We again thank all authors for their comments. Some of the comments have already been addressed in our previous response (1).
Stefan Hellstrand is suspicious that our statistical analyses might have distorted reality. The statistical analyses we have performed are standard procedures in epidemiological research. In his attached Figure 1, Hellstrand has calculated risks by combining information in Tables 1 and 2 to show that, in reality, there is a lower risk of mortality with higher milk intake and that the hazard ratios we present might be incorrect. Unfortunately, Hellstrand has combined the wrong data resulting in an erroneous analysis and incorrect results. Since we are using time-updated data for the cohort of women, a crude incidence rate would be obtained by dividing the number of cases with the person-time at risk in Table 2. To get a crude incidence, or risk, by using the numbers at risk in Table 1 (baseline data), these should be combined with the number of cases provided in supplemental Table H. By doing so, and by comparing with intake of <1 glass of milk, the risk ratios for total mortality would be 1.09, 1.22 and 1.22 for 1-2 glasses, 2-3 glasses and ≥3 glasses of milk, which is in line with the age-adjusted hazard ratios with results derived from only baseline data as presented in supplementary table H). Thus the statement that our data processing or analyses “change the results of consuming milk from being positive for health to being negative” is incorrect.
Hellstrand further compares the energy intake in our cohorts to the recommended intake in Sweden. Firstly, it is difficult to draw conclusions based on a comparison with ecological data. Secondly, it is unclear how Hellstrand has calculated the “fraction of subsamples that died” since this was incorrectly calculated in his Figure 1. Thirdly, and most importantly, we have used standard methods to compensate for misreporting of energy intake in epidemiological studies. This is done by first excluding those with implausible energy intake (≥3 SDs below or above the log-transformed mean energy intake) and then adjusting for total energy intake, as suggested by Walter Willett (2, 3). Willett states “Although the extremes within this range are rarely correct, adjustment of nutrient intakes for total energy intake will, to a large extent, compensate for overall underreporting or overreporting” (2). We would also draw the attention to the results for cheese and fermented milk products. High consumption also of these products was associated with also a higher total energy intake whereas the mortality and fracture rates were lower compared with a low consumption of these products.
Fredrika Hansson’s first question regarding lactose has been addressed in our previous response (1). Regarding the definition of outcomes, these are clearly defined in the methods section. We did not specifically study mortality after a fracture.
In response to Klaus W Aschern we would like to point out that the studies referred to regarding D-galactose were animal studies that, to the best of our knowledge, seem to be carefully conducted. He further speculates that a high intake of milk might reduce intake of other foods with positive health effects. The associations presented in the article were independent of a healthy eating pattern and also many nutrients, as shown in Appendix Table E.
Siddappa Gada proposes a prospective study to study various dietary products on mortality. Indeed, our study is a prospective study where participants indicated their consumption of milk and other dairy products and were followed until fracture or death. A long-term randomised study with exclusive adherence to separate dietary products is unlikely to ever be well performed.
Bodo C. Melnik presents an interesting alternative hypothesis as an explanation for our findings. An intriguing question arises – what is the concentration of microRNA in cheese and fermented milk products compared with that in non-fermented milk?
Henning K Antonsen proposes that gastroesophageal reflux or ulcer disease or its treatment might be the underlying cause of our findings since his patients relieve their symptoms by drinking milk. Interestingly, 50 years ago such diet to reduce the symptoms from peptic ulcers was reported to increase the risk of myocardial infarction (4). Thus, based on autopsy data both from the United States and Great Britain the relative risk of myocardial infarction was more than doubled in ulcer patients prescribed a milk diet compared with ulcer patients without such a diet (4). By use of data provided in the original publication (4), the RRs can be calculated to 2.3 (95% CI 1.4-4.0) and 4.7 (1.4-15.3), respectively. The pooled RR estimate is 2.6 (1.6-4.3).
Moreover, whether proton pump inhibitors increase the risk of fragility fractures have been debated since some years. A recent review of the current evidence concluded that patients prescribed PPI therapy tend to be more frail with more risk factors for fractures than those not given these drugs and it is probable that PPI therapy is not an independent risk factor for fracture (5). Even though a surrogate measure for fracture, bone mineral density in a RCT was significantly higher in a group that received PPI along with risedronate compared with those on risedronate alone (6). Nonetheless, we found no indication that high consumers of milk had been prescribed PPIs more frequently than low-consumers of milk (Attached Table 1).
1. Michaëlsson K, Byberg L. Reply to commentaries regarding our manuscript "Milk intake and risk of mortality and fractures in women and men: cohort studies". BMJ. 2014;349:g6015.
2. Willett W. Nutritional epidemiology. 3rd ed. Oxford University Press: Oxford; 2013.
3. Willett WC, Howe GR, Kushi LH. Adjustment for total energy intake in epidemiologic studies. Am J Clin Nutr. 1997;65(4 Suppl):1220S-8S.
4. Hartroft WS. The Incidence of Coronary Artery Disease in Patients Treated with Sippy Diet. Am J Clin Nutr. 1964;15:205-10.
5. Leontiadis GI, Moayyedi P. Proton pump inhibitors and risk of bone fractures. Curr Treat Options Gastroenterol. 2014;12(4):414-23.
6. Itoh S, Sekino Y, Shinomiya K, Takeda S. The effects of risedronate administered in combination with a proton pump inhibitor for the treatment of osteoporosis. J Bone Miner Metab. 2013;31(2):206-11.
Competing interests: No competing interests