1
00:00:05,910 --> 00:00:09,510
Welcome to Science with Sanjula, where we talk about anything global health.

2
00:00:10,140 --> 00:00:13,950
My name is Sanjula and I am a researcher at the University of Oxford.

3
00:00:14,550 --> 00:00:19,650
Join me as I speak to world leading scientists who tackle today's biggest challenges in health care.

4
00:00:20,250 --> 00:00:26,400
My first guest is Professor Sir Richard Peto, one of the most prominent epidemiologists of the last century.

5
00:00:27,000 --> 00:00:32,160
Professor Peto spent his impressive career investigating why people die prematurely.

6
00:00:33,030 --> 00:00:37,590
He's an Emeritus Professor of Medical Statistics and Epidemiology at the University of Oxford,

7
00:00:38,040 --> 00:00:42,990
and he even invented the meta-analysis and a theory known as Peto's Paradox.

8
00:00:43,740 --> 00:00:46,770
Professor Peto, first of all, may I call you Richard? Please.

9
00:00:47,310 --> 00:00:51,210
You mentioned that you fell in love with science at school at Southampton.

10
00:00:51,240 --> 00:00:55,950
Oh, yes. And further on, you actually fell out of love with science at the University of Cambridge.

11
00:00:55,980 --> 00:00:59,160
Well, I was still in love with it, but I certainly fell out of doing it well. 

12
00:00:59,390 --> 00:01:04,140
When I was at school, it was just so exciting. It was an ordinary state school,

13
00:01:05,220 --> 00:01:11,310
but the teachers were really interesting and trying to get across not the stuff in the O-level and the A-level curriculum,

14
00:01:11,640 --> 00:01:17,400
but what they found interesting and they did, they just put so much effort into the ways they taught.

15
00:01:17,400 --> 00:01:18,420
It was really excellent.

16
00:01:18,870 --> 00:01:30,910
And I did very well at school, got a good scholarship to Cambridge, and then somehow Cambridge was much less exciting to me in comparison with that.

17
00:01:30,930 --> 00:01:35,999
I don't know why. I really don't know why. I can't understand those years. But it was just a letdown.

18
00:01:36,000 --> 00:01:39,959
After the excitement of school, I wanted that same excitement to carry on and it didn't.

19
00:01:39,960 --> 00:01:44,010
And probably the fault was more in me than in Cambridge.

20
00:01:44,010 --> 00:01:50,510
I don't know. I can't really sort it out. But anyway. So I dropped completely out and finished up, you know, really failing at the end of three years.

21
00:01:50,520 --> 00:01:54,270
I mean, that's not not getting a first, second, third or even selected resit examination,

22
00:01:54,270 --> 00:01:58,470
and really failing at the bottom of the whole university.

23
00:01:58,860 --> 00:02:02,130
And they said to me, look, this isn't what should be happening to you.

24
00:02:02,580 --> 00:02:08,660
Come up during the summer holidays and we'll teach you some real mathematics.

25
00:02:09,440 --> 00:02:13,819
And so I did. And that really helped to get me interested again.

26
00:02:13,820 --> 00:02:18,260
And I just sort of bounced back off absolute bottom into being pretty useless.

27
00:02:18,860 --> 00:02:26,269
So at the end of the fourth year, this extra year, I hadn't done terribly well but thought, well, I'd like to stay on and maybe do an MSc at Cambridge.

28
00:02:26,270 --> 00:02:32,059
And they said, well, look, I know it sounds odd to say when you failed, you should stay on for a year and now you've partly succeeded, you should go.

29
00:02:32,060 --> 00:02:35,180
But I think it's time that you left Cambridge.

30
00:02:35,600 --> 00:02:43,309
So I did and just was fortunate enough to stumble into the statistics course that was being taught by David Cox, who really is a remarkable statistician.

31
00:02:43,310 --> 00:02:49,340
In one of your previous interviews, you've said that you had a blue motorcycle and long hair.

32
00:02:50,000 --> 00:02:55,100
Yeah, yeah, yeah. During your time at Imperial College in London.

33
00:02:55,460 --> 00:02:58,910
But as you said, you also managed to graduate. You were 24 years old.

34
00:02:58,910 --> 00:03:05,840
You just graduated from your Masters of Science at Imperial College, and you went on to get an interview with Richard Doll.

35
00:03:06,590 --> 00:03:14,840
And when he asked you why you wanted to work with him, you famously said, I do not know if I want to work with you.

36
00:03:14,840 --> 00:03:19,060
And actually I do not know if I want to work at all. Well, it was silly.

37
00:03:19,070 --> 00:03:23,360
I don't know what sort of world I was living in, what I thought I was going to be living off.

38
00:03:23,750 --> 00:03:27,800
But, you know, the sun was shining all the time. Even when it was raining. I got a big 650 motorbike.

39
00:03:28,070 --> 00:03:30,680
It was London in the 1960s, it was just such a lovely...

40
00:03:30,680 --> 00:03:36,230
The idea of just being somebody that went to the office and worked seemed very dismal in comparison.

41
00:03:36,680 --> 00:03:42,650
But when I started working with him, once you get your first result, you're addicted,

42
00:03:42,950 --> 00:03:49,160
even though in retrospect it's not particularly important, but you start getting results and that's very addictive.

43
00:03:49,400 --> 00:03:55,100
But the first ten or more years I was working with Richard Doll, actually first ten years I was working for Richard Doll,

44
00:03:55,550 --> 00:04:00,620
I got to the point where I could give good lectures, which some of the stuff I'd done was,

45
00:04:00,920 --> 00:04:05,899
was original, but nothing had ever actually been of any real use to anybody.

46
00:04:05,900 --> 00:04:08,480
I mean, nothing I've done had ever saved any lives at all.

47
00:04:08,990 --> 00:04:12,800
I've worked on trials and shown that if you do them properly, then the treatments didn't work.

48
00:04:13,280 --> 00:04:19,430
I've worked on, you know, various epidemiological things which had already been studied by other people.

49
00:04:19,910 --> 00:04:25,340
And I remember saying to Richard in the seventies, that after I'd been working more than ten years that nothing I've done saved any lives.

50
00:04:25,340 --> 00:04:31,399
He said, well, I think it's too soon to tell. You know, maybe some of the things that you've done will turn out to be of value.

51
00:04:31,400 --> 00:04:34,400
And that did turn out to be the case in the 1980s.

52
00:04:34,400 --> 00:04:39,080
We started to get results that really did save a lot of lives and did influence the way things were done.

53
00:04:39,440 --> 00:04:41,060
And of course, that makes life much easier.

54
00:04:47,070 --> 00:04:53,730
Now I would like to invite you to give a mini lecture of about one minute about any scientific topic you are most passionate about.

55
00:04:54,180 --> 00:04:58,100
Whenever you're ready. Please go ahead. Yeah.

56
00:04:58,110 --> 00:05:01,380
Well, death in old age is inevitable. Death before old age is not.

57
00:05:01,530 --> 00:05:06,300
That's the slogan we've got printed up on the wall in the Richard Doll building.

58
00:05:06,720 --> 00:05:12,480
And so we're trying to say, why is it that people die before, I'll just say age 70 arbitrarily?

59
00:05:13,200 --> 00:05:19,180
I'm 78, so I like talking about death before age 70. It means that I'm safe and you're not.

60
00:05:19,770 --> 00:05:23,390
Why do people die before the 70, well, how many deaths are there before age 70?

61
00:05:23,400 --> 00:05:27,540
In the world, about 30 million deaths a year and

62
00:05:28,500 --> 00:05:35,700
the death rates among people of a given age in general are going down, but the population is going up.

63
00:05:36,060 --> 00:05:38,310
And so that number of about 30 million a year,

64
00:05:38,310 --> 00:05:46,740
that'll probably continue to be true through the 2020s because the decrease in death rates are counterbalanced by the increasing population.

65
00:05:47,880 --> 00:05:58,170
And most of these deaths now are from non-communicable diseases, heart disease, chronic lung disease, cancer and the big causes of these are

66
00:05:59,880 --> 00:06:06,810
smoking. Blood pressure, blood lipids. And I don't know quite how to put in diabetes and obesity.

67
00:06:07,350 --> 00:06:18,150
Overweight obesity is a cause of high blood pressure, of dyslipidemia, diabetes, smoking, obesity.

68
00:06:18,810 --> 00:06:26,010
blood pressure, blood cholesterol. But smoking is the big one in general.

69
00:06:26,190 --> 00:06:30,780
In some populations, alcohol becomes very big. I mean, in Russia, particularly in the 1990s.

70
00:06:30,930 --> 00:06:37,230
It was this fast increase in deaths from vodka. And that's subsided, but it's still very high.

71
00:06:37,710 --> 00:06:39,630
I think in every interview I've listened to,

72
00:06:39,630 --> 00:06:48,750
you tend to quote that a moderate reduction in a big cause of death may be more life saving than a big reduction in a small cause of death.

73
00:06:49,200 --> 00:06:57,090
You were actually one of the first ones to, well, I might say, invent a new research technique called a meta-analysis.

74
00:06:57,780 --> 00:07:05,460
It's this business of getting all the trials, not taking just the famous ones, because the famous ones will be the ones with atypical results.

75
00:07:05,640 --> 00:07:14,150
But all of them, really find out all the randomised trials that have ever been done, get results from each of them, get them right and then add them up.

76
00:07:14,160 --> 00:07:19,410
Now this works only if you get really big large scale evidence in the process.

77
00:07:19,770 --> 00:07:23,009
How big should the trial be? Well, it depends on the question.

78
00:07:23,010 --> 00:07:26,910
How many depends on how big the treatment effect is. If you've got very striking treatment effect,

79
00:07:27,330 --> 00:07:31,139
then you hardly don't need to randomise and you didn't need randomisation to show that

80
00:07:31,140 --> 00:07:34,680
cigarette smoking is an important cause of lung cancer because the effect is so big.

81
00:07:35,370 --> 00:07:38,880
But when you've got the you know...

82
00:07:40,200 --> 00:07:45,179
Mostly, we've got treatments that have moderate effects, especially treatments for cancer, treatments for heart disease,

83
00:07:45,180 --> 00:07:51,690
treatments for chronic lung disease treatment and treatments for T.B. If you've got two plausible treatments,

84
00:07:51,690 --> 00:07:57,090
the difference between them is unlikely to be big. If there's uncertainty about it, it's unlikely to be big.

85
00:07:57,390 --> 00:08:00,420
But then moderate differences can really matter.

86
00:08:00,720 --> 00:08:06,320
And if you want to work out how to treat 10 million people, then why not randomise 10,000?

87
00:08:06,660 --> 00:08:15,360
But people just weren't thinking of numbers like that. So trial sizes ought to be for many questions in the tens of thousands.

88
00:08:16,260 --> 00:08:20,310
And then there's another debate about so-called subgroup analysis.

89
00:08:20,610 --> 00:08:28,079
Actually, there's one good example of something you did, because I think at some point you were publishing a paper in The Lancet.

90
00:08:28,080 --> 00:08:33,420
Oh, I'll tell the story of that paper. The most beautiful trial results that we ever got, really, was in the 1980s.

91
00:08:33,840 --> 00:08:38,999
We wanted to test aspirin in the middle of an acute heart attack. You know, you've had a horrible heart attack.

92
00:08:39,000 --> 00:08:43,410
God, you know, ring the ambulance, raced into hospital. Get in the hospital a couple of hours later.

93
00:08:43,710 --> 00:08:47,130
Are they going to die today? Are they going to die tomorrow? Will they be dead this week?

94
00:08:47,550 --> 00:08:54,780
And actually, we wanted to do a trial where you just give the patient half an aspirin, half an aspirin a day.

95
00:08:55,110 --> 00:08:59,100
You know, most doctors thought this was ridiculous, but we persuaded them to do it anyway. So we did it.

96
00:08:59,310 --> 00:09:03,840
And so we got half an aspirin a day or dummy pill that looked like half an aspirin but wasn't.

97
00:09:04,260 --> 00:09:07,530
And we finished up with 1000 deaths in those who got the dummy pill,

98
00:09:07,530 --> 00:09:12,239
800 deaths and those who got the real pill, and you can't get 1000 versus 800 just by chance.

99
00:09:12,240 --> 00:09:20,040
It had to be real. But it's really important because it applies to anybody who's come into hospital with a heart attack all over the world.

100
00:09:20,040 --> 00:09:25,079
I mean, this could apply to millions of people a year, you know, maybe 100 million a decade.

101
00:09:25,080 --> 00:09:28,320
I don't know.

102
00:09:29,470 --> 00:09:33,670
You know, when we sent it to the Lancet it's, you know, absolutely delighted.

103
00:09:33,670 --> 00:09:37,030
Great. You know, we've just finally got a result that really is going to save lives.

104
00:09:37,030 --> 00:09:41,710
And also it costs nothing as a treatment. Half an aspirin a day really isn't that expensive.

105
00:09:42,850 --> 00:09:48,700
And the Lancet reviewer said, no, no. Well, okay, you've got this result, but now you've got to tell us who benefits.

106
00:09:48,700 --> 00:09:53,350
You must actually provide us subgroup analyses, 

107
00:09:53,620 --> 00:09:58,009
which we've written again and again that subgroup analyses are a reliable machine for producing false negative results.

108
00:09:58,010 --> 00:09:59,470
So if you've got something that works,

109
00:09:59,950 --> 00:10:08,020
then you can reliably produce a false negative result by doing enough subgroups and picking out something that seems not to work.

110
00:10:08,290 --> 00:10:11,379
And so we say on principle alone. By chance alone, you'll get...

111
00:10:11,380 --> 00:10:11,800
That's right.

112
00:10:11,800 --> 00:10:17,830
If you've got something that works equally for everybody and you do enough subgroup analyses, you're going to get false negatives.

113
00:10:17,830 --> 00:10:21,909
And those false negatives, they become urban myths.

114
00:10:21,910 --> 00:10:25,420
Doctors love something about particular patients. They don't want to know what works on average.

115
00:10:25,420 --> 00:10:30,190
They want to know what works for the individual. That's the phrase personalised medicine, they call it.

116
00:10:30,670 --> 00:10:37,569
And so The Lancet said that they would publish our paper only if we did various subgroup analyses.

117
00:10:37,570 --> 00:10:40,900
So we said no. On scientific principle, no. They said, well, we're not going to publish a paper then.

118
00:10:41,290 --> 00:10:43,959
And so we just compromised our principles and said, well, all right then.

119
00:10:43,960 --> 00:10:50,710
But we also subdivided the patients according to whether they were born under the mediaeval birth signs

120
00:10:50,710 --> 00:10:54,220
of Libra, Gemini or Capricorn, you know completely stupid stuff.

121
00:10:54,550 --> 00:10:57,880
So we took 12 subgroups.

122
00:10:58,240 --> 00:11:03,190
And of course with 12 subgroups it's easy even when you've got a beautiful result like that to find...

123
00:11:03,190 --> 00:11:07,960
And it turned out that if you were born under Libra or Gemini, then it seemed not to work.

124
00:11:08,200 --> 00:11:11,830
If you're born under Capricorn, then it seemed to halve your risk of death.

125
00:11:12,280 --> 00:11:17,409
And of course that's so stupid that people realised that this was nonsense.

126
00:11:17,410 --> 00:11:20,890
So we put that out as the first subgroup analysis of our paper

127
00:11:21,310 --> 00:11:25,330
and The Lancet, when we sent it in they said, well, you know, you've done the subgroups so the paper is acceptable now,

128
00:11:25,330 --> 00:11:28,570
But you've got to delete this stuff about astrology. That's not serious. We said no,

129
00:11:28,810 --> 00:11:31,870
that is the only subgroup analysis there that is serious.

130
00:11:31,870 --> 00:11:36,660
That is the only serious scientific subgroup analysis. And this was true.

131
00:11:36,670 --> 00:11:44,960
Anyway, they just published it. They published it with the birth signs? Yeah, 13 August 1988, page 5, the top thing.

132
00:11:45,000 --> 00:11:51,909
And that struck people and it's been used again and again this astrological

133
00:11:51,910 --> 00:11:56,649
subgroups but still, nearly every trial that's reported finishes up with loads of

134
00:11:56,650 --> 00:12:02,230
subgroup analyses. It's completely meaningless and it often gives false negative results.

135
00:12:08,800 --> 00:12:14,650
Hi. My question for Professor Peto is if we already know that tobacco is bad for people and we know it kills people,

136
00:12:14,950 --> 00:12:24,549
why would we still bother investigating tobacco? Still smoking is a cause of about 20% of all male deaths in middle age and old age in old age,

137
00:12:24,550 --> 00:12:31,720
20% of all female deaths in middle age and in old age, and actually about 25% of all cancer deaths.

138
00:12:32,230 --> 00:12:37,330
So still it's a quarter of all cancer deaths still in this country despite that two thirds reduction.

139
00:12:37,900 --> 00:12:43,690
Smoking is causing more cancer deaths than every other known cause of cancer put together.

140
00:12:44,050 --> 00:12:46,050
But you wouldn't think that if you look at newspaper coverage.

141
00:12:46,060 --> 00:12:53,500
Look at COVID, look at newspaper coverage of COVID over the last two years. Actually in Britain, tobacco over the last two years,

142
00:12:53,500 --> 00:12:58,210
2020, 2021, killed about the same number of people as COVID.

143
00:12:58,510 --> 00:13:07,660
We really do need to know the long term effects of smoking seriously, and we need to know the long term effects of stopping at various ages.

144
00:13:08,110 --> 00:13:13,180
And we've done studies in various other countries, India, China, various other populations.

145
00:13:13,720 --> 00:13:18,550
And you've got to get Chinese evidence on Chinese deaths to get the Chinese government to take it seriously.

146
00:13:18,940 --> 00:13:24,870
And at the moment in China there is about a million deaths a year from smoking, nearly all male rather than female.

147
00:13:25,750 --> 00:13:31,270
And because if you take, say, people born in the 1960s in China,

148
00:13:31,270 --> 00:13:38,510
then the proportion smoking was about 100 times greater in males and females.

149
00:13:38,530 --> 00:13:43,870
So you're saying the numbers are actually still going up in other places and therefore it's still important that we investigate tobacco.

150
00:13:43,870 --> 00:13:49,149
Well, in the 2030s there'll be about 2 million deaths per year from smoking in China.

151
00:13:49,150 --> 00:13:52,540
By mid-century it'll be about 3 million. That's partly population growth,

152
00:13:53,260 --> 00:14:01,389
but there's got to be systems set up that just routinely monitor the extent to which smoking is killing people in different populations.

153
00:14:01,390 --> 00:14:09,550
And this will help bring forward the time when more and more effective controls get put in place.

154
00:14:16,490 --> 00:14:22,040
A few years ago, you were actually diagnosed with stage four cancer. That's right.

155
00:14:23,180 --> 00:14:27,080
And actually you've told me this story before. Would you mind sharing that with us?

156
00:14:27,740 --> 00:14:32,820
When I started working on intestinal cancer, basically everybody with stage four cancer died of it.

157
00:14:33,320 --> 00:14:36,350
There's no stage five right? Right.

158
00:14:36,350 --> 00:14:40,909
And Jeremy, one of my good friends died a year before I was diagnosed with stage four.

159
00:14:40,910 --> 00:14:45,950
He was diagnosed with stage four and died of it a few months earlier. So, you know, he was one of my very close friends.

160
00:14:47,890 --> 00:14:56,110
And so basically I knew I was dying and they said, oh, will you take this chemotherapy?

161
00:14:56,110 --> 00:14:59,230
And I knew that chemotherapy didn't work from my previous days.

162
00:14:59,800 --> 00:15:08,550
And, you know, working on trials of colorectal cancer in days when chemotherapy didn't work, I sort of took it almost to be nice to the doctor..

163
00:15:08,570 --> 00:15:12,640
You know, he's such a nice doctor, it just seemed rude not to take his chemotherapy.

164
00:15:12,970 --> 00:15:14,530
I didn't really believe it was going to do any good.

165
00:15:14,530 --> 00:15:21,159
I mean, I basically knew I was dying and they took a chunk out of my cancer to see if it had got any special vulnerabilities.

166
00:15:21,160 --> 00:15:24,340
No, it didn't. And it had spread to my lungs.

167
00:15:24,340 --> 00:15:31,899
That's what makes it stage four. And so and actually before my Festschrift, about a week before my Festschrift, I'd been told,

168
00:15:31,900 --> 00:15:36,490
well, the metastases in your lungs are still growing despite all the chemotherapy.

169
00:15:36,940 --> 00:15:41,110
So I'm afraid that means that even if we remove them, there's going to be other disease elsewhere.

170
00:15:41,110 --> 00:15:46,809
And this is incurable. So that was what I was told a week before my Festschrift meeting when it came to Studying The Bleeding Obvious.

171
00:15:46,810 --> 00:15:50,440
But then I finished up getting my lung surgery, you know,

172
00:15:50,470 --> 00:15:57,130
to take out the bottom of the lung with where they were, these some things that had been cancer.

173
00:15:57,580 --> 00:16:01,159
And when they looked at it down a microscope, they found it actually wasn't cancer.

174
00:16:01,160 --> 00:16:03,129
So the CT scan had been misinterpreted.

175
00:16:03,130 --> 00:16:08,470
And what it was was fibrotic consolidation of where cancer had been but had been destroyed by the chemotherapy.

176
00:16:08,470 --> 00:16:11,080
So it was just scarring where the cancer had been.

177
00:16:11,470 --> 00:16:17,740
And we were all going to go off on a family holiday to Greece, it was going to be our last family holiday to get in a goodbye to life.

178
00:16:18,310 --> 00:16:23,200
And instead, I got this news three days before we went and so we went and it was just amazing.

179
00:16:23,200 --> 00:16:25,600
I woke up this morning in Greece.

180
00:16:26,140 --> 00:16:34,629
It was a gentle breeze, Greek sunshine there and me lying in bed next to partner and you know, with about ten other family members within 50 metres,

181
00:16:34,630 --> 00:16:39,730
you know, it's just I felt like Orpheus coming back up out of the underworld when he sees that first beam of light.

182
00:16:40,000 --> 00:16:47,650
It was wonderful. Well, and then as a final question on this podcast, we ask all the professors, what advice would you give?

183
00:16:48,040 --> 00:16:52,240
I think I'd say whatever you do, you need to study something

184
00:16:52,510 --> 00:16:58,330
That's a serious question. And what do you mean, what is a serious question?

185
00:16:58,360 --> 00:17:07,690
Well, I've defined serious as being substantially relevant to mortality, you know, being potentially an important cause of death.

186
00:17:08,020 --> 00:17:12,690
I've defined it as that. But there's things other than death that matter, those aspects of the quality of life.

187
00:17:12,700 --> 00:17:24,520
And you could study... if somebody could really understand musculoskeletal, you know, the impact, pain or I mean, more things like depression or,

188
00:17:24,580 --> 00:17:30,820
you know, mental health where I mean, apart from suicide, they may not be terribly relevant to mortality.

189
00:17:31,060 --> 00:17:37,840
But of course, mental patients are much more like to smoke and other people and they finish up with higher lung cancer rates.

190
00:17:38,800 --> 00:17:43,650
I think just study...

191
00:17:43,860 --> 00:17:53,310
Study things that are serious and/or study something is beautiful, you know, just because beauty is sort of.

192
00:17:54,380 --> 00:18:01,340
Scientific quality control and results from studies have beautiful questions.

193
00:18:01,880 --> 00:18:05,690
They may not be immediately relevant, but they might well be relevant in the longer term.

194
00:18:06,200 --> 00:18:11,560
And people who just study things because of the delight of the subject and.

195
00:18:12,770 --> 00:18:21,470
Often. You know, they often do find things which certainly change our scientific understanding and may in the long term be of benefit.

196
00:18:21,530 --> 00:18:25,730
Well, thank you very much, Richard. That was a very, very interesting interview.

197
00:18:26,870 --> 00:18:30,170
We've reached the end of the first episode of Science with Sanjula.

198
00:18:30,940 --> 00:18:36,460
Episodes will be released weekly on Tuesdays, so make sure to subscribe on your favourite podcasting app.