#14 Evidence for low fat diet and decreasing saturated fat – PURE Study

Almost all guidelines are recommending that we should reduce dietary fats and restrict saturated fats. The PURE study has called to question these recommendations and in this episode, we explore the evidence.

It’s very hard to find a guideline that does not recommend reducing total fat and saturated fat intake:

  • The Australian Heart foundation guidelines
  • The heart association guidelines
  • The Australian dietary guidelines
  • The RACGP and Diabetes Australia diabetes guidelines
  • The American Heart Association Guidelines
  • The NICE cardiovascular disease guidelines
  • The World Health Organisation healthy diet guidelines
  • The World Heart Federation guidelines

…and the list goes on and on. It seems that every single evidence based guideline has made a statement on limiting fat intake and on avoiding saturated fats, but are these statements evidence based? A study from the Lancet from the PURE trial has resurfaced this controversy and in this episode, we will explore the evidence.

Ecological Evidence

Saturated fats are in eggs, animal meats and milk products like milk, cream, cheese and butter. The recommendation to avoid saturated fats is so well known and so widely adopted that you’d think there was pretty solid evidence to back it.  Well you might be surprised – because the recommendation to decrease fat and avoid saturated fats actually came from pretty weak evidence from ecological studies. This is where you look at different populations, see what they eat and count their heart attacks. If a population is having more heart attacks, see if you can blame something in their diet. The most famous of these studies was the seven countries study by Ancel Keys. Here, Keys showed that the Countries which ate the most saturated fat had the most heart attacks. Fascinatingly, he actually collected data from 21 countries but only reported on the 7. When a guy called Jacob Yerushalmy analysed the data from all 21 countries, the association was no longer there. But it was too late, so to speak, the cat was out of the bag. Fats were deemed bad. And saturated fats…worse.

For a more in depth discussion of the history of fats and diet and guidelines, there’s a book called “Good calories, bad calories” by Gary Taubes that goes into a lot more detail. If you can handle his passive aggressive and sort of salesman type tone, then give it a read.

Surrogate Marker Studies

The second piece of evidence used a few decades ago to launch the saturated fat recommendations, is that eating foods high in saturated fats increases cholesterol. (Actually, while it increases LDL it also increases HDL and some studies suggest it conveys a more favourable HDL:LDL ratio). Does it matter though? The reason we care about cholesterol is because of its link to cardiovascular disease.  It is a surrogate marker for cardiovascular disease. But there are plenty of drugs that lower cholesterol but have no impact on cardiovascular disease. So the fact that something lowers cholesterol doesn’t always mean it is good for us. Not only that, but treatments like the Mediterranean diet reduces cardiovascular disease without having any impact on cholesterol levels. So it’s best we use evidence that assesses the impact of low fat diets and low saturated fat diets directly on cardiovascular diseases, rather than on things like cholesterol.

Cohort Studies

Since this seven countries trial, studies that have tried to back up the claim that saturated fats are bad, have failed to come through with the goods.

The next level of evidence, up from ecological studies, are cohort studies – this is where you take a population, ask them how much fat they eat (among other things), and then follow them up to see how many got cardiovascular disease and whether it was associated with their diet.

A recent review of all of the studies that used this approach was published in the BMJ in 2015 (reference below). They found that there was NO association between saturated fats and death, heart attacks, strokes or diabetes.

Randomised Controlled Trials

The highest level of evidence are randomised controlled trials, where you actually randomise a group of people to either reduce their saturated fat intake or keep it the same. By far the biggest of these studies was the Women’s Health Initiative. If you look at systematic reviews like the Cochrane Review, this trial contributed to about 60% of the power of the review. This is a complicated trial that needs its own episode, but in one of the arms of the trial, 50,000 women were randomised to two groups: The first group simply continued their usual diet, and the second group got a very intensive program of dieticians and eduction to reduce their total fat and saturated fat intake. They followed them up for 8 years and found absolutely no difference in cardiovascular disease whether they continued their usual diet or reduced their fat and saturated fat.

The PURE Study

The PURE study is the latest to stab this saturated fat theory in the back. It’s a huge prospective cohort study – and probably the highest quality one we have to date.

They recruited 156,424 people and got them to complete surveys at the start of the study and then every 3 years after that.  They were interested in things like smoking, physical activity, medications, socioeconomic things like education and income and a full medical history.

They recruited people from 18 countries getting a good mix of  third world and first world countries. This is what sets it apart from other studies.

They recruited from three high income countries – like Canada and Sweden, 11 middle income countries – like Brazil, China, Poland and South Africa, and four low-income countries like India and Pakistan. And they based this on the World Bank classification.

They also tried to get the most accurate representation of what each of these people were eating. They did this by making participants fill out food questionarres on their diet. Every country had its own questionarre that was specific to that country but the questionnaires were standardized across the countries. The problem with this method is that there is good evidence that food questionaries are often not an entirely accurate depiction of what someone is actually eating. Imagine if I asked you what sorts of foods you eat –  How different would your answer be if you were in a good mood compared to being in a bad mood – your diet might not be different, but you might feel guilty and self loathing and inflate all the bad things you eat. If you just got out of gym, on the other hand, and are feeling amazing, you might inflate all the fruits and vegetables you eat. And then there’s the problem of recall – who can remember how many times they added salt to their food? And how are you to know how much butter or oil was in that eggs Benedict you ordered from the restaurant. This is one of the big limitations of any diet study, be it a cohort or even a randomised controlled trial. In order to verify the accuracy of the answers to these questions, they got a sample of about 50-250 participants from each country to keep a 24 hour food diary every now and then. That way they could compare how well matched their actual food intake was to what they had written on their food questionnaire – and they could try and adjust for this.

This PURE study is still ongoing. They are following this population continuously and publishing studies about them left right and centre. The specific study from the PURE data that we’re looking at today was titled: “Associations of fats and carbohydrate intake with cardiovascular disease and mortality in 18 countries from five continents (PURE): a prospective cohort study. It was published in the Lancet in August 2017.

Out of the 156,424 people in the PURE cohort, only 135,335 could be included in this study because they didn’t have all the data they needed. They also excluded people who had a history of cardiovascular disease. The average follow up was for 7 and a half years.

The aim of the study was to see if they could make any associations between diet and cardiovascular disease. They followed each patient up every single year to see if they had a heart attack or a stroke or any other cardiovascular outcomes.

They calculated the proportion of carbohydrates, fats and protein in everyone’s diet. They did this by converting the foods they ate into nutrients. They then split up the population into quintiles for each of these macronutrients. So for example, with saturated fat intake, they split the population into 5 groups based on the proportion of saturated fat they were eating. The top 20% – who ate the most saturated fats –  were put in the highest quintile and the 20% who ate the least amount of saturated fat were put in the lowest quintile and so on and so forth. They did this for all the macronutrients. And then for each quintile, they checked to see how many died, how many got strokes, heart attacks and so on.


Saturated Fats

Put that skinny flat white down and and take a seat for this one because you could be in for a bit of a shock. The recommendation is that we eat less than 6-10% of our total energy intake from saturated fat, it differs slightly depending on the guideline. The lowest quintile of saturated fat consumers in this study were eating 2.8% of their total energy intake from saturated fat – so they were well and truly within all the guidelines. And they were the most likely group to die! They were also the most likely group to develop stroke! And while it wasn’t statistically significant, they seemed to be at the greatest risk of all the other cardiovascular outcomes.

There was no difference between any of the other quintiles, only this lowest quintile of saturated fat consumers where at a higher risk. So it didn’t matter if you were getting 13.2% of your total energy from saturated fat or 9.5% or 7.1% or 4.9% – there was no difference in death or cardiovascular disease. But if you were eating a really low amount – 2.8%, then your risk significantly increased. All cause mortality was 7.2% in the lowest saturated fat eaters and this went down to around 4.5% as you increased the amount of saturated fat in your diet.

Total Fats

Again, the lower your total fat intake the more likely you are to die or develop a stroke. The highest quintile of fat eaters were getting 35.3% of their energy from fat and they were the least likely to die. Total death and stroke was more common when total fat made up 10% of your diet compared to if it was 35.3%.


This is where things get interesting. When I was in primary school there was a national program to teach kids about the food pyramid. On the bottom of this food pyramid – the foods we should be eating the most – were all carbohydrate foods – breads, cereals and pasta. Well, what does the PURE study have to say about that? It showed that as your intake of carbohydrates increased, your risk of death and cardiovascular disease also increased. And if you look at the graph it really looks like it’s in a linear fashion. So in the lowest quintile of carbohydrates eaters (those who obtained 46.4% of their total calories from carbs) – 4.1% died during the follow-up period. But among those who ate the most carbohydrates (77.2% of their total calories) – 7.2% of them died.

Other Macronutrients

Interestingly, animal protein intake was associated with lower rates of mortality.


  • Why do we insist on making recommendations when we don’t have sound evidence to support these recommendations. Especially when they involve a radical change in lifestyle.  Using a different example,  most dietary guidelines, up until recently, used to strongly recommend that parents should avoid giving their babies foods that are allergenic. The American Academy of Pediatrics told all mothers not to eat peanuts during pregnancy, while breastfeeding and not to give them to their child until they reach 3 years of age. Then the LEAP trial came along in 2015 and showed that actually, the earlier you give peanut products to babies, the less likely they are to develop allergies. It is clear that these recommendations had actually caused a lot of food allergies.  The difference is that the paediatric nutritional guidelines have been quick to adjust their recommendations after the LEAP trial came out, but the total fat and saturated fat recommendation don’t seem to show any signs of slowing down despite an increasing body of evidence that we were probably wrong.
  • This isn’t perfect evidence. Nothing beats a randomised controlled trial in terms of telling us whether something is causative, but it’s almost impossible to do a good quality randomised controlled trial on diet. Firstly blinding is impossible. Secondly, if I told you to stop eating saturated fats, or you start eating much more saturated fats, would you be able to do that? For 5 years? It’s very tricky. This is one area of medicine that you argue that cohort studies are probably better at giving us answers than randomised controlled trials – just because it’s so hard to control the bias in RCT’s. But even still, we do have randomised controlled trials – which do not find any benefits from a diet low in saturated fats. It would be very unusual not to pick up an association from cohort studies if there was something very bad about saturated fats as the guidelines suggest.  When you look at cohort studies of smoking – there was like 8-10 times higher rates of lung cancer in those who smoked. It was a very robust association. But here, cohort studies just don’t seem to find an association.
  • One might argue that people who are health conscious might exercise more, smoke less and choose their diet carefully – therefore skewing the results. But if anything, this would skew the results towards low-fat diets looking better. Because healthy people are eating less fat and less saturated fat – because that’s what society has been telling them to do – so if the people in the low-fat group are exercising more – this would make the results look even more impressive.
  • Perhaps people who can’t afford meats which are high in fat are more likely to die not because they are eating less fat, but because they don’t have money to pay for healthcare, etc. Interesting potential bias. So the researchers did additional analyses where they adjusted for socioeconomic status and found the same results.

Bottom Line

This is a direct quote from the conclusion of the study:

Global dietary guidelines should be reconsidered in light of the consistency of findings from the present study, with the conclusions from meta-analyses of other observational studies and the results of recent randomised controlled trials.

References and Links

#13 Lowering cholesterol with evolocumab and cardiovascular risk

There’s a new cholesterol lowering drug in town and its ability to lower cholesterol is like nothing we’ve ever seen before. It’s called Evolocumab, (trade name Repatha) but is it any good at reducing cardiovascular disease? This week, we delve into the FOURIER trial – a humongous randomised controlled trial that will answer this question.

Development of Evolocumab

Evolocumab is a monoclonal antibody PCSK9 inhibitor. It’s actually an interesting story how this drug came to be.  A group in Paris who do a lot of research on families with familial hypercholesterolemia had long known about a mutation on Chromosome 1 that was associated with some of these families. They had no idea where and what the gene did but they were aware of it. Then, in another part of the world, researchers in Canada had discovered a new protein involved in cholesterol regulation whose gene was also located on Chromosome 1. The two teams got together and eventually, in 2003, discovered it was all the same gene. The gene was for PCSK9 and certain mutations that over-activated this gene seemed to be linked to familial hypercholesterolemia.  But it was also discovered that people with mutations that de-activated this gene had very low levels of cholesterol and perhaps even reduced cardiovascular disease. The more PCSK9 the higher your cholesterol and the less PCSK9 the lower your cholesterol. So what if we could block PSK9? Well, this multi billion dollar idea was quickly developed by Amgen who made Evolocumab – a fully human monoclonal antibody that binds to PCSK9 and inhibits it. Early phase clinical trials have shown that its ability to reduce LDL is out of this world, like nothing we’ve ever seen before. but we’ve been waiting for big trials on whether it has any impact on cardiovascular disease.


Wait no longer! Because the FOURIER trial has arrived. It stands for Further Cardiovascular Outcomes Research with PCSK9 Inhibition in Subjects with Elevated Risk. …..bit of a stretch if you ask me.

This trial was published in the NEJM in March of 2017.


They recruited 27,564 patients from 49 countries between he ages of 40 and 85 with cardiovascular disease. They had to either have had a stroke, a heart attack or symptomatic peripheral vascular disease. People with angina or stents were not considered having cardiovascular disease.

The patients also had to have other risk factors that put them at even higher risk than your average secondary prevention patient, like diabetes or smoking.

They had to have at least an LDL of 1.8mmol/L (70mg/dL) and they had to already be on a statin.

They were then randomised to either evolocumab or placebo injections. If they were randomised to the evolocumab group, then they could choose between having a 140mg injection every 2 weeks or 420mg injections every month. And the gave the same choice to those in the placebo group to ensure blinding.


They ended up with 27,564 patients. The average age was 63 and one-quarter of them were women. 80% of them had had a history of myocardial infarction and 20% had a history of stroke. 70% were taking a high dose statin and the rest were taking a moderate dose. 90% were on aspirin. 30% of them were smokers….even after having their heart attack or stroke!

The average LDL was 2.4mmol/L (92mg/dL), and they were able to lower this by 60% with evolocumab….which is huge! Most were already on maximal dose statin so this is very impressive.

They followed them up for just over 2 years.

The primary outcome was a composite of cardiovascular death, myocardial infarction, stroke, hospitalization for unstable angina, or coronary revascularization.

So what did they find?

The primary outcome occurred in 11.3% of the placebo group, and in the evolocumab group…it went down to 9.8%. So this gives a 1.5% absolute reduction. The relative risk reduction is 13%

Where relative risk reduction is very useful is if the baseline risk is different for everyone. So in this study, it showed that evolocumab had an absolute risk reduction of 1.5%….but that’s only if your baseline risk is 11.3% over 2 years. If your baseline risk is 5% over 2 years, then this number is no longer true. And this is where relative risk is useful. Because we can apply it to people’s baseline risk to determine their individual absolute risk reduction. Using the relative risk reduction of 13%, if our patient has a 5-year cardiovascular risk of 10%, then giving evolocumab for 5 years will reduce that by 13% to 8.7% – making an absolute risk reduction of 1.3% for that patient. But if our patient has a 5-year risk of 3%, then evolocumab will only reduce it down to 2.6% – making an absolute risk reduction of 0.4%.

Relative risk is also good when comparing it to other drugs. Compare Evolocumab’s 13% relative risk reduction to high dose statins which reduce risk by 35%!. Or to ezetimibe which reduces risk by 6%. Or antihypertensives which reduce risk by about 25%. So hopefully that puts evolocumab in perspective.

Now, it’s important to see exactly where this 13% risk reduction came from. Because it was a 13% reduction in any of heart attacks, strokes, unstable angina or revascularization, and I think we can all agree that a 13% reduction in risk of revascularization or angina is very different to a 13% reduction in cardiovascular death.

So firstly, there was absolutely no reduction in death – cardiovascular death or death from any cause. Not even a signal of benefit there – it was 3.1% in the placebo group and 3.2% in the evolocumab group. There was a reduction in non-fatal myocardial infarction, this went from 4.6% down to 3.4%, and stroke also went down from 1.9 to 1.5

In terms of harms – they could not find any difference. And they looked at a lot: cognitive impairment, muscle aches, bleeding, allergic reactions. There was a slight increase in injection site reactions but this was rare.


  • Sponsorship: The manufacturer of the drug sponsored the trial – they were involved in designing the trial and collecting the data. They then gave the raw database with all the data to an independent group called the TIMI group and they were responsible for analysing the data. So industry was involved but this trial would never have happened and probably wouldn’t have been as robust if not for industry funding.
  • Lipid Hypothesis: This trial and the positive results of the trial, has re-opened this whole debate about the lipid hypothesis. This is the debate as to whether lowering cholesterol is important – no matter which drug…or whether the drug is important and it doesn’t matter what it does to cholesterol. There is probably no bigger debate in evidence based medicine than this. A lot of proponents for the lipid hypothesis consider this trial a great victory to their belief system but I wouldn’t be so quick to jump on the band wagon because there are a few holes to the theory in this trial. Firstly, the investigators divided the patients in this trial based on their starting LDL. In the lipid hypothesis, we would expect a greater benefit from reducing cholesterol in those who have a very high cholesterol – The higher the cholesterol the greater the risk and so the greater the reduction in cardiovascular disease with treating such high cholesterol. We see this in hypertension, reducing blood pressure from 180 systolic with an antihypertensive gives you a much greater benefit than reducing a 140 systolic with that same antihypertensive. But this was not seen…there was the exact same benefit if your starting LDL was 3.4mmol/L or if it was 1.8mmol/L. Secondly, there’s a group called the cholesterol trialists who did a big meta-analysis of cholesterol lowering in 2012, and their conclusion was that it doesn’t matter what you use to lower cholesterol, every 1mmol/L reduction in LDL will result in a 23% relative reduction in cardiovascular events over 5 years. This trial showed that evolocumab reduced LDL by 1.6mmol/L – so if we apply this rule of the clinical trialists, this should equate to a 37% reduction in cardiovascular disease. But it didn’t – It led to a 13% reduction in cardiovascular disease. So clearly the impact of cholesterol lowering is not consistent across the different drugs.
  • Another point to mention is that these were not your average secondary prevention patients. It’s very common for a drug to be released onto the market with a very narrow indication but this tends to expand over time. It’s difficult to say whether this medication will benefit your average secondary prevention case. After all, not only had these patients already had a heart attack or a stroke, but they also had risk factors on top of that like smoking or diabetes. Even worse, my fear is that this drug will creep into the primary prevention space even though it’s very doubtful it has any meaningful effects in primary prevention.,
  • Cost: Evolocumab currently costs 14,300 USD per year of treatment. So let’s make a calculation. This study showed a number needed to treat of 67 over 2.2 years. So if you multiply 14,300 by 2.2 and then by 67 – this drug costs 2.1 million dollars to avoid one non-fatal heart attack or non-fatal stroke.  That will not save any lives, just reduce a heart attack….for 2.1 million dollars. Can you imagine how many heart attacks you would avoid by putting 2.1 million dollars into exercise programs? Currently, the Government in Australia will pay for the medication in those with homozygous familial hypercholesterolaemia.

Bottom line:

We now have 3 medications that lower cholesterol and reduce cardiovascular disease. The latest is the monoclonal antibody, Evolocumab (Repatha), which is able to reduce cardiovascular disease by a relative reduction of 13%. Compare this to high dose statins which reduce cardiovascular disease by 35% and ezetimibe which reduces CVD by 6% (relative). Evolocumab does not reduce death. It currently costs 14,000 US dollars per year, which equates to 2 million dollars per nonfatal heart attack avoided.

References & Links:

#12 Magnesium for Leg Cramps

Night time leg cramps are very common and can ruin sleep. One study surveyed 490 veterans and found that 60% of them had suffered from night time leg cramps and 1/4 of those had them regularly. A lot of people use magnesium to ease their nocturnal cramps and they often swear by them, but do they actually work? This week, we delve into the evidence of magnesium for leg cramps.

A study published in JAMA in May of 2017 has tried to answer this question.


This was a randomised, double blind, placebo controlled trial that was conducted in Israel. They first recruited people using flyers and posters inviting anyone who suffers from leg cramps. Then, for anyone who expressed interest, they put them through a 2 week screening phase where they didn’t give them any treatment but asked them to keep a diary where they had to document how many night time cramps they got and the nature of these cramps. They then used this diary to include or exclude people based on whether they matched the criteria. The criteria was that they had to have at least 4 episodes of nocturnal leg cramps during this 2 week screening phase. And the criteria for what they called a cramp was fairly strict: it had to be painful, occurring at rest, involuntary and causing a palpable knot in the muscle. They excluded pregnant women and those with renal failure.

166 people responded to the ads, and 94 of those passed the 2 week screening phase and made it into the trial. On average these people were getting 8 night time leg cramps per week. 60% of them were women, and the average age was was 65.

They were randomised, with concealed allocation, to receive either 865mg of magnesium oxide daily (520mg of elemental magnesium) or placebo to take at night before bed.

The primary outcome was the change in the number of cramps per week. Secondary outcomes included severity and duration of the cramps, quality of life and quality of sleep. The way they got their data was again with that diary –  Every morning when they woke up they’d write down how many episodes they had, how severe it was from 1-10 and how long they went for. The investigators sent them a text message every morning to remind them to fill out their diary and investigators called them twice per week to see whether there were any adverse effects, to make sure they were taking the magnesium or the placebo and to make sure they were filling out their diary.

They followed them up for 4 weeks and here’s what they found:


Magnesium was able to reduce the number of cramps – there were 3.41 less cramps per week when the participants were taking magnesium. Coming from an average of 8 cramps per week that’s about a 40% reduction in the number of cramps. Which is pretty impressive.

The problem is though….that in the placebo group there were 3 less cramps per week. Which is also pretty impressive. And while magnesium was able to reduce cramps by 0.4 episodes per week over placebo, this was not statistically significant.

Now this doesn’t necessarily mean it isn’t true. It wasn’t powered with enough people to detect such a small difference. They designed the trial to have enough power to detect a difference of 1 episode of cramping per week, so they would have needed 2-3 times more people to accurately say whether this reduction of 0.4 episodes per week was true. But even if it is true is a reduction of 0.4 nocturnal leg cramps per week a significant one? This equates to 1 less cramp every 2-3 weeks. So in someone getting 8 cramps per week is this meaningful? I don’t think so.

They also presented the results in a dichotomous way which is what you need to get a number needed to treat. They defined a “minimum clinically important difference” as a reduction in leg cramps by 25%. They found that in the placebo group, 63% of them reached this minimum clinically important difference, but in the magnesium group, 75% of them reached it. Again, this was not statistically significant, the P value was 0.21, but again, that doesn’t mean it isn’t true – it just means that there is a 20% probability that these results occurred by chance.

Therefore, in this study, 12% more people obtained a minimum clinically important improvement in their symptoms in the magnesium group compared to placebo. If this is true, it means the number needed to treat for magnesium over placebo is around 9.

There was no difference between magnesium and placebo for any of the secondary outcomes, like quality of life or sleep quality.

In terms of side effects, there was no difference between magnesium and placebo. But other studies have shown increased rates of diarrhoea with magnesium.

Other Evidence

There have been 4 other randomised controlled trial looking at magnesium for idiopathic leg cramps (Cochrane), and they all showed pretty much the same thing. That amazingly, there was a great improvement with placebo, but no greater improvement with magnesium. 1 of the trials also showed this very slight improvement with magnesium over placebo but it also was not statistically significant.

There have also been three trials looking at using magnesium specifically in pregnant women and while one found it was better than placebo, the other 2 didn’t.

Bottom Line

In those who suffer nocturnal leg cramps, placebo and magnesium are very affective at reducing the number of cramps, reducing them by almost half. There was a non-statistically significant improvement with magnesium over placebo which, if true, equates to a reduction in the number of leg cramps by 0.4 episodes per week or a number needed to treat of 9 for a minimum clinically important improvement (P=0.21).

So should we stop using magnesium? Well, given that the placebo response is so good for this condition, and given there is absolutely no other effective treatment for nocturnal leg cramps and given that the side effects are minimal….perhaps magnesium is a reasonable option. What are your thoughts? Leave a common

References and Links:


#11 Antibiotics for Otitis Media with Paul Glasziou

Otitis media, or middle ear infection, is a very common issue. It is the second most common reason for a child to see a GP and is the most common reason antibiotics are prescribed to children. According to a large Australian database called the BEACH data, 80% of kids with otitis media are prescribed an antibiotic.

The dilemma

A study was published in the NEJM in December of 2016 by Hoberman, et al.,  that compared 10 days of amoxycillin/clavulanate to 5 days for treating kids with acute otitis media. They randomised 520 kids between the ages of 6 and 23 months and this was a very well designed trial. They found that 10 days was far superior to 5 days – Clinical failure occurred in 34% of those getting 5 days of antibiotics versus 16% in those getting 10 days. This makes a number needed to treat of 6 – so one in every six kids will avoid clinical failure if they are treated with 10 days of antibiotics versus 5 days. This paper got a lot of press – “should we be treating otitis media with antibiotics again?”, “Did we get it wrong with the recommendation that we shouldn’t treat?, “Should we change the guidelines to recommend 10 days of antibiotic use?”

But this is weird! because when you compare antibiotics to placebo, the benefit is nowhere near as good.  So comparing antibiotics to less antibiotics was way better than comparing antibiotics to no antibiotics! How does this make sense. To put it more starkly, the exact same research team that did this study, did a different study in 2011, also published in the NEJM, where they compared 10 days of amoxycillin/clavulanate to placebo. And here there was no difference. So the same team of researchers with almost identical study design on the one hand showed that 10 days of antibiotics doesn’t work, yet on the other hand showed that 10 days of antibiotics is better than 5 days! Whats going on here?

The answer lies in a wonderfully written opinion article published in the BMJ by Paul Glasziou entitled: “How to hide trial results in plain sight”

But before going into that, let’s first cover the previous evidence on treating otitis media with antibiotics

Cochrane Review

A Cochrane review on antibiotics for otitis media was done in 2015 which found 13 randomised controlled trials that were of a low risk of bias making a total of 3,400 children with otitis media. These studies compared antibiotics versus placebo.

They looked at a heap of outcomes but the main ones are:

  • There was a reduction in symptoms at 2-3 days in those who took antibiotics. 84% of children had improved by 2-3 days in the placebo group versus 89% in those who took antibiotics. So antibiotics increase the chance of recovery by 5% making a number needed to treat of 20.
  • There was no difference in hearing loss, no difference in suppurative complications like meningitis or mastoiditis.
  • There was a reduction in ear drum perforations – this went from 5% in the placebo group to 2% in the antibiotic group.
  • And in terms of adverse events, the number needed to harm for vomiting, diarrhoea or rash was 14 in those taking the antibiotic.

The Answer

Almost every other study had used the child’s symptoms as their primary outcome, including the Cochrane review and even including the 2011 Hoberman study that compared 10 days of amoxycillin/clavulanate to placebo. This is the AOM-SOS score or the Acute Otitis Media Severity Of Symptoms Scale. This scale assesses things that parents and children really care about: Ear pain, ear tugging, irritability, decreased play, eating less and fever.  The 2016 Hoberman study that compared 10 days to 5 days of amoxycillin/clavulanate used a different primary outcome. Their primary outcome was any of these three things:

  • If they still had otoscopic signs of infection
  • If they did not have a complete or nearly complete resolution of signs and symptoms by the end of treatment
  • Worsening infection

There are three problems with this outcome:

  1. Who cares about how the ear looks otoscopically? This is a disease oriented outcome rather than a patient oriented outcome. As a parent I care if my child is eating, if they’re not in pain and if they’re not miserable. I really couldn’t care less if their ear drum was still red or not.
  2. This outcome is very subjective “if they did not have complete or nearly complete resolution of signs and symptoms”.
  3. All the other trials have used the AOM-SOS scale as their primary outcome which is a patient oriented outcome, why did they change it?

The authors did also use the AOM-SOS scale but used it as a secondary outcome, and it was very difficult to find the results for this – you had to look at different rows from two different tables to find this result. Paul Glasziou has done that for us in his article and has even draws us a nice graph comparing the severity of symptoms over time. And when you compare this change in symptoms, there was really no difference between the 10 days or the 5 days.


  • 1 in 3 kids in this trial had diaper dermatitis (nappy rash). Now as a parent of 2 kids who are still in nappies – this is a nightmare. They’re in pain when you wipe them, you’re forever putting creams on and getting peed on all the time as you try to maximise nappy-free time.
  • 1 in 3 had diarrhoea. 

Other Issues

This study was done out of hospital setting, so these kids were going to the emergency department. And they had very strict criteria for which kids were allowed into the study. They had to have obvious bulging of the tympanic membrane and the presence of middle ear effusions. A lot of the children had a fever as well. So if we struggle to show a benefit for antibiotics in these patients, how much more so will they be useless in those with a slightly red drums, a bit grizzly, coming to their GP and not hospital.

The lead author, Hoeberman, apparently has a patent on an amoxycillin clavulate combination. And this doesn’t make the data wrong, but given that there is potential gain by him proving more amoxycillin/clavunate is better, it definitely warrants a more skeptical eye when looking at the methods and results.

Bottom Line:

The evidence suggests that antibiotics will increase the chance of recovery by 5% over placebo, it will decrease the risk of ear drum perforation by 3%, but has no impact on long term hearing or other complications like mastoiditis or meningitis. Using amoxycillin/clavulanate for 10 days will improve otoscopic signs compared to 5 days of use but will not impact patient symptoms. 1 in 3 children will develop diarrhoea and 1 in 3 will develop nappy dermatitis with the use of amoxycillin/clavulanate (whether for 5 days or 10 days).


#10 Likelihood of chronic opioid use after first prescription

Opioid addiction has been increasingly recognised as a public health crisis:

  • Deaths from prescription painkillers have now surpassed the deaths from heroin.
  • Opioid use disorder has risen 500% from 2010 to 2016 in the USA
  • For the first time in history, the Surgeon General sent a letter to every single doctor in America. That’s over 2 million doctors! And the letter was a summary of the recommendations on opioid prescribing from the CDC. This clearly points to the importance of this issue. 

The issue is that opiates can be a very good drug. They can be a life saver in terms of pain management and they can be used in patients with kidney disease or peptic ulcer disease when other pain killers, like NSAIDs, can’t.

The scary thing about them though, is that some people become dependant.  I always have this strong fear in the back of my mind whenever I hand over an opioid script like ”am I going to ruin this patients life forever?”

This week we look into a study that asks the question: “when handing a patient a script for opioids for the first time, what is the chance that they will become a chronic user?” It also asks: ” Is there something that a doctor can do when giving a patient their first opioid prescription to reduce the risk of them becoming a chronic user?”




The state of Oregon in the USA has whats called a “Prescription Drug Monitoring Program” where they can track and monitor all prescriptions for controlled substances. Firstly, this is brilliant, because clinicians can use this data to see if a patient is drug seeking. In Australia we rely on doctors to report to a doctors shopper line, but a lot of drug seekers fall through the cracks with this system. The other reason why this monitoring program is amazing, is that it creates an unbelievable amount of great data on opioid use. The authors of this study used this data in a really genius way.

They looked at every patient that was prescribed an opiate between the years of 2012 and 2013 who were otherwise opiate naive. They classified someone as being opiate naive if they had not filled a script for opiates for 12 months prior to this first script being filled. There were 536,767 of these patients who were prescribed an opiate for the first time. And this was all outpatient prescriptions, so it didn’t include opiates dispensed in the hospital.

They then checked how many of these became chronic users. They defined that as someone who filled 6 or more scripts for opioids over the subsequent 12 month period. They had good evidence to back this up as a good marker of chronic use.

The opioids they included where analgesics, antitussives and migraine medications. They had to exclude tramadol and buprenorphine-naloxone combinations because they were not included in the prescription drug monitoring program during 2012 and 2013.


  • They wanted to exclude people who were prescribed opiates with the intention of being a long term prescription, like those with cancer pain or other palliate care patients. This is because the intention of the study was to find out how many people become long term users when we don’t want them to be long term users.  The problem is that this data set doesn’t include the patient diagnosis. To get around this, they excluded patients who died within 1 year of this first opioid prescription assuming that these patients might have been palliative care patients.
  • They also excluded anyone under the age of 11 because they are thought to be at low risk of chronic use and because this age group would mostly be prescribed opiates in the form of antitussive medication. 
  • Another issue they had with the data was that all they had to work with was the number of pills prescribed and the dose of the pills. So they couldn’t work out from that whether it was a PRN prescription or whether they were prescribed to take 2 tablets per day or 5 tablets per day. But they used a different approach to get around this. A good marker of what dose they were using was how many times they refilled a script within the first 30 days of this first script.  So they know the dose of the pills and the amount of pills that were prescribed, and they know now how long it took them to refill the script they could now figure out what dose they were using. 


  • Of the 536,767 patients who were prescribed an opiate for the first time, 26,785 became long term opioid users. Which makes 5%. 
  • 85% of these first prescriptions were combinations of opiates with paracetamol. In Australia we have a combination product of paracetamol and codeine known as Panadein or Panadein Forte. America seem to have a lot more combination options: they have acetaminophen and hydrocodone, acetamenophine and oxycodone as well as acetaminophen and codeine.
  • The more repeat fills they had in the first 30 days after the initial prescription, the more likely they were to become chronic users: The risk of chronic use went from 2.9% if they just filled that first script in the first 30 days, to 10% if they had another refill in the first 30 days, to 26.1% if they had 4 or more refills in the first 30 days. 
  • Now here is possibly the most interesting finding in the entire study: those who were started on long acting opioids had a much higher chance of chronic use than those started on short acting opiates. It was 25% versus 3.5%. 

Bottom Line

There is a 5% risk that a patient will become a chronic opioid user after being given their first script for an opioid medication. The likelihood of long term use increase greatly as the number of script refills within the first 30 days increased – From 2.9% with just that 1 opioid script fill, to 26% if they filled 4 or more.  Initial prescribing of extended release preparations are associated with much higher chronic use rates compared to immediate release forms (25% versus 3.5%)






#9 Delayed Antibiotics

We want to reduce the use of antibiotics but at the same time we also want to keep our patients satisfied. Otherwise, they’ll just go and find another doctor.  This week we look into a really great study that compared refusing to give that antibiotic script, to giving them the script straight away, to a delayed prescription – which is where you tell them to only use the script if symptoms don’t get better in a few days. Which method is the best for reducing antibiotic use? and which method is best for keeping patients satisfied? Let’s take a look.


There is a lot of antibiotic prescribing in primary care. A new report on the Australian BEACH data found that around 80 to 90% of all antibiotics prescribed by GP’s for respiratory tract infections were completely unnecessary. And this is likely to be an overestimate because the doctors volunteering to  participate in the BEACH study are probably different to the average GP.

The whole issue of antibiotic resistance has been getting constant attention and is considered a massive public health threat. So simple methods to reduce antibiotic prescribing should be welcomed with open arms.

One such method is called “delayed prescribing”  where the doctor writes out a prescription for the patient but in one way or another gets them to only use it if their symptoms don’t improve in a few days.

I’m pulling out this particular trial published in the BMJ in 2014 out of the large body of evidence on delayed prescribing – mainly because I think it was one of the more excellent trials.  But there has been heaps of trials like this, and amazingly they all show similar results. They will be discussed later.


So this trial was published in the BMJ in 2014 and it randomised patients who didn’t need immediate antibiotics into 5 groups.

They recruited 53 GPs in 25 different practices and they gave them each a bunch of envelopes. Anytime a patient would come in with a respiratory tract infection the GP would decide if the patient needed immediate antibiotics. If they didn’t need immediate antibiotics, they picked an envelope at random and in that envelope it would tell the doctor which group the patient should be assigned to. It also had an advice sheet to give to the patient based on which intervention they were getting.

The five group were:

  1. No prescription – so here the doctor would read the envelope and use their own approach to not give them a prescription, but they had to offer advise to come back if things got worse. So they said things like: “this is viral, you don’t need antibiotics, this will get better on its own, etc.“
  2. The other 4 groups used different approaches to delayed prescription. The second group was given the script and asked not to use it unless they didn’t get better in the next few days.
  3. The third group was given the script for antibiotics like the other group, but the script was dated in the future, so they couldn’t fill it until that date had come some days later
  4. In the fourth group, the doctor left the script with reception and told the patient they could come at any time without booking an appointment to collect the script if their symptoms don’t pick up over the next few days.
  5. And finally, the fifth group had to actually call the clinic and leave a message for the doctor that things weren’t getting better and they wanted the script – then the doctor would leave it at reception for the patient to pick up.
  6. They also analysed the data together with the participants who got an immediate script. So really there were 6 groups: 1 immediate antibiotic group, 1 no antibiotic group and 4 delayed antibiotic groups.

The advice on how long they should delay taking the antibiotics was different depending on the type of respiratory tract infection. So if it was an ear infection they would say: “if things don’t get better in 3 days then use the antibiotics”, or “come pick it up from reception” or “call the clinic” (depending on which group they were assigned to). If it was a sore throat the delay was for 5 days, and if it was for an acute bronchitis, the delay was for 10 days.

They managed to recruit 889 respiratory tract infection suffering patients but 333 were given immediate antibiotics because the doctor thought they needed them. So only 556 were randomised into the five delayed prescribing groups.

The outcomes they assessed were symptom severity –  to see if those with delayed prescriptions had worse symptoms, they also looked at how much antibiotic was actually being used in each group, and patient satisfaction. Anyone over the age of 3 with any respiratory tract infection was included in the study. Including: cold, flu, sore throat, middle ear infection, bronchitis or sinusitis.


Severity of Illness

In terms of severity of illness, there was no difference between the groups. There was also no difference in the duration of the respiratory tract infection nor in the amount of analgesia used. Even those who were given the antibiotics to take immediately had no difference in their illness compared to the others. So telling your patient: “no, you can’t have these antibiotics” or offering them one of the four different delayed prescription strategies did not effect their actual illness compared to giving them the antibiotics straight away.

Antibiotic Use

There was a massive reduction in the amount antibiotics used with the delayed prescriptions:

Of those who were given the immediate antibiotic script, 97% ended up taking the antibiotics. This is remarkable because in general, if a doctor gives a patient a script for almost anything else, study after study has shown that only about 70% ever go and fill that script. So there is something special about antibiotics. Or maybe there is something special about respiratory tract infections and the desperation to get rid of them as quickly as possible.

If the patient was denied a script for antibiotics 26% of them used antibiotics. And this is very interesting. Many just went to a different doctor and got their script. Makes you think hey?

In the delayed groups, there was no difference in antibiotic use no matter which method of delayed prescribing you used. Around 37% took antibiotics in the delayed groups: whether they were given a script and told to only fill it if they didn’t get better in a few days, whether they had to come to reception to collect their script, whether they had to call the doctor a few days later if they wanted their script or whether they got a script which was dated in the future, it didn’t matter. 37% of them used antibiotics.

So offering a delayed antibiotic script reduced antibiotic use by doctors 60%. And denying them a script only reduced antibiotic use by 10% over what a delayed script did. (Note: this difference in antibiotic use between delayed prescriptions and no prescriptions was not statistically significant)

Patient Satisfaction

With regards to patient satisfaction: 80% of patients were very satisfied with the consult if they didn’t get a script at all. So the fear that patients always want antibiotics and will not be satisfied if they don’t get them is not indicated in this study. Or perhaps the doctors where very good at explaining why they didn’t need antibiotics. The highest level of patient satisfaction was 89% and this was in the patient led delayed script  group- where they were given a script and told only to fill it if things didn’t get better.

Other Evidence on Delayed Antibiotic Prescribing

A Cochrane Systematic Review was done in 2013 (before this study was published), and it found 10 trials making a total of 3157 patients comparing delayed prescribing for respiratory tract infections.

Just like in the BMJ study, if given an immediate script for antibiotics 93.3% will use antibiotics.

Again, just like our study, when given a delayed script only 32% will use antibiotics. Looking at the forrest plot, all the studies in this Cochrane review showed this same result with regards to this massive reduction in antibiotic use with delayed prescriptions.

Of those who were denied antibiotics, 14% still used antibiotics according to this Cochrane review. This is a little bit lower than the 27% in the BMJ who used antibiotics even though they were refused them initially.

Again, this review found no difference in the clinical outcomes whether you got immediate antibiotics, delayed antibiotics or no antibiotics.


So the results are very consistent. Offering a script and asking the patient to only use it if things don’t get better in a few days, will reduce antibiotic use by 60% and that’s the absolute reduction. Thats huge.


Bottom Line

Offering a delayed antibiotic prescription for a respiratory tract infection –  where you give the patient a script and tell them only to use if things don’t get better in a few days – will reduce antibiotic use by 60%. Completely refusing to give a script for antibiotics will only marginally decrease antibiotic use compared to a delayed prescription.  There is no difference in clinical outcomes when giving an immediate script, a delayed script or no script.


Firstly,  no other tool that has been studied is as effective at reducing antibiotic use as this. Delayed prescribing achieves a 60% absolute reduction in antibiotic use.  The trials that look at patient education and doctor education do not come anywhere near this in terms of reducing antibiotic use.

Secondly, the qualitative evidence on why doctor prescribe antibiotics for respiratory tract infections seem to indicate four main reasons:

  1. The first is the fear that patient will develop a bad complication from their illness – and you can understand this  because if you’re a GP who has been working for several decades, it’s almost inevitable that you’ll send someone home with a mild infection that will then end up getting really sick. And that might scare the bageebes out of you and change the way you deal with these respiratory tract infections.
  2. The second reason is that we think the patient wanted the antibiotic – and it turns out that we’re usually off target with this one, patients often don’t want them at all. They just want an explanation. In my own experience, I’ve had patients who I could have sworn were just coming because they wanted antibiotics and after I explained why I didn’t think it was necessary, to my shock they’re like: “oh few….I was so scared you were going to tell me to take antibiotics…I really hate taking them”.
  3. The third reason antibiotics are prescribed when they probably shouldn’t be is because of diagnostic uncertainty. So here the doctor is not sure whether its viral or bacterial, so to er on the side of caution they give antibiotics.
  4. Finally, the fourth reason is to do with time. Doctors fear that spending the time to explain why a patient doesn’t need antibiotics will take much more time than just printing a script and giving it to the patient. And will make them run late for the rest of the day.

Delayed prescriptions are brilliant because they tackle all four of these reasons. If a doctor fears the patient will develop a nasty complication from not getting antibiotics, they still have a script should things get worse.  If the doctor fears the they will lose the patient by not giving them a script because they think the patient really wants one – well in delayed prescribing the patient still gets a script. And if the doctor is worried about time, it’s just as quick and easy to give a script and say use these in a couple of days if things don’t get better.

Lastly, all four methods of delayed prescribing led to the same amount of antibiotic use, so it makes sense to use the method that’s easiest for the patient – which is to give them the script and let them drive their decision take it or not, rather than make them call me or leave it with reception.


#8 Does Ezetimibe IMPROVE-IT?


The reason doctors like to lower their patients cholesterol is to decrease their risk of developing cardiovascular disease. There is no point decreasing cholesterol levels if it has no impact on cardiovascular disease.

Ezetimibe (Zetia, Ezetrol  or Vytorin when in combination with simvastatin), was on the market for 13 years before the manufacturers finally published a study to assess whether it actually had any benefit for cardiovascular disease. This was called the IMPROVE-IT trial and is the topic of this episode.


The FDA approved Ezetimibe in 2002 without any evidence that it had patient oriented benefits like reducing cardiovascular disease. The approval was based entirely on trials that showed that it was able to lower LDL cholesterol over placebo.

This drug was marketed so aggressively in the following years that it had managed to make $5 billion dollars in its first 5 years, well before any trials were done to test whether it reduces cardiovascular disease. Remember, we want to lower cholesterol for the sake of reducing cardiovascular disease….not just to make blood tests look better.

Before the IMPROVE-IT trial came in 2015 the manufacturers of ezetimibe intermittently published studies – some on other surrogate markers like carotid artery wall thickness and other smaller ones on cardiovascular outcomes but they never looked good for ezetimibe. Let’s just go through some of these so we can get a complete picture of ezetimibe.

In 2008 they published the ENHANCE trial where they compared simvastatin and ezetimibe versus simvastatin alone for carotid intimal thickening – which is a measure of atherosclerosis in the carotid arteries. Carotid artery thickening,  is just another surrogate marker for cardiovascular disease – the thought being that the thicker your carotid arteries, the more likely one is to develop cardiovascular disease, and you could argue that this is a bit better than just showing its impact on cholesterol levels. In this study, there was no difference between the two groups. In fact the carotid artery walls were actually thicker in the ezetemibe group but this was not statistically significant.

Then, also in 2008, came the CEAS trial which compared ezetemibe plus simvastatin to simvastatin alone for 1,873 patients with Aortic Stenosis. They followed them up for 4.5 years to see if there were any differences in ischaemic events or aortic-valve events but there were none.  Concerningly, this trial showed a statistically significant increase in cancer with ezetimibe.

In 2009 came another one of these carotid artery wall thickness trials. This time they compared statin plus ezetimibe versus statin plus niacin. Niacin being another drug that has shown to reduce LDL and increase HDL. Again they were looking at the thickness of the carotid artery – and with this the niacin was better than the ezetimibe.

In 2011 there was finally a trial that showed benefits for hard outcomes with Ezetimibe, the SHARP trial. The problem, was that they compared statin plus ezetimibe versus nothing at all! ….Ezetimibe…together with a drug we already know reduces cardiovascular disease…..to nothing at all. So was it the statin doing all the work? Who knows? Its impossible to tease out how much of the benefit came from statins versus how much came from ezetemibe.

So up to this point we have a bunch of trials showing ezetimibe lowers cholesterol, and it consistently does this, lowering it by about 20%. But any trials that have tested hard outcomes have either shown no benefit for ezetimibe or have been a rediculous comparison.


That was all until 2015. 12 years after they got approval by the FDA they published the IMPROVE-IT trial

This came out in 2015 in the NEJM and was a multi-centre, international, double blinded placebo controlled trial to test whether ezetimibe with simvastatin is better than simvastatin alone in terms of cardiovascular outcomes. In other words, what is the added benefit of ezetmibe over statins?

Guess that acronym

Let’s play….guess that acronym…

So IMPROVE-IT stands for: IMProved Reduction of Outcomes -Vytorin Efficacy International Trial. So they took the I the M and the P from IMProved ….well played vytorin…well played.

And double points because it nicely played on the title of the PROVE-IT trial which was a trial comparing high dose statin to medium dose statin.

Study Design

They took  the highest risk patients you can possibly study. Men and women over the age of 50 who had just had a heart attack (within the last 10 days). They had to have an LDL cholesterol of over 1.3mmol/L (50mg/dL). They excluded those with kidney and liver disease. While they originally recruited 5000 patients, this number seemed to increase to 18,144 for reasons which will be discussed later.  They also extended this trial…twice. So the follow up ended up being 7 years on average.

They randomised these 18,000 patients to one of two group. One group got simvastatin at 40mg and the other group got simvastatin 40mg plus ezetimibe 10mg.

The primary outcome they were assessing was a composite of: Death from cardiovascular disease, myocardial infarction, unstable angina requiring hospital admission, revascularisation and stroke. A pretty controversial composite outcome because some argue that unstable angina and revascularisation which is just stents and things, should not be grouped together with heart attacks and strokes. They call them “softer” outcomes. But as we’ll see later from the breakdown of each of these outcomes, revascularisation and angina didn’t contribute much to the results of this composite.

Baseline characteristics of the patients in the study

All of the patients in the study had just had a heard attack within the last 10 days. But 20% of them had had a heart attack on another occasion as well as this one, so this was their second or third heart attack. The average age was 64. A quarter of them were women and a quarter had diabetes. The average LDL cholesterol was 2.4mmol/L (93.8mg/dL)


LDL cholesterol decreased from 1.8mmol/L in the statin only group to 1.4mmol/L in the statin/ezetimibe group. So about a 20% reduction in LDL which is what we’ve seen in other trials with ezetimibe.

With regards to the primary outcome, the composite of heart attacks, strokes, angina and revascularisation – it  went from 34.7% in the statin only group to 32.7% in the ezetimibe-statin group.

So there was a 2% reduction in cardiovascular disease when adding ezetimibe to a statin for these extremely high risk patients when given for 7 years. To put it another way: 98% of people who have just had a heart attack and take this drug for 7 years will not derive any benefit from it.

To put these numbers in perspective: 35% developed cardiovascular disease over 7 years in this trial. A statin would reduce a 35% risk to about 26% if it was low dose and to about 23% if it was high dose. Ezetimibe reduced it from 35% to 33%.

While this trial required 18000 patients and 7 years of follow up to show a 2% reduction in CVD, a different trial that gave statins immediately after a heart attack only needed 3000 patients and 16 weeks to show  2% reduction in CVD. This was the MIRACL trial.

The relative decrease in cardiovascular disease was about 6% with ezetmibe (range from 1%-11%). Compare this to statins which give about a 25% relative risk reduction. Going from low dose to high dose statin even gives you more bang for your buck than ezetimibe – it offers about a 10%-15% further reduction in cardiovascular disease on top of that already given by the low dose statin

Now lets break down that composite outcome. The good thing, is that the bulk of the 2% reduction came from hard outcomes. They mostly came from heart attacks and strokes. There was a 1.7% reduction in non-fatal heart attacks and a 0.6% reduction in strokes.

But! there was no difference in cardiovascular mortality. Or in overall mortality. Compare this to statins which consistently show a reduction in mortality – at least for secondary prevention by about 10%. But alas, ezetimibe could not show a difference in mortality even in this very high risk population

In terms of harms, they did a good job of looking very thoroughly at all sorts of side effects, and there was really no difference in harms between the two groups. Importantly, no difference in cancer rates, because previous studies had signalled an increased risk of cancer with ezetimibe.


  • Firstly, 42% of patients stopped therapy in each group. Now that is a very high drop out rate. So make of that what you will.
  • Secondly, there were 5 amendments to the protocol during the study! One being to increase the sample size. Was this because they didn’t like the results they were seeing? They reported in their methods that they were looking for a certain number of events 5,200 to be exact. But was that the plan from the beginning? Or did that become the plan while the study was going on? There was a lot riding on this trial. 3 billion dollar a year in sales, they really couldn’t afford not to show a benefit. Even at the expense of sound methodological quality perhaps?
  • When they ran the statistics before the study, they calculated that they needed 5000 people to detect a 9% relative risk difference in cardiovascular outcomes, but they ended up with 18,144! So my question is: where was that interim analysis for 5000 people? They must have done one to then decide to increase the number to 18,000…where is that report?
  • This drug is being heavily marketed for use in primary prevention. But, there is no evidence that it works at all in primary prevention. You could argue it barely works for secondary prevention.  But let’s assume that this 6% relative reduction carries true even in primary prevention, what benefits might patient who haven’t had heart attacks be getting from this drug? In Australia, guidelines recommend initiating cholesterol therapy once the 5 year risk reaches 15%. A statin would reduce that risk to about 10% over that five years. Whereas ezetimibe would reduce it to 14.1%. So a 0.9% reduction in cardiovascular disease if taken every day for 5 years. In America and the UK their guidelines are recommend starting therapy at a 10 year risk of 10%. So if a patient with this risk took ezetimibe every single day for 10 years, only 0.6% of them would get any sort of benefit from this drug.
  • Another criticism of this trial that has been published is that standard of care after a myocardial infarction is high dose statins at a dose of 80mg of simvastatin. While the majority of the cardiovascular benefit comes from the first 10mg, there is evidence that 80mg provides a small but significant benefit on top of this. This study only used 40mg, which is below the recommended post MI dose of statin. So would we still see a benefit for ezetemibe on top of the recommended 80mg simvastatin? That is anyones guess.
  • This is the second drug that lowers cholesterol and also lowers cardiovascular disease. So I guess it adds to the argument for the lipid hypothesis. But its important to not get carried away and use drugs that lower cholesterol but don’t have evidence on cardiovascular outcomes There have been several drugs like: fibrates, torcetrapib and niacin that improve the lipid profile but have shown no improvement in cardiovascular outcomes – some even showing harms. So while the authors, in their conclusion make a very bold statement that “this study clearly proves the lipid hypothesis” I am still cautious.

Bottom Line

For patients who have just had a heart attack, adding ezetimibe to statin therapy will at best (given other studies showing no benefit) reduce cardiovascular disease with a 6% relative reduction. Compare this to statins and blood pressure medications which offer a relative risk reduction of around 30%.  Ezetimibe will not reduce death, compared to statins which consistently show reductions in death in secondary prevention. There is no evidence that ezetimibe has any benefit in primary prevention.


#7 Testosterone Trials – Should we treat age related testosterone deficiency?


Testosterone levels seem to naturally decline as men age. There has been an enormous rise in the amount of testing and treatment of this natural decline in testosterone.  Testosterone clinics have been popping up all over the place and promise that treatment will restore energy, improve depression, increase libido and enhance overall well-being.  In America, the direct to consumer advertising about low T is everywhere- on television, radio, magazines and billboards.  There has been a 100 fold increase in testosterone prescriptions over the past three decades and this is despite the complete lack of any new indications for the drug or any new evidence for the use of testosterone. A trial that looked through the medical records of 112,000 veterans who had been prescribed testosterone. found that only 3% of them met the laboratory  criteria to actually confirm the diagnosis of low testosterone. Around 13% had contraindications to testosterone therapy. Including 1.5% with prostate cancer! And 16% never even had their testosterone levels tested.

So either doctors really believe in testosterone therapy or patients are pressuring doctors to prescribe it. And until now, there has been no good evidence to support OR discredit the treatment of age related testosterone deficiency. The  Testosterone Trials are the first well designed piece of evidence to shed some light on the correct approach. These trials assessed the impact of testosterone treatment on age related testosterone deficiency, not testosterone deficiency due to pathology of the reproductive system.

Study Design

This one study was actually a series of 7 studies. Each was a randomised, placebo controlled trial. They screened 51,085 but only 790 met criteria for inclusion. And then this group 790 men were used across the 7 studies. participants could be included in more than 1 of the trials depending on which inclusion criteria they matched. The general inclusion criteria included:

  • Age >65
  • Low serum testosterone levels on two seperate readings (<9.5nmol/L or <275ng/dL)
  • No past history of prostate cancer and low risk for prostate cancer
  • No cardiovascular disease and low risk for cardiovascular disease

They randomised half of the participants to testosterone and the other half to placebo. The testosterone they used was 1% Androgel in a pump bottle, they initially started 5g topically, daily, but they titrated the dose to keep the serum concentration to a normal level. They considered a normal level as the normal range for someone aged 19 to 40. To maintain placebo, they also did fake dose titrations to those getting the placebo gel.

Vitality Trial

To be recruited to the vitality trial patients had to have low vitality as determined by a vitality scale called the FACIT-fatigue scale (Functional Assessment of Chronic Illness Therapy-fatigue). This is a scale out of 65 which assesses symptoms such as “I feel tired”, “I have trouble starting things because I’m tired”, “I feel weak all over”, “I need help doing my usual activities”, etc. Out of the 790 men in the overall study, 474 qualified for the vitality trial.

The primary outcome was how many men in each group had a greater than 4 point increase in this vitality score. They chose 4 because that is supposed to be the minimally clinically important difference.

Bottom Line: There was no difference in the number of people who had clinically important improvements in vitality with testosterone compared to placebo

Physical Function Trial

To be included in the physical function trial participants had to have difficulty walking or climbing stairs and they had to be slow –  with a gait speed of less than 1.2 metres per second on the 6 minute walk test. 387 patients out of the 790 were enrolled in this trial. The primary outcome was how many people achieved a 50 metre increase in their 6 minute walk test.  This is what was considered to be a clinically important improvement in their gait speed

Bottom line: There was no difference in the number of people who had clinically important improvements in their physical function.

Sexual Function

To be included in the sexual function trial, participants had to have low libido as per a low score on a libido scale called the DISF-M-II scale. They also had to have a partner willing to have sexual intercourse with them at least twice per month. 459 patients qualified for this study. The primary outcome was the improvement on a sexual function scale called the PDQ-Q4 scale (Psychosexual Daily Questionnaire – Question 4). This score asks people about their sexual encounters. For example:  Did you experience any of the following today: sexual daydreams, sexual intercourse, flirting, ejaculation. There are 12 of these in total, each with a “yes” or “no” making a total possible score of 12.

The average baseline score was 1.4 in both groups. In the placebo group, it went to an average of 1.3. In the testosterone group went to 1.9. Making a 0.6 point increase in this 12 point scale over placebo with testosterone therapy.

They assessed erectile dysfunction as a secondary outcome using a score called the IIEF erectile function score. This increase by 2.6 points with testosterone over placebo. Compare this to phosphodiesterase type-5 inhibitors (Viagra, etc.) which in meta-analyses of elderly men has shown to increase this score by 5.7 points.

Bottom Line: Treating men with low testosterone and low libido with testosterone therapy will given them about half an extra flirt, or half an extra spontaneous erection, or half an extra sexual daydream per day. Though direct comparisons have not been done, testosterone appears to be inferior to phosphodiesterase type-5 inhibitors.

The Cardiovascular Trial

By way of background, there are trials which point to an increased risk of cardiovascular disease with testosterone therapy. One randomised controlled trial of 209 elderly men in 2010 found that those in the testosterone trial were 6 times more likely to develop a cardiovascular adverse event than those in the placebo group. Because the Testosterone Trials had men with low cardiovascular risk, it was not powered to pick up a difference in cardiovascular events. So instead they used the surrogate marker of non-calcified coronary artery plaque volume and coronary artery calcification score. 170 men were enrolled in this part of the trial and they put them into a CT coronary angiogram machine at the start of the study and at the end of the study to see who had more atherosclerotic plaque built up in their coronary arteries. There was more atherosclerosis and a higher coronary artery calcification score in those receiving testosterone

Bottom line: Testosterone therapy will increase the amount of atherosclerotic plaque in the coronary arteries as well as the coronary artery calcification score. While this study was not powered to detect a difference in patient oriented outcomes of cardiovascular disease, previous studies have signalled an increase in CVD with testosterone therapy

The Cognition Trial

493 of the 790 men met criteria for Age Associated Memory Impairment and there was no improvement in cognitive functions or memory with testosterone

Bottom Line: testosterone therapy does not improve cognition in men with age associated memory impairment

The anaemia Trial

126 patients out of the cohort of 790 were anaemic, half of which had no known cause. The primary outcome was how many people achieved a rise in haemoglobin by 1g/dL. 54% of men did this when given testosterone versus 15% who were given placebo.

Bottom Line: Treating low testosterone with testosterone therapy in anaemic patients  will improve anaemia with a number needed to treat of 3

The Bone trial

Here they took 211 patients and checked whether their Volumetric Bone Mineral Density changed using quantitative CT scanning. They also checked BMD with DEXA scanning. And there was a slight increase in bone mineral density with testosterone therapy. What we really care about is fractures though, because after all, this is why we care about BMD. They checked for fractures at regular intervals in all the 790 participants. During the 12 months of the study – 6 got a fracture in the testosterone group and 6 got a fracture in the placebo group.

Bottom Line: Testosterone therapy increases bone mineral density but there is no evidence that it reduces fractures

Adverse Events

There was no difference in adverse events between the two groups

Patient Global Impression of Change

In the above summary for each of the trials, only the primary outcomes have been listed. These T-trials had a lot of secondary outcomes and some of them were statistically significant while others were’t. But one of the secondary outcomes deserves mention.

It great when trials actually ask the patient…”did you feel better on this drug?” And this trial did that. They did each of the 790 participants these questions: “Since the beginning of the trial, has your sexual desire been better, not changed or worse?”, “Has your walking ability been better, unchanged or worse?” and “has your energy been better, unchanged or worse.?

With regards to physical functioning, around 11% of those who got placebo said that their walking ability got much better compared to 18% of those who got testosterone. So an absolute difference of 7% in the patients impression of their improvement in physical functioning. Making a number needed to treat of 15.

For vitality, around 10% of those in the placebo group said their energy levels were much better, compared to about 21% in the testosterone group. Making  a number needed to treat of around 10 for patient impression of improvement in vitality.  With regards to sexual function, 5% of those getting placebo said that their sexual desire was much better, compared to 20% of those who got testosterone. The NNT is 7, So for every 7 patients with low testosterone and low libido who are treated with testosterone therapy, 1 will report an improvement sexual function.

Final Bottom Line:

For elderly men who have low testosterone levels that is not due to pathology of the reproductive system, restoring their levels with testosterone therapy to that of a young man, will not improve objective measures of vitality or physical functioning but will result in a slight improvement in objective measures of sexual functioning which is of marginal clinical significance and is likely inferior to phosphodiesterase type-5 inhibitors. But, there will be a subjective improvement in sexual function, vitality and physical function in about 10% of patients. Testosterone will not improve cognition in those with age related memory impairment. It will improve anaemia with a NNT of 3 in those with anaemia. It will improve Bone Mineral Density but not reduce fractures. Concerningly, while this study was not powered to detect a change in cardiovascular disease, it did find an increase in the amount of atherosclerotic plaque in the coronary arteries as well as the coronary artery calcification score with testosterone therapy. This is on a background of previous studies which have showed increased cardiovascular disease with Testosterone therapy.


#6 Medication to speed up kidney stone passage

Calcium channel blockers and alpha blockers are both smooth muscle relaxants and so could theoretically relax the smooth muscle of the ureter and speed up the passage of kidney stones. Given that kidney stones can take several weeks to pass and often are rated as 10/10 pain, treatments to speed up the passage of stones would be greatly welcomed.

A Cochrane meta analysis in 2014 found 32 studies, making a total of 5864 patients with kidney stones, and showed that alpha blockers, mainly tamsulosin, increased the likelihood that a stone would pass by 4 weeks with a number needed to treat of 4. Tamsulosin was better than the calcium channel blocker, nifedipine, in this review.

The problem is that the 32 trials in the review are of fairly poor quality. They were small, the largest having 150 patients and the others have less or much less than that. And most of the studies had an unclear risk of selection bias, which as we discussed last week, is one of the most important measures of trial quality. A lot of the trials also didn’t do any blinding.

Since this Cochrane review a very high quality randomised controlled trial has been published in The Lancet in 2015. They randomised 1, 167 patients with kidney stones less than 10mm to one of three group: One groups were given the alpha-blocker tamsulosin at 400micrograms per day, the second group got the calcium channel blocker nifedipine at 30mg per day and the third group got a daily placebo.

On average, these patients were 43 years old, 80% were men and 20% were women. The average stone size was 4.5 mm, though 25% were greater than 5mm.

The primary outcome was how many people had passed their stone by 4 weeks in each group.

The results were very different to the Cochrane review. They showed no difference – about 80% had passed the stone by 4 weeks no matter whether they got tamsulosin, nifedipine or placebo. There was a 10% absolute increase in stone passage with tamsulosin if the stone was greater than 5mm. It went from 61% with placebo and nifedipine to 71% with tamsulosin, though this was not statistically significant. You could argue that as only 25% of the patients in the trial had stones bigger than 5mm, this analysis was not powered to detect a 10% difference in outcomes. This improvement in stone clearance with tamsulosin for larger stones certainly fits with what the Cochrane review had showed. What we really need now is a large, well designed study, like this one but only looking at stones greater than 5mm to see whether this holds true.

In terms of adverse events, 3 people developed serious adverse events in the nifedipine group and 1 in the placebo group. There were no serious adverse events in the tamsulosin group.  In terms of how many people stopped due to adverse events, for nifedipine the number needed to harm was 10 and for tamsulosen the number needed to harm was 25

So what do we do with all this?

We now have a very high quality, well designed trial that shows that tamsulosen and nifedipine are of no benefit in improving stone passage rates or reducing the need for some sort of intervention, but we have data from 32 other smaller and crappier studies saying that it does. Which is the better evidence – a large, well designed randomised controlled trial or a meta analysis of smaller studies?

And while this can frustrate some people I think this is the beauty of evidence based medicine. Evidence is not black and white. It’s often grey.  And we can use our own opinions or our patients opinions to guide what you do based on the scenario in front of us

You could easily argue:  1) There didn’t seem to be any major adverse events with tamsulosen, there is some evidence that it works, especially for stones greater than 5mm, it’s only for 4 weeks, maybe I’ll give it a try. Or 2) If they couldn’t find a benefit in a large well done study it probably doesn’t work.  And I guess both would be right.

It’s a beautiful thing.

What would you do if you had a kidney stone base on all this evidence? I think we can agree that there is no benefit for nifedipine but would you take tamsulosin? Say, if you had a stone greater than 5mm?

Currently, the guidelines in Europe (European Association of Urologists) and America (American Urological Association) both recommend medical expulsion therapy if the stone is less than 10mm and provided there is no indication for urgent stone removal.

The scary thing is that this is not the only example of where a meta analysis was “proved wrong” by a large, well designed randomised controlled trial. A study has found that large, definitive randomised controlled trials differ in the results of the meta-analysis that preceded it about 35% of the time.



  1. http://www.thelancet.com/journals/lancet/article/PIIS0140-6736(15)60933-3/abstract
  2. http://onlinelibrary.wiley.com/doi/10.1002/14651858.CD008509.pub2/full
  3. http://www.nejm.org/doi/full/10.1056/NEJM199708213370806#t=abstract
  4. Music Polyrhythmics – El Fuego

#5 Caesarian Section versus Vaginal Delivery for Breech Presentation – The Term Breech Trial

For show notes including all the results for this trial: http://www.ebmpodcast.com/breech

The Term Breech Trial is one of the most important clinical trials in the history of obstetrics and has changed the way we manage breech presentation across the world.

The term breech trial was as randomised controlled trial comparing planned caesarian section to planned vaginal delivery for management of breech presentation at term.

2088 pregnant women, who were presenting in a breech presentation at 37 weeks or more, were randomised to one of two groups: planned vaginal delivery OR planned caesarian section. Women were recruited from 26 different countries using 121 clinical centres.

The women could either have frank breech or complete breech, but they were excluded if they had a footling presentation. Women were also excluded if they were having very large babies –  They defined really big as an estimated weight of 4kg or more which is 8.8 pounds (for the Americans)

Allocation was concealed and the plan would be that if they were randomised to the C-section group, they would schedule it in at anytime from 38 weeks onwards. On the day of the C-section, they would make sure the baby was still in breech, and if it had managed to move to be in cephalic presentation they would then just plan a normal vaginal delivery.  About 40% of the mums where randomised at the time of labour, and so they would have been pretty quickly rushed off to get their Caesarian section after being randomised to that group. If they were randomised to the planned vaginal delivery group, they would just wait for them to labour spontaneously, unless for whatever reason they needed induction of labour and they had a whole protocol on how induction of labour was to be managed.

The primary outcome was the death of the baby either during labour or in the first 28 days of life, or like really bad things happening to the baby. And there was a heap of these: birth trauma, which included subdural haematoma, intracerebral haemorrhage, spinal-cord injury, basal skull fracture and peripheral-nerve injury or clinically significant genital injury; seizures occurring at less than 24 h of age or requiring two or more drugs to control them; Apgar score of less than 4 at 5 min; cord-blood base deficit of at least 15; hypotonia for a least 2 h; stupor, decreased response to pain, or coma; intubation and ventilation for at least 24 h; tube feeding for 4 days or more; or admission to the neonatal intensive care unit for longer than 4 days.

Of those assigned to C-section – 90% were actually delivered by C-section. The 10% who were delivered vaginally, did so either because they managed to turn to cephalic presentation, or they had progressed too quickly in labour that it was too late for C-section, or because the mother changed her mind last minute and decided against a C-section.

Of those assigned to the planned vaginal birth – only 56% ended up delivering vaginally. The rest were delivered via C-section and the most common reason being that they just weren’t progressing in labour or the baby was too big for the pelvis.

So what were the results? Well that composite primary outcome – with all the bad stuff that can happen to a new born baby, coupled with some less bad stuff – went from 5% in the planned vaginal delivery group to 1.6% in the C-section group. So there was a 3.4% decrease in what they called serious neonatal morbidity or neonatal/perinatal mortality with planned C-section over planned vaginal delivery. The number needed to treat was 30, which means for every 30 breech babies we deliver via planned C-section instead of planned vaginal delivery, we will prevent 1 bad thing happening to the baby.

So let’s look at the breakdown of the components of this primary outcome:

Starting with perinatal or neonatal death – this went from 1.3% with planned vaginal delivery, to 0.3% with planned C-section, so a 1% decrease in baby death with planned C-section

Birth trauma went from 1.4% to 0.6% with planned C-Section and this was mainly because of decreased bone fractures and brachial plexus injuries

Hypotonia which lasted greater than 2 hours went from 1.8% to 0.2%

There was a decrease in Low Apgar score at 5 minutes, it went from 0.9% to 0.1% with planned C-section

Rates of intubation and ventilation also decrease – from 1.3% to 0.3%

And neonatal ICU admission decreased from 3% to 1.5%

In terms of maternal outcomes – there was really no difference here. No difference in postpartum bleeding, mortality, genital tract injury, wound issues or infection. Having said that…they didn’t look at pain and return to normal function which I think we know would have been better in the planned vaginal delivery group.

Interestingly, they re-analysed the data, to see if there was a difference in baby outcomes depending on how well trained the obstetrician was in doing vaginal breech deliveries. And it did seem to make a  difference: instead of that 5% risk of the primary outcome with planned vaginal delivery  – If the only looked at obstetricians was super trained – with more than 20 years experience in delivering breech babies, this primary outcome went down to 3.2%. Which is still greater than C-section which if you remember was 1.6% but still, better than the vaginal breech delivery average.

This study quickly had a tremendous impact on practice almost immediately after publication:

For breech delivery in the Netherlands for example, the caesarian section rate went from 50% to 80% within 2 months of this study being published.

The American College of Obstetrics and Gynaecology changed  their guidelines within a few months from the publication of this study to now recommend a planned caesarian section for all term singleton breech presentations.

In Australia, by 2005, only 3.7% of babies in breech presentation were delivered vaginally. And the number of obstetricians confident In performing a vaginal breech delivery is diminishing across the world.

In 2006 a survey was done on Australian obstetrics and gynaecology trainee’s who were in their final year of training, and only 11% of them said they would OFFER a vaginal delivery to their women in breech. I imagine this would me much less if done today.

The term breech trial researchers published another paper 2 years later where they followed up a selection of the original babies in the term breech trial to see if there was any difference in outcomes after a couple of years. Only 85 of the 121 birthing centres were pre-selected to be involved in this 2 year follow up. So out of the 2088 women in the actual term breech trial, 1159 were to be included in this 2 year follow up study. As the child reached 2 years of age, the investigators sent out questionares to these 1159 women about their Childs development. Questions about communication, gross motor skills, fine motor skills, problem-solving skills, and person-social skills. The plan was to see whether there was any difference in child death rates outside of the initial neonatal or perinatal period OR if thee was any neurodevelopment delay. They only managed to get data from 920 and they found no difference in death or neurodevelopemental delay between the planned C-section and the planned vaginal delivery from breech babies at 2 years follow up.

So whats the bottom line? Well I’ll leave the lead author of the study to answer that in a quote from a response to a commentary she published in the BMJ:

“for those women preferring a vaginal breech birth, they should be reassured that although planned caesarean section reduced the risk of perinatal or neonatal mortality or serious neonatal morbidity, compared to planned vaginal birth, in the Term Breech Trial, 95% of babies in the planned vaginal birth arm did well. Also, although our statistical power was limited, we did not find planned caesarean to be associated with better outcomes for the children at 2 years of age”


Term Breech Trial: http://www.thelancet.com/journals/lancet/article/PIIS0140-6736(00)02840-3/abstract

2 year follow up trial: https://www.ncbi.nlm.nih.gov/pubmed/15467565

music by Polyphonics song El Fuego