Scientific statement from WASH member Professor F.P Cappuccio on Jama Paper
20 January 2015
Kalogeropoulos AP et al. Dietary sodium content, mortality and risk for cardiovascular events in older adults. The Health, Aging, and Body Composition (Health ABC) Study. JAMA Intern Med 2015; January 19, 2015 on-line
Prepared by WASH member Prof F P Cappuccio, University of Warwick
Summary of study
The present study reports the results of a 10-year longitudinal analysis of the association between sodium intake and morality in a population of 2,642 men and women, aged 71-80 years (mean 73.6 years).
Sodium intake was assessed by Food Frequency Questionnaire (FFQ), outcomes were all-cause mortality, incident CVD and incident heart failure.
Analyses were presented using sodium intake both as continuous variable and using three cut-off points of <1500, 1500-to-2300 and >2300 mg sodium per day, equivalent to <3.75, 3.75-to-5.75 and >5.75 g of salt per day.
The study had a 80% statistical power to detect a 20% increase in all-cause mortality risk per 1g of sodium intake.
The results show: (1) a linear association between dietary sodium intake and mortality (Figure 1) with estimate snot improving by more complex non-linear modelling. Hazard Ratio was 1.09 [95% CI 1.04-1.16, p=0.001] for crude estimates and 1.03 [0.98-1.09, p=0.27] for fully adjusted model; (2) in group analysis (Table 2) sodium intake >2300 mg per day (>5.75 g salt per day) was associated with higher mortality than the mid group, whereas the group with sodium <1500 mg per day (<3.75 g salt per day) showed no significant increase in mortality; (3) the results for incident CVD and heart failure are only presented by sodium groups (Table 2) and follow the pattern of mortality.
The authors conclude: (1) sodium intake was not associated with 10-year mortality, incident CVD or heart failure; (2) consuming >2300 mg sodium per day was associated with non-significant higher mortality
This study contains methodological flaws in many domains that have been highlighted recently and that make the study and the results unreliable and in danger of misleading researchers and the public.
It is an observational study and does not imply cause-effect relationship. Nevertheless these studies may be of interest if interpreted in the wider context of the available evidence and their numerous pitfalls acknowledged. Sadly the authors do not do satisfactorily address them.
Systematic error in sodium assessment. The study relies on FFQ as a method for assessing sodium intake. The method is not only imprecise, but may lead to biased assessment. The authors accept that FFQ may be less accurate than 24h urine collection and that they are poorly correlated with it. Furthermore their validations (quoted in ref. 31 and 32) are very old and may be out of date given the significant reformulation of foods that has occurred globally since 2006, especially for sodium content. Finally, in Table 1 they describe 11% of the sample has having an intake of sodium <1500mg per day, at variance and in contrast with the evidence they quote that achieving this level is extremely difficult and that in the NHANES data only 1.3% of people over 51 years achieve that level. Clearly there is a problem with their assessment of sodium intake and/or with the selection of their population sample, admittedly selected on the basis of voluntary participation and good functional capacity, process that introduces concerning biases. Finally, discretionary use of salt is not factored in as contributor to total sodium intake.
Reverse causality. There is no mention of sensitivity analyses after exclusions of deaths in the first couple of years from sodium assessment. Table 1 suggests some selection bias in the characteristics of the three sodium groups that might be explained by several facts. Males are less represented in the low sodium group, possibly reflecting survival bias. People in the low sodium group eat significantly less calories than the other groups but have comparable body mass index, suggesting that the reason for lower calorie intake is not reduced size but likely ill-health or adopted life-style as a result of high risk. Furthermore, Table 1 shows that, in the lowest salt group, more individuals had coronary heart disease, cerebrovascular disease, hypertension and particularly uncontrolled hypertension. The mean systolic blood pressure was 3 mmHg higher in lowest salt group compared with the middle salt group. These indicate that the lower mortality and CVD events in this group are almost certain to be due to reverse causality.
Potential for residual confounding. The referred imbalance in key confounders across sodium categories (Table 2) has the potential to alter the direction of associations.
Insufficient power. The study had a statistical power of 80% (p=0.05) to detect a 20% increase in mortality risk per 1g of sodium intake (assuming a linear association). The authors accept (Figure 1) the assumption of a linear association, they are able to detect a highly significant 9% increased risk in the crude analysis (Table 2) and fail to detect statistical significance in the adjusted model which shows a 3% increase in mortality for higher sodium intake. These results indicate that the study was not powered to detect a smaller effect of 3%, still of population significance if it were true. Any additional analysis by incident CVD and heart failure, as well as by sodium groups is certainly underpowered and not helpful. In their discussion the authors often refer to ‘signals’ in their results, not a scientifically acceptable terminology. In conclusion, their results do not rule out any effect smaller than a 20% difference in risk for a 1g sodium difference and are unable to confirm the presence of a U or J shaped relationship between sodium intake and mortality.
Finally, it is surprising that in support of the argument that there is ‘controversy’ on the efficacy of a moderate reduction in salt intake for the prevention of cardiovascular disease authors, reviewers and editors have allowed the citations of reference 49 , that was withdrawn by the Cochrane Review Centre in June 2013. The existence of the raw data and the validity of the results published in the trials of heart failure have been questioned to the point that the journal Heart also had to retract a paper using the same trials and co-authored by Taylor RS . It is important that the readers are fully aware of these facts, not acknowledged here but all into the public domain, that raise issues about the scientific rigour applied on this occasion by authors, reviewers and editors.
Conclusions. Following previous recommendations, we should consider what the data show : using inaccurate measurement of sodium intake and a self-selected population of older man and women, the study confirms that a higher salt intake is associated with higher mortality even in this group and that there is no evidence of harm for levels of sodium below the current global recommendation of 5g of salt per day (equivalent to <2000mg of sodium) .