My last role before retiring was a nurse educator for an international rare disease foundation. It was clear to me how much research you did for your newsletter and the reason I value it so much. Much gratitude for the time, effort and hard work to bring us the accurate info.
Thank you, Jan. I read your comment after four hours of responding to questions and comments, including a few that were a little, shall we say challenging. You gave me a lift that I very much appreciated, and reminded me why I write the newsletter.
It’s obvious you put a lot of effort and time selecting articles and crafting the final results. The only criticism I would make is the example you chose. I have a problem now believing studies based Human diets which is complicated from the tendency of the subjects to supply inaccurate info based on keeping food diaries. Maybe you could address. Is the size of the study big enough we can ignore that concern?
Hi Michael. It’s definitely a concern, but I do think well-conducted dietary studies can give us reliable and useful information.
Food-frequency questionnaires won’t record every meal or portion perfectly. That isn’t really what they’re designed to do. Their purpose is to estimate people’s habitual diets well enough to distinguish, for example, those who regularly obtain more protein from plants from those who obtain more from animal foods. Validated questionnaires are reasonably good at making those comparisons, particularly in large prospective studies.
Some people will misremember or misreport what they eat, and diets may change over time. However, random inaccuracies generally blur real differences between groups and make associations harder to find, rather than manufacturing consistent associations from nothing. Studies that reassess diet periodically are stronger because they can follow changes in eating habits.
Study size still matters. Dietary effects are often modest, and outcomes such as deaths from cardiovascular disease take many years to accumulate. Following hundreds of thousands of people can produce enough events to distinguish a real but relatively small association from random variation, and to test whether the finding is broadly consistent across important subgroups. It gives us greater statistical precision, but of course it can't compensate for poor dietary measurement if that leads to systematic rather than random errors.
The protein example drew on two large studies involving different groups of Americans. One measured diet only at the beginning, which is a limitation. The other reassessed diet every four years. Both nevertheless found a broadly similar result.
While estimated percentage differences are only estimates, I do see the overall conclusion as credible, especially because it fits a much wider body of evidence showing better cardiovascular outcomes when foods such as red and processed meat are replaced by legumes, nuts, whole grains and other healthy plant foods.
At the end of the day, we're not going to be able to shut 10,000 people in a hall somewhere and control their diet for 20 years, so prospective observational studies are the only practical way to answer these nutritional questions - though newer Mendelian randomisation studies can play a part, too. If we see similar results in multiple studies, across different ethnicities and geographical regions, I think it's hard to believe that those findings are simply down to poor recall.
My last role before retiring was a nurse educator for an international rare disease foundation. It was clear to me how much research you did for your newsletter and the reason I value it so much. Much gratitude for the time, effort and hard work to bring us the accurate info.
Thank you, Jan. I read your comment after four hours of responding to questions and comments, including a few that were a little, shall we say challenging. You gave me a lift that I very much appreciated, and reminded me why I write the newsletter.
It’s obvious you put a lot of effort and time selecting articles and crafting the final results. The only criticism I would make is the example you chose. I have a problem now believing studies based Human diets which is complicated from the tendency of the subjects to supply inaccurate info based on keeping food diaries. Maybe you could address. Is the size of the study big enough we can ignore that concern?
Hi Michael. It’s definitely a concern, but I do think well-conducted dietary studies can give us reliable and useful information.
Food-frequency questionnaires won’t record every meal or portion perfectly. That isn’t really what they’re designed to do. Their purpose is to estimate people’s habitual diets well enough to distinguish, for example, those who regularly obtain more protein from plants from those who obtain more from animal foods. Validated questionnaires are reasonably good at making those comparisons, particularly in large prospective studies.
Some people will misremember or misreport what they eat, and diets may change over time. However, random inaccuracies generally blur real differences between groups and make associations harder to find, rather than manufacturing consistent associations from nothing. Studies that reassess diet periodically are stronger because they can follow changes in eating habits.
Study size still matters. Dietary effects are often modest, and outcomes such as deaths from cardiovascular disease take many years to accumulate. Following hundreds of thousands of people can produce enough events to distinguish a real but relatively small association from random variation, and to test whether the finding is broadly consistent across important subgroups. It gives us greater statistical precision, but of course it can't compensate for poor dietary measurement if that leads to systematic rather than random errors.
The protein example drew on two large studies involving different groups of Americans. One measured diet only at the beginning, which is a limitation. The other reassessed diet every four years. Both nevertheless found a broadly similar result.
While estimated percentage differences are only estimates, I do see the overall conclusion as credible, especially because it fits a much wider body of evidence showing better cardiovascular outcomes when foods such as red and processed meat are replaced by legumes, nuts, whole grains and other healthy plant foods.
At the end of the day, we're not going to be able to shut 10,000 people in a hall somewhere and control their diet for 20 years, so prospective observational studies are the only practical way to answer these nutritional questions - though newer Mendelian randomisation studies can play a part, too. If we see similar results in multiple studies, across different ethnicities and geographical regions, I think it's hard to believe that those findings are simply down to poor recall.