← More or Less
More or Less
Is salt bad for you?
00:27:56
More or Less has the latest on salt, 'zero tolerance' policing, and how to predict the...
Loading summary
Transcript68 lines
- [00:00]
Tim Harford
Thanks for downloading the More or less podcast from BBC Radio 4. Here's Tim Harford. Hello and welcome to More or Less. Whether you love numbers or just want to be friends with them, we're the programme for you. And as always, we'll be turning to the numbers for insights about the news and, more importantly, the world. Later, we'll ask whether zero tolerance policing really works.
- [00:22]
Bernard Harcourt
We call it Newton's Law of Crime. What goes up must come down, and what goes up the most tends to come down the most.
- [00:30]
Tim Harford
And we'll ask whether it's possible to predict the future height of growing children.
- [00:35]
Charlotte Wright
The best predictor of what height an individual is going to be is the height they are in childhood.
- [00:40]
Tim Harford
But first, let's talk about the most delicious condiment in existence salt. Last month a Daily Express headline declared,
- [00:47]
Charlotte Wright
now salt is safe to eat.
- [00:50]
Tim Harford
And that sentiment was echoed in the. I think it's fair to say more academically minded Scientific American magazine, which is which said, it's time to end the war on salt. The zealous drive by politicians to limit our salt intake has little basis in science. These and similar reports cited a Cochrane Collaboration review which, in its own words, set out to assess whether advice to cut down on salt in foods altered our risk of death or cardiovascular disease. The Cochrane Collaboration is a highly respected network of researchers which evaluates the evidence for medical interventions. Their review was a meta analysis of seven studies, each of which was a randomised controlled trial or RCT. RCTs are considered to be the best kind of study for medical research, and the plain English summary of the Cochrane review itself does seem to support the Daily Express headline writer.
- [01:43]
Graham McGregor
Cutting down on the amount of salt has no clear benefits in terms of
- [01:47]
Tim Harford
likelihood of dying or experiencing cardiovascular disease. All this might surprise you. Haven't public health bodies been telling us to eat less salt for years? In Britain? Nice. The National Institute for Health and Clinical Excellence wants people to reduce their intake of salt to 6 grams per day per adult by 2015 and to 3 grams per day by 2025, from over 8 and a half grams a day at the moment. So has the Cochrane Collaboration paper really challenged this advice? The paper was published in the American Journal of Hypertension, which is edited by Dr. Michael Alderman. Dr. Alderman thinks it has.
- [02:27]
Michael Alderman
Well, I think it's important because they attempted to assemble all of the available randomized clinical trials. That's the highest level of evidence for medical interventions. They were only able to find seven, and in analyzing those as a group, in a process called meta analysis, they were Unable to find sufficient evidence to determine whether not reducing sodium intake, which is the issue, would be safe or beneficial.
- [02:58]
Tim Harford
We spoke to Rod Taylor, one of the authors of the Cochrane analysis, and he very clearly said he thinks people should cut down on their salt and this shouldn't be interpreted as meaning that salt is harmless.
- [03:12]
Charlotte Wright
What do you think?
- [03:12]
Michael Alderman
Well, I mean, I think everybody's entitled to their opinion. I think the fact of the matter is that the findings of the analysis were there's not evidence of harm or benefit. You know, when you're talking about going from the 3,500 milligrams a day that we in the United States and Great Britain eat now down to levels of 2,300 or 1500, that's cutting it by more than half. It seems to me that there has to be really solid, powerful, overwhelming, robust information that this is going to be beneficial and safe.
- [03:48]
Tim Harford
I'm going to be talking to Professor Graham McGregor. He's been using mathematical modeling and he reckons that cutting down salt intake could save many thousands of lives. I mean, do you dispute the methods he's using?
- [03:59]
Michael Alderman
Those models, of course, have been used for years and years and it's traditionally in science that's a technique we say, as Professor McGregor and others have done. Since we know that reducing sodium intake by half will lower average blood pressure by a couple of millimeters of mercury, and we know that a difference of a couple of millimeters of mercury in a population will prevent so many reduced deaths by so much, they believe it's inevitable that salt reduction of this amount will save those lives. But that's a hypothesis. It has to be tested. And the reason the hypothesis is so weak in this particular case is that in addition to lowering blood pressure a little, a reduction of sodium by half will also increase sympathetic nerve activity. It'll increase aldosterone secretion, increase insulin resistance and activate the renin angiotensin system and other enzyme system in the body. All of those things are negative, increase the risk of cardiovascular events. The health effect of reducing sodium is going to be the net impact of all of those factors. That's why you have to do randomized clinical trials.
- [05:13]
Tim Harford
That's Dr. Michael Alderman, editor of the American Journal of Hypertension. Now, as we researched this item, it became increasingly obvious to us that the question of salt reduction is hotly contested. There are in fact lobby groups on both sides. On the pro salt side stands the Salt Institute, which describes itself both as a non profit salt industry trade association and the world's foremost source of authoritative information about Salt. The anti salt lobby complains that Michael Alderman has an association with the Salt Institute. Dr. Alderman explained that association to us in an email. He I first attended a meeting of a scientific advisory committee of the Salt Institute in 1995 or 6, I believe. I received a $750 honorarium. I attended most subsequent annual meetings until, I think, 2008, for which I received no further honoraria. I have received no speaking fees or other individual support from the Salt Institute or any commercial interest for me or my research. The bulk of my research is NIH supported and NIH is America's Health Research Agency, a part of the US Government. For another view of the Cochrane paper, I spoke to Graham McGregor, a professor of cardiovascular medicine at Queen Mary University of London. Professor MacGregor lobbies against salt as chairman of Cash Consensus Action on Salt and Health.
- [06:40]
Graham McGregor
I think on initial reading I was flabbergasted that they could, from a few studies which were not designed as outcome studies, come to conclusions which were wrong, that salt reduction did not cause a reduction in strokes and heart attacks. And when we started to look at the individual studies, it was clear that this was a statistical misinterpretation. I mean, if you excluded one study, it was in patients with heart failure who were on huge doses of tablets and treatment and were very ill. And they're not relevant to the public health aspects of salt reduction studying that there were reductions in mortality, there were reductions in stroke and heart attack mortality and reductions in stroke and heart attack events within their analysis, but they weren't significant. Now, to anyone, even if when you
- [07:32]
Tim Harford
say significant, do you mean that they weren't of practical significance?
- [07:35]
Graham McGregor
No, the reductions were not statistically significant. That is they failed to reach a 5 in 100 chance of it happening by chance. That's the degree of significance we look for. And they weren't significant. Based on their results, you can say nothing. You can just say, well look, it looks interesting. It looks as though salt reduction does lower stroke and heart attacks and mortality. But it wasn't significant because of the smallness of the studies and the number of people involved. And even on their own figures they worked out, you needed 28,000 people randomized in a study for five years on a low and high salt intake to come to some conclusion. And I think there were only about 6,000 patients in the total study anyway. Now when we then re looked at the results and reanalyzed them, where we combined the normal people with normal and high blood pressure together, which is something you do anyway, because where you define High blood pressure is entirely arbitrary. We actually found, contrary to what they were claiming, that there was a significant reduction in stroke and heart attacks. So a 2 gram reduction in salt intake did cause a 20% or 1/5 reduction in strokes and heart attacks, which is a huge amount. And the results really, that they claimed are completely the wrong way around. A. There were reductions in everything they looked at. When we looked at them, combining the two groups together, they were significant.
- [09:01]
Tim Harford
Nevertheless, do you think we really have enough evidence for public bodies to be lecturing people on reducing their salt intake?
- [09:07]
Graham McGregor
We're not lecturing people. We are only putting forward the huge amount of evidence. We have more evidence about salt than we do for tobacco. We have epidemiological studies studying different communities of the world. We have migration studies where people have moved from a low to a high salt environment. All of the animal studies show that every animal, in order to develop high blood pressure, has been given a high salt intake. We have evidence in chimpanzees that giving them the sort of amounts of salt puts up their blood pressure. We have genetic studies in humans. We also have treatment trials which clearly show that when you reduce salt intake, you it reduces blood pressure. Blood pressure is the biggest cause of death in the Western world and the developing world from the strokes and heart attacks it causes. No. The case for salt is, in our view, sufficient to take public health action.
- [09:58]
Tim Harford
Professor Graham McGregor, both Michael Alderman and Graham MacGregor have strong positions on this issue. Does that mean we should take them both with a pinch of salt? Well, it's certainly worth finding out what others think. I and there is a degree of consensus among policymakers, at least, that we should reduce our salt intake. We've decided to give the last word to Rod Taylor, professor of Health Services Research at the University of Exeter. Professor Taylor was the lead author of the Cochrane study which triggered this latest row over the evidence.
- [10:29]
Rod Taylor
What our results show is that reducing salt reduces blood pressure. What we were unable to demonstrate was a significant improvement in either mortality or the cardiovascular events.
- [10:42]
Tim Harford
When you say significant, do you mean statistically significant or practically significant?
- [10:47]
Rod Taylor
Statistically significant. We found a large effect for the reduction of cardiovascular events, so, in other words, very practically of potential benefit. But it didn't reach what we call statistical significance.
- [11:03]
Tim Harford
I'm just looking at the plain language summary of your research from the Cochrane Collaboration website and in big bold letters it says, cutting down on the amount of salt has no clear benefits in terms of the likelihood of dying or experiencing cardiovascular disease. It does really sound as though you're saying there's no harm in salt.
- [11:23]
Bernard Harcourt
Yeah.
- [11:24]
Rod Taylor
No, I think in retrospect we acknowledge that maybe the plain language summary is not as helpful as it could be. And we've actually redrafted the plain language summary. So can I read you what the headline now reads?
- [11:37]
Tim Harford
Please do.
- [11:38]
Rod Taylor
Advice to reduce the amount of salt eaten reduces blood pressure. But there is insufficient evidence to confirm the predicted reductions in people dying prematurely or suffering cardiovascular disease. And we think that's a much better and more clearer interpretation of our findings.
- [11:57]
Michael Alderman
Okay.
- [11:57]
Tim Harford
It's not just the plain text summary. I mean, I'm just looking at the version of your report that was published in the American Journal of Hypertension, says we found no strong evidence that salt reduction reduced all cause mortality or CVD morbidity in normotensives or hypertensives. In other words, no evidence that salt reduction improves your health.
- [12:19]
Rod Taylor
That's exactly correct. That's exactly what we said. But the important thing is the detail in that term, no strong evidence. So no strong evidence is different to no evidence.
- [12:29]
Tim Harford
Dr. Alderman basically thinks that what you've found really indicates that campaigners against salt don't have enough evidence to back them up. And we need much better evidence before we lecture people on reducing their salt. Do you have sympathy with his view?
- [12:45]
Rod Taylor
I think it's important that people see our research in the context of the totality of the evidence. There is a large body of other sorts of evidence, evidence from animal studies, evidence from non randomised studies, that when one puts our study together in the context of all of the evidence, we would agree with Professor MacGregor that there probably remains sufficient evidence to continue to take public health action here.
- [13:17]
Tim Harford
You may recall Graham MacGregor saying that Taylor's team had made statistical errors and when he reanalysed the data, he found that in fact low salt diets did save lives on a statistically significant basis. He wrote up his results, didn't share them with Rod Taylor or his team, and sent them off to the Lancet. Rod Taylor wasn't impressed. Professor McGregor has actually re analyzed your data and combined people with normal blood pressure and high blood pressure. And he reckons once you do that actually there is statistically significant evidence that salt reduction causes a reduction in strokes and heart attacks and so on. What do you make of his reanalysis?
- [13:55]
Rod Taylor
We firmly believe that combining evidence on healthy and hypertensive populations is not appropriate in this case. There's a five fold difference in the risk of developing cardiovascular disease if you have diagnosed high blood pressure versus not having diagnosed high blood pressure. So we think combining those two bits of evidence is like combining apples and oranges. And we would argue that that really breaks a central tenet of statistical analysis.
- [14:23]
Tim Harford
Next, we move to the question of statistical significance. In this clip, you'll hear Rod Taylor talking about a P value. Just to explain that briefly and a little loosely, this is the probability that the results might have emerged purely by chance, assuming, in fact that salt has no effect on health at all. P of.01 means a probability of 1%. A P of.05 means a probability of 5%. By convention, and it is just a convention, a p value of 0.05 or lower is regarded as statistically significant. A higher P value is not statistically significant. But of course, that does make statistical significance a high bar to jump. In some cases, you can be 93% confident that the results didn't happen by chance and still not reach statistical significance. And in fact, Rod Taylor found himself in exactly that position.
- [15:13]
Rod Taylor
The P value for memory in the MacGregor study, the probability that the result was indeed true was a p value of 0.04. In other words, there was a 4 in 100 chance that the result was due to chance, whereas our p value is 0.07. In other words, a 7 and a 100 probability that the result was due to chance. Now, are you saying that those results are that different? Well, of course they're not. And the question that you asked me earlier about practical significance as opposed to statistical significance, in both cases, the average result shows here, whether you look at the McGregor analysis or the Taylor analysis, if I can call them such, that both analyses show that on average, there's approximately a 20% reduction in cardiovascular events, no matter how you slice and dice the data. So we're agreeing.
- [16:05]
Tim Harford
Professor Rod Taylor from the University of Exeter. One of the stories of the week was the arrival of American super cop Bill Bratton as an advisor to the Prime Minister. In the wake of the recent riots and looting. Bratton was even touted to fill the vacant job of Commissioner of the Metropolitan Police, although in the end, the government ruled out appointing a foreign national. We were curious about the statistical evidence on Bill Bratton's record as the Chief of Police, first in Boston, then New York and then Los Angeles. What did he actually do? Did it work? And if it did, did it work for the reasons Bill Bratton's supporters claim? Bratton was most famously associated with three things. First, there's the broken windows theory of crime, which is that graffiti, fair dodging and other misdemeanors encourage A sense of permissiveness in which much more serious crimes are then committed. Second, there's the idea that zero tolerance policing is the answer to that. And third, the fact that on Bill Bratton's watch, crime rates fell through the floor. I spoke to Professor Bernard Harcourt of the University of Chicago to run us through the evidence. Professor Harcourt began by pointing out that Bill Bratton was a top cop in Boston, New York and then Los Angeles at a very favorable time.
- [17:20]
Bernard Harcourt
Crime has fallen across the country in the United states since about 1991 in absolutely remarkable ways. For more than two decades now, the crime rate, homicide rate in the United States has plummeted.
- [17:35]
Tim Harford
So clearly, one can't just naively say, well, Bill Bratton took over in New York, crime fell in New York, he took over in la, crime fell in la. Therefore, Bill Bratton is a policing genius. We've got to look a little bit more closely at these numbers. So what are the different possible explanations?
- [17:51]
Bernard Harcourt
Well, there have been huge national level factors that have affected the rate of criminality in the United States over the past two plus decades. And some of those factors have been mass incarceration. Actually, mass incarceration is probably the leading factor we now incarcerate in the United States, 1% of our adult population.
- [18:11]
Tim Harford
Couldn't we credit Bill Bratton with that? I mean, if he's advocating tough policing, surely tough policing means more people going to prison.
- [18:19]
Bernard Harcourt
It's interesting because to a large extent, broken windows policing was embraced by progressives. The idea was that instead of engaging in these very harsh sentencing policies, we could just get rid of minor disorder. And then there are other factors that contribute as well. Changes in drug use patterns across the country. So the decline of crack cocaine, for instance, has been an important factor. Another factor that's been identified by some of the criminologists is the raw number of police officers employed.
- [18:53]
Tim Harford
This is very relevant in the UK case since the government plans to cut the number of police officers.
- [18:58]
Bernard Harcourt
Right. Indeed. What the research seems to indicate in an interesting way is that the strategies that are employed are not themselves so important as the mere fact of the raw numbers of police officers.
- [19:13]
Tim Harford
Let me just push back a little bit. Defend Bill Bratton for a moment because he did increase police numbers, but not that much in Los Angeles. And sure, crime fell across the United States, but it seemed to fall particularly sharply in New York. So, I mean, is it fair to say he gets no credit at all?
- [19:33]
Bernard Harcourt
Okay, let's go to New York for a second. The drop in crime in New York City from 1993-96. Those were the first years of broken windows policing. See declines in homicide rates and robbery rates that are high. 70% declines in homicide, 60% declines in robbery. Now, in Los Angeles during the exact same period, Note Bratton's not in Los angeles yet. It's 10 years later that he'll get there. But what's going on in LA during this time? It's a defunct police department that's on its way to being in receivership under federal court supervision. And yet robbery rates were declining at faster rates than they were in New York City. Now, there's a reason that a big city like New York City and Los Angeles, which were former homicide capitals, saw some of the sharper drops in crime. And a colleague of mine, Jens Ludwig, and I explain it in a statistical study that we conducted in 2006. We call it Newton's Law of Crime. What goes up must come down, and what goes up the most tends to come down the most.
- [20:38]
Tim Harford
I think more or less listeners are reasonably familiar with this idea that basically there are always certain transitory factors at play.
- [20:45]
Charlotte Wright
And.
- [20:45]
Tim Harford
And if those transitory factors come in and make things really bad in a particular location, if you come back a few years later, that location will appear to have improved.
- [20:55]
Bernard Harcourt
Right. That's reversion to the mean, and that's what we would expect. And in part, it explains why large cities like New York and like la, that saw the sharpest increases in homicides during the bad times that became the murder capitals of the US, would also see the sharpest drops in crime when the entire country sees a reduction in criminal offending.
- [21:24]
Tim Harford
So, thinking about everything you've told us, is there any statistical evidence that Bill Bratton is a super cop?
- [21:32]
Bernard Harcourt
There's no reliable statistical evidence that he's a super cop. You know, I just opened up the LA Times the other day. Violent crime in Los angeles fell almost 10% for the first six months of 2011. That's remarkable. Bill Bratton isn't the chief there anymore.
- [21:50]
Tim Harford
Bernard Harcourt of the Law School of the University of Chicago. You're listening to More or Less with me, Tim Harford, in association with the Open University. You can contact the programme by emailing us at more or lessbc.co.uk Andrew Porter did just that after I wondered aloud how big whales would be if the crinkly bits were flattened out. If one considers that the steepest gradient is of the order of one in three, then a reasonable average gradient is
- [22:19]
Rod Taylor
likely to be lower than 1 in 6.
- [22:21]
Tim Harford
Arctan of 1 over 6 is less than 10 degrees. Cosine of 10 degrees is around 0.985, so the hypotenuse is just 1.5% longer than the base. So flattening Wales would increase its area by about 3%. Wow. Well, I never did quite master trigonometry at school. Always more of a statistics man myself. But thanks to Andrew, and please keep your emails coming. We were recently asked a question by a rather short man who is married to a rather tall woman. He was wondering whether, as an old piece of homespun wisdom claims, sons are always taller than their mothers, in which case his two boys will grow to be big, tall lads. But is there any truth to this or is it just a tall tale? Well, the medical statistician, Professor Tim Cole, set about trying to predict the likelihood that a son would be shorter than his mother. One way to do this is to take a theoretical approach based on our underlying knowledge of how common it is for men and women to be of any given height. You just look at the statistical distributions of heights for men and women, figure out the chance of getting a short enough man and a tall enough woman. One complication is that the heights of parents and children are correlated. But Professor Cole took that correlation into account and did his sums. With this result.
- [23:37]
Rod Taylor
There is a chance of 1 in 61 of the man being shorter than his mother. That's 1.6%.
- [23:46]
Tim Harford
You said this is a theoretical approach. Presumably you could go and just grab a big bunch of data and find out that way. Has anybody done that?
- [23:55]
Rod Taylor
It could be done, and in fact, I've just done it, or rather my colleague has just done it for me this morning. And we've looked at data from the 1958 birth cohort, where there are about six and a half thousand boys who, as men were measured at age 33, and their heights can be compared with their mothers when the mothers were measured in pregnancy. And using those figures, we discover that the number of men shorter than their mothers is about 106,500, which is again a 1.6% figure, which is almost exactly what I arrived at with my theory, which is rather serendipitous.
- [24:37]
Tim Harford
There are two reasons why Professor Cole was lucky here. The first is that you just wouldn't expect theory and practice to produce exactly the same result quite so elegantly. And the second is there's an additional complicating factor which Professor Cole left to one side in giving us his figure of 1 in 61. This complication is that the human race is getting taller. This is known as a secular trend. In his original sums, Tim Cole was assuming that the child's generation had the same average height as the parents generation. What he needed to do was to factor in that the child's generation is taller on average by between 1 and 3 cm, depending on which country you're looking at. This is what happened when he factored that in.
- [25:18]
Rod Taylor
If we introduce a slight secular increase in height from generation to generation, the chance for the sun it goes down to 1.1%.
- [25:28]
Tim Harford
Professor Tim Cole of University College London. So theory and evidence didn't overlap quite as elegantly as it first appeared. Phew. Normal services resumed. So just to reiterate Professor Cole's analysis, There's nearly a 99% chance that a son will be taller than his mother. And for those of you who are interested, there's a 6% chance that a daughter will be taller than her father. More generally, how do doctors estimate how tall a child is eventually going to be? I spoke to Charlotte Wright, a professor of community child health at the University of Glasgow and a consultant pediatrician who's been working on this question with Tim Cole. They argue that looking at a child's parents height, they may actually be a distraction.
- [26:12]
Charlotte Wright
A technique that has been widely used in paediatrics is to compare children's expected height to their parents heights. And unfortunately it's rather developed into an idea that you can predict a child's height by looking at the parents heights, when in fact it really should be used to test whether a child's expected height is within limits given how tall the parents are, which is a slightly different question.
- [26:35]
Tim Harford
So what you're saying is you shouldn't sort of say, well look, dad's six foot three and so the son's going to be a big strapping lad. You should say, well look, the son's 7 years old, he's this height, that's actually not that high for a 7 year old. And okay, he might be a slow developer, but basically looking at his height as a seven year old is probably going to tell you more about how high he's going to be, how tall he's going to be as an adult than looking at his father's height.
- [27:00]
Charlotte Wright
Absolutely. I mean, if you think about it, anybody who's seen a family of children grow up, adults think about their own siblings, you know that you're all different heights. A family of boys, a family of girls, they're all different heights. So that just because their parents are a set height, it doesn't mean you're all going to be the same height. There's going to be a very widespread. And the best predictor of what height an individual is going to be is the height they are in childhood, because children basically do tend to track along roughly the same trajectory.
- [27:27]
Tim Harford
Professor Charlotte Wright of the University of Glasgow. That's all we have time for this week, I'm afraid. But do please keep your suggestions coming in via Twitter or More or less@BBC.co.uk. our website is BBC.co.uk more or less. And we'll be back next week. Until then, goodbye. More or Less was presented by Tim Harford, the Financial Times undercover economist. The producer was Richard Knight. The programme was made in association with the Open University.