Fruit Consumption and Risk of Type 2 Diabetes 2

Fruit consumption and risk of type 2 diabetes: results from three prospective longitudinal cohort studies. Isao Muraki et al. BMJ 2013;347:f5001 (published 28th August 2013) 

This recent study published in the British Medical Journal (28th August 2013) triggered a number of articles in the Guardian newspaper, questioning the health benefits of fruit juice and smoothies, and highlighting the importance of eating more whole fruit.

The study analysed results from three large-scale longitudinal cohort studies carried out in the US since 1984: the Nurses Health Study (66,105 women), the Nurses Health Study 2 (85,104 women), and the Health Professionals Follow-up Study (36,173 men). Associations between total and specific fruit and fruit juice consumptions and the later incidence of type 2 diabetes, were expressed as hazard ratios. A hazard ratio of 1.0 indicates no association, while <1.0 indicates a protective effect and >1.0 indicates a risk. 95% confidence intervals indicate the range around the result within which we can be 95% confident that the true value lies (or put another way, if you repeated the study 100 times, the result would be expected to lie within the stated interval for 95 of those studies).

Here are the results:

Hazard Ratio 95% Confidence Interval
Whole fruit (3 servings/week) 0.98 0.96-0.99
Blueberries 0.74 0.66-0.83
Grapes and raisins 0.88 0.83-0.93
Prunes 0.89 0.90-0.96
Apples and pears 0.93 0.90-0.96
Bananas 0.95 0.91-0.98
Grapefruit 0.95 0.91-0.99
Peaches, plums and apricots 0.97 0.92-1.02
Oranges 0.99 0.95-1.03
Strawberries 1.03 0.96-1.10
Cantaloupe 1.10 1.02-1.18
Fruit juice 1.08 1.05-1.11


What are we to make of a study that indicates that if we eat blueberries, grapes or raisins we are less likely to get type 2 diabetes, but if we eat strawberries or cantaloupe melon, or drink fruit juice, we are putting ourselves at risk? To make matters even more complicated, the authors go on to state that in eight previous prospective studies, the association between total fruit consumption and risk of type 2 diabetes was examined, and the results were mixed. Similar to previous analyses in the Nurses Health Study and the Finnish Mobile Clinic Health Examination Survey study, the current findings supported an inverse association between total fruit consumption and type 2 diabetes, but not in other studies […] In four prospective studies, consumption of citrus fruit was not associated with a lower risk of type 2 diabetes. Apple consumption was inversely associated with risk in the Women’s Health Study and in the Finnish study, but not in the Iowa Women’s Health Study. In addition, greater consumption of berries was associated with a lower risk in the Finnish study, but not in the Iowa Women’s Health Study.

The problem is something called sampling variation. If you were to repeat a study a hundred times (even a very big study like the Nurses Health Study), you would get a hundred different results for the same variable or association, and you would be able to see that all the results fall within a certain range. This is what the 95% confidence interval estimates statistically (see above), based on the size of the study sample, because the larger the sample the more likely it is that the result will approximate the true value. Therefore the range is likely to be narrower for a hundred large studies than for a hundred small studies, and we can be more confident of the results. Most very big studies are not repeated a hundred times however, and so we cannot be totally sure where within this range the observed result sits – whether it is one of the majority that cluster around the true value, or whether it is one of the few outliers that is miles away from it. If the range around the true value does not include one (in the case of hazard ratios), we can at least be fairly confident that the true value falls on one side or other of no effect, i.e. that if the result indicates a protective effect then a protective effect really exists, even if we can’t be so sure of its magnitude (or actual clinical significance in the real world – how many blueberries would a person actually need to eat in order to offset one bowl of cantaloupe sorbet??)

So the point is that if your 95% confidence interval includes 1 (as it does for peaches, plums, apricots, oranges and strawberries), it may well be that the result you have seen is simply due to sampling variation. Even if it does not include 1, there is still a 5% probability that the result you have seen has occurred by chance, when in fact no such association really exists. Alternatively, an association may exist but be far too small for your study to detect (in which case it may not be clinically important anyway). It is only when lots of studies all looking at the same thing start to get broadly similar results that we can begin to say with some confidence that an association is true.

Lots of people seem to have been working on the question of whether fruit is good or bad for you in terms of type 2 diabetes risk. One of the problems though is looking at the fruit in isolation. Eating an apple as part of an apple pie, loaded with white sugar and topped with pastry made from white flour, more sugar and butter, smothered in custard made with eggs, milk and sugar, and washed down with a glass of pasteurised fruit juice drink made from concentrate (with added sugar), is quite different from simply eating an apple, perhaps after you have finished eating your fat-free, plant-based salad. Proponents of plant-based nutrition suggest that we should actually be eating at least 8-10 portions of fresh fruits and vegetables per day for optimum health, in addition to whole grains, nuts, seeds and legumes. Some even go so far as to suggest that all this food should be eaten raw. They may well be right, but this study gives figures for increments of 3 servings of fruits per week because this is consistent with the consumption levels in the study populations. We can only guess (by multiplying the observed results) at the potential effects of more dramatic dietary changes, such as may occur if a type 2 diabetic eating a standard Western diet were to switch to a whole-food, plant-based one.

It is notable that the authors do not define fruit juice for the purposes of this study, and as all the subjects are from the United States, it is reasonable to think that this category might include such horrible concoctions as (I shudder at the thought of this)… Sunny Delight, which I remember being given routinely every day to all the kids I looked after in summer camp in Indiana back in 1997, when all of these three studies were under way, and which has worryingly made its way across the Atlantic in more recent times. Sunny Delight is about as similar in composition to the freshly squeezed juices I make at home as a fruit fly is to an elephant. They are simply not the same animal.

But before we get bogged down in any more technical details, it is worth stopping to ask ourselves one very important question…


Is the study question a clinically important one?

One of the first questions that needs to be considered when reading a scientific paper is whether or not the study is seeking to answer a clinically important question. If it isn’t, there is not much point reading on – the results will be neither here nor there, and you might as well spend your time in the pub/gym/park/spa…

To answer this, it is important to understand how such research projects are generated, four years after the end of the last of these three studies. When very large prospective cohort studies (such as the Nurses Health Study) are carried out, huge amounts of data are collected on all kinds of dietary variables, in as much detail as possible. Some of the study questions will be clearly stated at the design stage (as they should be), and certain pieces of information will be sought specifically in order to answer these questions. However these studies also generate large quantities of information that researchers look at after it has been collected and think “what can I do with this data?” Or “can I get a publication out of this?” They then start to look for associations between variables, so someone will have said “let’s do an analysis of the relationship between fruit consumption and type 2 diabetes”. They won’t have known what the results would be when they “asked the question of the data”, but nevertheless they did not design the study specifically to answer that question. Publication bias may then rear its ugly head if they then only publish positive (read expected or hoped-for) associations and not negative ones.

Is it important to me to know whether eating more fruit protects me against Type 2 diabetes? Yes, that probably is helpful to know, particularly as there are people keen to suggest that we avoid eating fruit because of its high sugar content. As this study points out, there are several chemical components of whole fruit, other than sugar, that may offer powerful disease-protection: resveratrol, chlorogenic acid and naringin (no, I had never heard of any of these either!) being just three examples given.

Is it useful to me to know that in mice with diabetes, bilberry extract rich in anthocyanins can activate adenosine-monophosphate-activated protein kinase, enhance glucose uptake and utilization in white adipose tissue (that’s fat, to you and me) and skeletal muscle, and reduce glucose production in the liver? No not really. That is of interest mainly to scientists who want to sound clever at dinner parties (or make lots of money selling bilberry extract to diabetics).

What is very useful for me to know is that people with type 2 diabetes who switch to a whole-food, plant-based diet, generally lose weight and their diabetes goes away. There are lots and lots of anecdotal cases of this occurring, but as far as I know no randomized intervention study has been carried out to find out if this is truly the case.

So here’s the really clinically important question: Does switching to a whole-food, plant-based diet lead to an improvement in type 2 diabetes, or even a cure (if it does, it is also likely that eating such a diet would also prevent the occurrence of diabetes in the first place, although this would require a different sort of prospective study to analyse properly)?

To find out the answer, the study that needs to be done is this: patients diagnosed with type 2 diabetes are randomly allocated to one of two intervention programs. One group is advised to switch to a whole-foods, plant-based diet and given lots of good education and support to enable them to understand and achieve this. The second group (the control group) is given standard NHS dietary advice along with an equivalent amount of education and motivational support.

Of course this study requires a great deal more resources than sitting down at a computer and running some analyses on a pre-existing dataset. Which is probably why when I approached a few professors of nutritional epidemiology recently with a related question and study idea I was asked if I had any idea what doing such a study would entail (I did), told that it was too much even for a PhD, and advised that perhaps I should consider whether I could extract some useful information from their pre-existing datasets instead (I couldn’t – their datasets all consisted of adults, most of whom were eating the Standard UK Diet (SUKD). I am interested in children eating plant-based diets – a rather niche group, and at present I am not even sure how many of them there are).

And therein lies a fairly major obstacle for those interested in the potential benefits of plant-based nutrition. In order to commission the important studies that need to be done, you need to at least have a PhD and a position of power and authority within the scientific community (such as a professorship… don’t you??). So where does that leave the rest of us mere mortals, including healthcare professionals who want to advise their patients to adopt a plant-based diet, and need that advice to be evidence-based? We need to be able to cite the best-available evidence to date, and studies such as this one, demonstrating a protective effect of fruit consumption in type 2 diabetes, all lend support to the assertion. Ultimately though, in order for a patient to be able to trust the advice they are given sufficiently to be motivated to make the big changes required, the message must be unequivocal and entirely consistent: “do this and you will definitely be cured” is quite different from “well, the current evidence lends support to the idea that if you adopt these changes you will have a good chance of seeing some improvement in your condition, but we’re not totally sure and not everyone agrees…” The person hearing the latter message is unlikely to be able to comply, will fail to see benefits, and so will in his or her own mind prove the advice they were given. This is where the science of evidence based on statistical significance meets motivational theory, and the two do not necessarily add up. Non-scientists wishing to help people to change their life right now often throw caution to the wind and adopt the first approach, even when the current evidence is insufficient to warrant such confidence, engendering the criticism of people like Ben Goldacre of Bad Science fame. But those afflicted by type 2 diabetes today can’t really wait for me to get my PhD and a position in a leading academic institution, and then carry out a large-scale study over a number of years in order to prove (or indeed disprove – it could always happen in Good Science) absolutely, my suspicions.

In the meantime they could read Dr Fuhrman’s website, blog and new book, The End of Diabetes. On his website Dr Fuhrman points out the scale of the problem faced in the US (and the UK is following suit):

Diabetes has been cited as the most challenging health problem in the 21st century. Over 25 million people in the United States have diabetes. Excess weight promotes insulin resistance and is the chief risk factor for type 2 diabetes. Currently 68.8% of adults are overweight or obese. The number of people with this disease has been increasing steadily, largely due to the increasing numbers of overweight people. According to the U.S. Centers for Disease Control and Prevention (CDC), the prevalence of type 2 diabetes has more than tripled in the past 30 years, and if current trends continue unabated, one-fifth to one-third of all Americans will have diabetes by the year 2050.

There is a risk that studies like the one discussed above blind us with so much science that we fail to see the enormous elephant in the room: EXCESS WEIGHT IS THE NUMBER 1 BIG RISK FACTOR FOR TYPE 2 DIABETES, whatever you eat (although obviously eating healthily will help with that hugely). Take the weight off, and most people (excluding those few whose diabetes is genetic) will get better.


The day I finished writing this post, I picked up a free magazine from Diabetes UK, in partnership with Tesco, entitled Type 2 Diabetes: What You Need to Know. Of course I was very interested to read what it had to say, which is representative of the standard messages being given to type 2 diabetics in this country. I’ve listed a few of them below, and I’m interested in both the explicit and implied messages (e.g. don’t change too much, don’t take control of your own health, there’s nothing much you can do) being given out. I’m not going to discuss in more detail what I think is so wrong with these statements here (some have been touched on elsewhere), but what do you think? Please discuss:

  • Having diabetes doesn’t mean having to have a sugar-free diet […] they should still be able to enjoy a wide variety of foods, including some with sugar.
  • Beware of [web]sites offering a cure for diabetes. Research advances every day, but there’s no cure. […]
  • don’t change treatments or make any alterations to your lifestyle until you have discussed them with your healthcare team.
  • No food is off-limits when you have diabetes – it’s fine to have a treat every now and again, just don’t overindulge.
  • Bread, rice, potatoes and pasta contain the all-important nutrient carbohydrate, which is broken down into glucose and used by the body’s cells as fuel [illustration: French stick, white pasta (4 varieties), crispbreads/crackers, white rice, brown bread, potatoes]
  • Aim for at least 5 portions of fruit and vegetables per day. Fresh, frozen, dried and tinned all count.
  • Technically, your body doesn’t need any foods in [the high in fat and sugar group – including butter, chocolate, sweets, crisps, olive oil and lollipops (illustrated)], but eating them in moderation will still mean you are following a healthy, balanced diet.
  • There’s no need to throw out recipes that you know and love… But making some small changes will have you cooking up a healthier version in no time
  • When you feel that nothing but cream will do, spoon extra-thick single cream on to fruit or puddings and use whipping cream for filling cream buns or cakes instead of double cream
  • Burgers are a quick and easy option when you’re on the go, but you could cut down the fat by asking for no mayonnaise and going easy on the cheese
  • Try and order a smaller portion [of fish and chips], or even remove some of the batter
  • Can people with diabetes follow a vegetarian diet? Yes, although following a vegetarian diet does not necessarily mean a healthier diet. You still need to have a good balance of different foods
  • Achieving your ideal weight is often easier said than done. But being overweight makes diabetes control difficult
  • While many of us might like to slim down to a weight we were when we were 18, it’s important to be realistic about whether you can really achieve this. For most of us, this may not be practical

Having considered these statements, you might be interested to take a look at Dr Fuhrman’s Nutritional Education Institute website, where he has some interesting short videos on the subject of coaching and motivating people to achieve excellent health, here: Coaching Modules

Leave a comment

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>

2 thoughts on “Fruit Consumption and Risk of Type 2 Diabetes

  • Jo Marsh

    Your journal club entry is fantastic and I thoroughly enjoyed reading it…..and look forward to more. The Tesco leaflet is indicative of their target audience………..would a Waitrose handout have a different message? Tesco mention foods which should not be part of a regular diet in my opinion! However, these foods form the staple of a lot of families’ diets. If the leaflet was to offer best advice, it may be deemed to be seen as radical or controversial, and tesco would assume that NO advice would be followed. But it is not being honest and pandering to the British culture, which is wrong.

    • admin Post author

      Thanks Jo. I think you have hit the nail on the head when you say that best advice may be deemed controversial. Sadly many doctors and policymakers also subscribe to the view that the truth is too much for people and needs to be toned down. Of course if people did start to follow best advice it would have serious implications for the food retail industry. I don’t know about Waitrose, but I gather that Tesco directs much of its marketing at the “harried housewife”. It’s probably in their interests that she remains harried and without enough time or energy for home cooking…