Recovery and validation of MCE data and study materials

We recovered raw data from the Minnesota Coronary Experiment (MCE) stored on two 9-track magnetic tapes (‘tape 2’ and ‘tape 4’) (see fig A below), using similar methods as we previously employed to recover missing Sydney Diet Heart Study (see the web appendix of 13 ). Technical expertise in data recovery and conversion was provided by John Svee (Data Conversion Resource, Inc., Aurora, Colorado, USA). Computer Logics software was used to read the raw tape to disk via a pertec interface with 9-track equipment attached to Windows 98 boxes in pure DOS mode. Tape 2 contained data written to tape on the 327 th day of 1988. The raw MCE ‘tape 2’ data was split into 2 logical files with standard zero-length separation blocks between files. These data were stored on the tape in fixed length American Standard Code for Information Interchange (ASCII) text files. Tape 4 contained data written to tape on the 183 rd day of 1985. We determined that it was written using Kronos operating system, version 53. This Kronos data was found to be expressed in 6-bit, rather than standard 8-bit characters. The data format and the exact character conversion table code (Table B-2 on page 106 from 14 ) were identified and translated by trial and error, ultimately resulting in readable ASCII characters which were found to represent a related series of punched cards. Each “record” of punched card data ended in a series of “:” characters. These were translated from the original varying number of colon characters per line into PC record marks (character 13 + character 10). Next these punch card records were programmatically assembled into fixed length ASCII text records. This file of 57,664 records was split into three files representing Type D sub-records (MCE data collection form 011, see Appendix Section VI.), Type R sub-records (MCE form 02) and Type X sub-records (MCE forms 02/11).

[...]

Dietary fats and cardiovascular disease

Ramsden and colleagues conclusions on replacing saturated fat with linoleic acid are not supported by the data. Contrary to their claims, strong and consistent evidence from both randomized clinical trials and prospective observational studies indicate that replacing saturated fat with healthy fats (including n-3 and n-6 polyunsaturated fats) has favorable effects on cardiovascular disease (CVD).

There were major problems with both the design and interpretation of the data from the MCE. The most severe of these are the duration of the trial and the very high rates of dropout. Only ~25% of the study participants received the study diets for more than a year with an average exposure of ~3 years. The short-term duration of the intervention and unusually high rates of non-compliance called the validity of the study into question.

Second, their data on the degree of atherosclerotic progression confirmed through autopsy was on a very small number (n=149). It is highly unlikely for a 1-year intervention to alter plaque that has been built up for decades. In fact, well-known dietary intervention studies such as the Los Angeles Veterans Administration trial1 and the Finnish Mental Hospital trial2 lasted more than 5 years and found that replacing saturated fat with polyunsaturated fats reduced both serum cholesterol and CHD incidence.

Third, the authors found that the excess mortality with the serum cholesterol lowering intervention (“cholesterol paradox”) was primarily confined to patients 65 years and older. Analyses in this sub-sample are plagued by the same issues that we observe with the “obesity paradox3”, where weight loss is often associated with increased mortality in older individuals or hospitalized patients. Also, among those aged ≥65 years, the hazard ratio for death was similar in the intervention group and the control group suggesting that the intervention, per se, had no effect on death.

[...]