For healthcare professionals only

CCBYNC Open access

Low carbohydrate-high protein diet and incidence of cardiovascular diseases in Swedish women: prospective cohort study

BMJ 2012; 344 doi: (Published 26 June 2012) Cite this as: BMJ 2012;344:e4026

Re: Low carbohydrate-high protein diet and incidence of cardiovascular diseases in Swedish women: prospective cohort study

Reply to letters by Drs. Campillo-Soto and Freedhoff

In his criticism of our paper Dr. Campillo-Soto repeatedly invokes “biases”, but uses a technical epidemiologic term rather loosely and unconventionally, to say the least. Selection bias is not really a concern in a cohort study. Nor is measurement bias - assessment of exposures cannot be differentially affected by the outcome when the latter is not known at the time the exposure is reported and non-differential misclassification is expected to attenuate a true association. In his reference to “compliance bias” and “expectation bias”, again interesting terminology, Dr. Campillo- Soto confuses interventional with observational epidemiological study designs. As for follow-up bias, it is relevant when losses to follow up are concerned, which is not the case in our study - linkage to registries allowed us practically complete follow up of all our study subjects.

Criticizing our paper, Dr. Freedhoff indicates what is already stated in our manuscript: that the diets of our study population are less extreme in their low carbohydrate content compared to advertised dietary regimes. The associations we detected, however, are monotonic, indicating that the risk is expected to be higher in more extreme low carbohydrate-high protein regimes. We disagree that the study is misinformative – all aspects are clearly presented in the manuscript, but more importantly, we disagree that the increase in risk is miniscule.

Both Dr. Campillo-Soto and Dr. Freedhoff criticize our use of a food frequency questionnaire, but they disregard the fact that such questionnaires, notwithstanding their limitations, are a standard tool used in large nutritional epidemiologic cohort studies. Also, both Dr. Campillo-Soto and Dr. Freedhoff point out that the assessment of the diet at recruitment only in our study generates misclassification. We agree, but in a cohort study this misclassification is non-differential and, thus, much more likely to attenuate an existing association rather than generating it.

Pagona Lagiou1, 2
Sven Sandin3
Marie Lof3, 4
Dimitrios Trichopoulos2, 5
Hans-Olov Adami2, 3
Elisabete Weiderpass3, 6, 7

1 Department of Hygiene and Epidemiology, University of Athens Medical School, 75 M. Asias Street, Goudi, GR-115 27, Athens, Greece.
2 Department of Epidemiology, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115, USA.
3 Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, SE-171 77, Stockholm, Sweden.
4 Department of Clinical and Experimental Medicine, University of Linkoping, SE-58185 Linköping, Sweden
5 Bureau of Epidemiologic Research, Academy of Athens, 28 Panepistimiou Street, Athens, GR-106 79, Greece
6 The Cancer Registry of Norway, Montebello, N-0310, Oslo, and Department of Community Medicine, University of Tromso, Tromso, Norway.
7 Folkhälsan Research Center, Samfundet Folkhälsan, Helsinki, Finland.

Competing interests: No competing interests

06 August 2012
Pagona Lagiou
Professor of Hygiene and Epidemiology
Sven Sandin, Marie Lof, Dimitrios Trichopoulos, Hans-Olov Adami, Elisabete Weiderpass
University of Athens Medical School
75 M.Asias Street, Goudi, GR-115 27, Athens, Greece