Archive | November 2015

Dietary advice by the Health Council of the Netherlands [part 1, eating less red meat?]

Recently, the Health Council of the Netherlands published their revised dietary guidelines.  I must say the council did quite a thorough job and they also published a decent collection of background papers. For any of you who understand dutch, everything is made available here.

We won’t be looking in dept at the background papers, the point is, the details are all meticulously looked at and as we will see, getting lost in the details can sometimes make people loose track of a bigger picture. What is the number one health risk we would want dietary guidelines to focus on? Are they markers for poor health like blood levels of certain proteins? Is it cancer? cardiovascular disease?  No, the number one high level health risk is ‘death’. We shouldn’t really care that much if our food gives us a higher risk of a stroke or a heart attack or cancer, or even if our food contributes to a higher probability of committing suicide. The one number we actually care about is called all-cause mortality.

So lets look at some of the high level dietary advice’s that the council gives us and lets look what we can find about their impact on all-cause mortality. A while ago I created a simple generic data-set browser that can help to get a high level view of a multi-variable data-set. The tool shows per-variable distribution densities and than allows you to select a second variable and get a heat-map showing the rough relationship between the two variables.  One useful source of data to do this with is the data from the China Study. It contains both nutritional information and data on mortality rates including all-cause mortality. The data browser for the China Study II is available here. So lets look at ten points of advice we should be able to have a look at in our data-set to see if the advice agrees with the data.

So what does the council advice that could possibly be confirmed or rejected with our data-set:

  1. Eat less red meat
  2. Eat fatty fish once a week
  3. Prefer plant based oils over solid and animal fats
  4. Eat legumes one time per week
  5. Eat 15 grams of nuts per day
  6. Eat 200 grams of vegetables per day
  7. Eat 200 grams of fruit per day
  8. Limit added salt to 6 grams a day
  9. Consume no alcohol at all, or at most one glass a day.
  10. In general, eat more plant based and less animal based products

Over the next few days I will look at each of those independently to see how they pan out. Today I’ll look just at red-meat. I’ll discuss a few more in my next blog post soon. Before we start though, a word of caution. All cause mortality in our data set is divided up into age groups. The units used for the different age-groups differ though. For the age groups 0..4, 35..69 and 70..79 the mortality numbers are in yearly deaths per thousand people. In the age groups 5..14 and 15..34 however, mortality is significantly lower, and an other scale is used where the numbers represent the number of deaths per one hundred thousand people per year. Please note also that the low number of deaths in these later groups make the results statistically much less conclusive if we look at the data from these groups. To get a rough idea of the kind of mortality rates we are talking about, a few normalized examples. We normalize to the number of yearly moralities per 100000 people in the given age group :

  • 0-4 : 200-1600
  • 5-14 : 27-150
  • 15-34: 70-360
  • 35-69: 850-2350
  • 70-79: 3000-15600

We need to keep these numbers in mind when looking at the data set.

Eat less red meat ?



OK, lets start with an interesting one: red meat. What does the data set tell us about red meat? Well, the distribution of average red meat consumption has a nicely shaped form and ranges from a large group who eat no red meat at all to a thin tail where red meat consumption reaches levels up to 116.5 grams of red meat per day. Lets first look at the correlation with all-cause mortality for the different age groups. While we know we should not put to much value on the 5-34 age-group results, the positive correlation for the 5-14 age group is absolutely massive. Way too massive to ignore. Also for the 0-4 age group there is a quite convincing correlation there. As such, eating less or probably eating no red meat for under 15 year old would seem like a solid advice indeed. When however looking at the other side of the spectrum, 70 to 79 year old people, we see a massive negative correlation, that given the numbers involved we can be statistically confident about.  A negative correlation means that eating more red meat actually correlates with lower mortality rates. The lowest point on the heat map actually is around the 100 grams of red meat a day level.

Two age groups we didn’t look at yet: 15-34 and 35-69.  Lets look at the most statistically significant one first. We see a solid overall negative correlation, but we see something else that can often be seen in heat-maps like these: We see a basically V shaped mortality graph. There seems to be some magic amount of red meat, somewhere around 50 grams a day.  Great, looking at current consumption in the Netherlands, again the council seems to be spot on for this age group as it was for the children. But are they? Chinese people tend to be smaller than dutch people, and for many nutritional aspects, the amount of food per kg of lean body mass is actually what counts. If we look at the weight questionnaire from the same data set, the average weight of a person, depending on the locality is between 46 and 61 kg. This means we need to add almost 50% to all our figures here. That is, for an average dutch person in this age group, it is likely that the sweat spot should be more like 70 grams per day if we trust these heat maps. Still OK if everyone eats about the same as the average, than cutting on red meat consumption would be good advice. Now lets look at our last age-group left. As stated, less statistically solid, but still.  The correlation has become virtually zero for the overall graph, but look closely. The same V shape from our 35-69 group is also visible with the exact same sweat spot we just identified. The hypothesis that this heat-map depicts a transition from the 5-14 age group graph to the 35-69 age group graph appears very likely indeed. But as said, for this crucial age-group the total number of deaths is to low to extract any solid info on this proposed transition from our data-set. So what conclusions should we draw from this with respect to the councils advice to eat less red meat?

  1. With respect to children and adolescents the advice is solid and possibly even a bit conservative. Eating no red meat at all would seem safest when looking at our data set.
  2. With respect to adults with an average eating pattern the advice is solid as is it for young adults under an as yet undetermined age threshold who eat very little red meat to begin with.
  3. For adults after a certain as yet undetermined age threshold who already didn’t eat all that much red meat, eating less might not be in their best interest. For this group, (assuming a strong body-mass correlation) the data set suggests an optimum of somewhere under one gram per day per kg of total body mass.
  4. For senior citizens, eating less red meat seems ill advised. In fact, senior citizens when looking purely at our data set may actually want to significantly increase their dietary intake of red meat.

Note that this is just a preliminary visual inspection of the data-set using just two variables at a time. A more thorough multi-variable analysis of this advice will be done as a use-case in my upcoming book, so more on this to come. I hope at any rate that this analysis shows that even good solid dietary advice can potentially be detrimental to  specific groups of people as in our case here the senior citizens.

In my next few blog posts I’ll be discussing some more of the councils dietary guidelines.

This entry was posted on 5th November 2015.