It’s all very well to show that the LCHP (LowCarbohydrate-HighProtein) score doesn’t completely and accurately reflect protein and carbohydrate intake, but what about the claimed results of an increased risk of mortality with an increasing score?
How was the relationship to mortality calculated?
The relationship to mortality was calculated through Cox proportional hazards models which are used to relate the risk of death with either the deciles of nutrient intake or the LCHP score, taking into account confounding factors (gender, age, socioeconomic status, smoking, BMI, physical activity, alcohol consumption). One thing to note about Cox proportional hazards models is that they represent the assumption of a model in which the thing being investigated (in this case carbohydrate vs. protein consumption) is being assumed to be directly causal in the decease of the affected individual. Cox models were originally used in drug and treatment trials where a terminally-ill patient’s survival might be expected to vary directly – and over a relatively short time-period (months to only a few years) depending on whether they were receiving drug or placebo. While it is likely that diet does have an impact on mortality, it is not at all clear that the link is as direct and immediate as assumed by these models.
Four different models (i.e. essentially differently constructed equations) were used to hunt for links between the supposed diet culprits – low carb/high protein – and death. In model 1 (not energy adjusted), increased protein intake (per decile) was found to be significantly linked to increased mortality (1.13 increased risk, CI: 1.03-1.23). However, as protein intake was the best predictor of total energy intake (correlation coefficient 0.93) it is likely that the one is a proxy for the other and the authors note that‘mortality … increased with … total energy intake’ (p.577)
Model 2 assessed the LCHP score but found that an association with increased risk of mortality was non-significant (P=0.14; CI: 0.97-1.20). Model 3 looked at energy-adjusted individual nutrient intake and found that energy-adjusted carbohydrate intake was inversely related to mortality: reduced risk 0.94 (CI:0.89-0.99) but the apparent link with protein had disappeared (possibly because of the energy-adjustment). However, according to the authors, this model is flawed because it does not take into account
‘the complementary changes that have to be introduced for the preservation of total energy intake, when carbohydrates and proteins change’ (p.578).
Most interesting is model 4 which finds a significant relationship between the LHCP (using energy-adjusted protein/carb deciles) and mortality (increased risk 1.08 CI:1.03-1.13, P=0.001). What this actually means is that increased mortality is associated with a decreased fat intake! Yes, because a 2-point increase in the LCHP score is caused 57% of the time by increasing intake of protein and/or carb. In an energy-adjusted formulation, when one or more macronutrients is increased, then necessarily the others decrease – as a proportion of energy intake – so what this finding may really be telling us is that – as fat intake declines as a proportion of energy, mortality increases!