of pregnancy outcomes, so we thought we ought to do that.

318
00:37:01,150 --> 00:37:05,530
And we go back to the yellow card reports that I mentioned before with all the

319
00:37:05,530 --> 00:37:11,020
problems that they entail and now instead of sending in yellow cards by post,

320
00:37:11,320 --> 00:37:16,930
you can fill it in online. All the data have been gathered together, digitised, whatever,

321
00:37:16,930 --> 00:37:26,379
and you can get it all through the MHRA website and this website called Interactive Drug Analysis Profiles and the Internet.

322
00:37:26,380 --> 00:37:36,760
These this interactive drug analysis profiles tell you we have received 60,000 reports on this drug and there are 2000 cases of headache,

323
00:37:36,760 --> 00:37:39,970
3000 of this, 4000 to that 100 of the other.

324
00:37:40,390 --> 00:37:51,670
And you can add them all up and see if there are any differences. So we decided to look at unintended pregnancies as an outcome in this database,

325
00:37:52,510 --> 00:37:58,030
and we chose to amalgamate the data from nine broad spectrum antibiotics.

326
00:37:58,030 --> 00:38:06,370
There they are listed and we chose to comparing to competitor groups.

327
00:38:06,820 --> 00:38:13,600
We chose negative controls. We took nine drugs at random from a formulary.

328
00:38:14,260 --> 00:38:21,100
And these are all drugs that women of childbearing age might from time to time be expected to take for one reason or another.

329
00:38:21,340 --> 00:38:25,420
Citalopram is an antidepressant. Ibuprofen is an anti-inflammatory drug.

330
00:38:25,780 --> 00:38:35,950
Lansoprazole is for, let's call it dyspepsia and so on, paracetamol, theophylline for asthma, zolpidem to help you sleep, whatever.

331
00:38:37,690 --> 00:38:39,519
So these are negative controls.

332
00:38:39,520 --> 00:38:47,460
We would not expect these drugs to make you more susceptible to pregnancy when you're taking the oral contraceptive and the.

333
00:38:47,530 --> 00:38:50,650
Then we looked at a group of positive controls.

334
00:38:51,670 --> 00:38:56,020
These are the enzyme inducers that we know should increase your risk.

335
00:38:58,210 --> 00:39:07,870
So we expect these 12 drugs as a group to increase the risk of unintended pregnancies compared with the controls.

336
00:39:08,410 --> 00:39:13,600
That's our expectation. If we don't find that, then there's something wrong with our study.

337
00:39:15,130 --> 00:39:19,930
We don't know what the relationship between these two groups would be.

338
00:39:20,560 --> 00:39:31,090
So that's what we're trying to find out. And we are, in fact, trying to prove the null hypothesis that there is no interaction of antibiotics.

339
00:39:32,770 --> 00:39:37,840
With supposed oral contraceptives through the measure of unintended pregnancies.

340
00:39:40,450 --> 00:39:46,929
We've also looked at a range of events, adverse events, not merely unintended pregnancies.

341
00:39:46,930 --> 00:39:55,840
That's our outcome of interest. We've looked at two negative events negative control events, cardiac arrhythmias and headache.

342
00:39:56,140 --> 00:40:00,820
We do not expect any of the three groups to affect these at all.

343
00:40:01,750 --> 00:40:06,579
The negative controls we've looked at congenital abnormalities.

344
00:40:06,580 --> 00:40:14,300
This is a positive control. We expect that the enzyme inducers should increase the risk of congenital abnormalities.

345
00:40:14,320 --> 00:40:21,129
We expect that if we don't find it, then there's something wrong because not all of the enzyme induces,

346
00:40:21,130 --> 00:40:24,940
but quite a few of them are known to cause congenital abnormalities.

347
00:40:24,940 --> 00:40:32,620
So that's our positive control. And then we've got a confounder or a potential confounder in the mix diarrhoea,

348
00:40:33,160 --> 00:40:41,320
because if an antibiotic causes diarrhoea, it may hurry the pill through the gut and you don't absorb it.

349
00:40:41,680 --> 00:40:45,700
And diarrhoea is known to be a risk factor for failure of the pill.

350
00:40:46,630 --> 00:40:56,110
So if we see an interaction or an effect of diarrhoea, then that might be a confounder for our antibiotics.

351
00:40:56,110 --> 00:41:00,550
Effect might explain explain the apparent effect of antibiotics.

352
00:41:00,760 --> 00:41:05,980
It's due to diarrhoea, not due to the mechanism of inhibiting the conjugation.

353
00:41:07,210 --> 00:41:12,640
So that's why we chose all those and we could have chosen others. But there's only so much time in the day.

354
00:41:14,140 --> 00:41:17,800
So now what were the results? There they are.

355
00:41:18,460 --> 00:41:30,010
There's the controls. Just look at the unwanted pregnancies. Five out of 60,000 reports, five out of 60,000 unintended pregnancies, I should say.

356
00:41:30,010 --> 00:41:34,870
These these are data we got on our first passage through the the database.

357
00:41:35,830 --> 00:41:40,000
We've since updated them with the latest and the results are very similar.

358
00:41:42,550 --> 00:41:49,060
Here are antibiotics of 45 out of about the same number, 68,000.

359
00:41:49,180 --> 00:41:54,130
So that's many times more than with the controls.

360
00:41:55,300 --> 00:42:00,550
And here's our positive controls, 39 out of half the number.

361
00:42:00,880 --> 00:42:05,560
So try doubling that 78, roughly out of 60,000.

362
00:42:05,980 --> 00:42:08,050
So you're beginning to see the pattern emerge.

363
00:42:09,550 --> 00:42:25,030
You might also want to look at the congenital abnormalities 819 and I'll show you 179 with the antibiotics and 234 with the controls.

364
00:42:25,030 --> 00:42:28,240
So many more with the enzyme inducers as we expected.

365
00:42:28,630 --> 00:42:34,390
So here's the summary 545 and 39.

366
00:42:38,020 --> 00:42:41,710
Hard to know what's going on. Diarrhoea certainly isn't increased.

367
00:42:42,820 --> 00:42:47,170
We thought we were a little surprised. We thought the antibiotics might increase the risk of diarrhoea.

368
00:42:47,170 --> 00:42:54,639
It doesn't seem to why the enzyme induces a lower, I don't know, but I suspect that the confidence intervals are quite wide on that.

369
00:42:54,640 --> 00:42:59,740
I'll show you that. And nothing much again with the cardiac arrhythmias, the congenital ones,

370
00:42:59,740 --> 00:43:04,360
clearly a signal coming from the enzyme inducers and not much from the headache.

371
00:43:04,960 --> 00:43:07,450
So when we look at these per 100,000,

372
00:43:08,230 --> 00:43:17,440
we can see that the ratios for the antibiotics is about eight times the controls and 14 times for the enzyme inducers.

373
00:43:17,950 --> 00:43:22,630
So we have the expected effect with the enzyme inducers a very large effect.

374
00:43:23,320 --> 00:43:30,190
And that's the answer to our question about antibiotics, not as big an effect, but nonetheless an effect.

375
00:43:30,490 --> 00:43:34,270
And the confidence intervals on the raw data do not overlap.

376
00:43:34,270 --> 00:43:38,290
It's really quite striking. And here's the congenital abnormalities.

377
00:43:38,560 --> 00:43:46,870
The signal comes from the enzyme inducers and not from the antibiotics as expected, which is what it should be.

378
00:43:47,940 --> 00:43:50,950
And if we hadn't seen that, we would have doubted the results.

379
00:43:51,640 --> 00:43:56,740
But that's what comes out. And the others really are nowhere.

380
00:43:56,830 --> 00:44:00,580
The ratios are all under one, one or less.

381
00:44:02,890 --> 00:44:08,500
So we think that this is and there's the there's the data highlighted.

382
00:44:09,400 --> 00:44:18,459
So we think this is evidence of an interaction and certainly one that constitutes a signal that ought to be looked at further,

383
00:44:18,460 --> 00:44:23,980
perhaps in other databases. There are, of course, problems with the database.

384
00:44:24,430 --> 00:44:29,740
These are spontaneous reports, people just sending in reports when they see something happening.

385
00:44:30,130 --> 00:44:32,260
And there is undoubtedly going to be a risk of report.

386
00:44:32,430 --> 00:44:40,740
Bias, but we think that the positive and negative controls we've used minimises the risk of that.

387
00:44:40,830 --> 00:44:50,790
But we can't rule it out entirely. But the data do suggest that antibiotics increase the risk of unintended pregnancies.

388
00:44:52,410 --> 00:44:56,729
Certainly the evidence that there's no risk is very weak and in our view,

389
00:44:56,730 --> 00:45:03,810
ought to be disregarded and bone concentrations should be measured if you're using that particular surrogate marker.

390
00:45:05,250 --> 00:45:12,750
And whatever you think, actually, it's our view that the precautionary principle should apply.

391
00:45:13,890 --> 00:45:17,930
And there is the precautionary principle. I'll read it out.

392
00:45:17,940 --> 00:45:26,070
If an action or policy has a suspected risk or suspected risk of causing harm and the absence of consensus.

393
00:45:27,030 --> 00:45:30,440
The burden of proof falls on those taking that action.

394
00:45:30,450 --> 00:45:40,530
In other words, the burden of proof that there is no interaction falls on the obstetricians and while they cannot prove it definitively.

395
00:45:41,400 --> 00:45:46,240
The precautionary principle, in our view, should apply. QED.

396
00:45:46,260 --> 00:45:46,650
Thank you.