A study published today in BMJ Open on salt and stroke has gotten a lot of press. I decided I’d better read it: 1) since my daddy died of a stroke 2) I have high blood pressure when I’m not regularly exercising and 3) I love salt.
The evidence to date has been pretty unconvincing. Plagued by poor quality and inconsistent results, it doesn't provide the consistent, strong evidence of a causal relationship between salt intake and death from stroke or heart disease that I seek. The inconsistency in study results leads to controversy and different interpretations of the data we have. This is what often happens when the data is just not clear.
So I read the new study, which was easy to do since it’s open access. You can read it, too.
What I didn’t expect was to find another ecological study behind all the news hype.
I’m an epidemiologist, so I have no problem with use of observational studies as evidence. For some questions, particularly about possible harms and lifestyle exposures, they can provide compelling evidence when randomized trials would be unethical. Cigarette smoking is one example. Sir Richard Doll and others provided us with consistent, strong evidence of an association between cigarette smoking and lung cancer. Later observational studies established a similar relationship between cigarette smoking and deaths from heart disease and stroke. Remember this; it will be important to this story, too.
There are good study designs using observations, most notably cohort studies and sometimes case-control. I’ll spare you the details of these study designs for now, but they make use of good records or use a long period of following subjects to establish the association between the exposure, in this case eating salt, and the outcome, in this case stroke and heart disease. The ecological design does not. Instead it makes use of convenient data.
A common characteristic of this study design is that data is not available for individual people. Ecologic studies make use of group data sources to supply information for the different components of interest, and then put the data together to look at patterns. The salt consumption in this study was collected as part of periodic surveys in the UK and the death data came from the Office for National Statistics for England and Wales. Like other non-randomized study designs, it is very important to adjust for other possible factors that are associated with death from stroke and heart disease and may be associated with eating salt. This adjusting is a bit trickier with ecologic studies than other designs. So I looked for how the authors handled potential confounding risks. Imagine my surprise when I read, then re-read the study to find that the authors did not perform any adjusted analyses to evaluate the relationship of salt intake to deaths from stroke and heart disease. NONE.
They did do some adjusted analyses when looking at a relationship between blood pressure and salt intake. They adjusted for a lot of things, as shown in their summary statement:
“…our findings that, in untreated individuals, there was a fall in BP of 2.7/1.1 mm Hg after taking into account age, sex, ethnic group, education level, household income, alcohol consumption, fruit and vegetable intake and BMI suggest that the reduction in population salt intake, which occurred between 2003 and 2011, is likely to be an important contributor to the falls in BP. “
Please read those factors for me again. Do you see cigarette smoking? You do remember my caution from above? Cigarette smoking is a very important risk factor for deaths from stroke and heart disease and for high blood pressure.
I checked the Table 1 to be sure that smoking information was available to the authors. Yes, indeed, it’s right there in the table, and yes, there was a significant decrease in smoking over the same time period that salt consumption decreased.
Why wasn’t this in the adjustment for blood pressure? And why was there no adjustment for stroke and heart disease deaths? I did not find explanations in the published study. Did you?
Were the editors not curious about this? They had the opportunity to ask the authors to please clarify this for the readers before publishing this work. How much of the population reduction in deaths from stroke and heart disease might be associated with the reduction in smoking? Smoking dropped from 19% to 14% in the population.
So ultimately I’m not able to use this study to clarify the question I still have about salt consumption and my stroke risk. I already knew I should not smoke since I knew that evidence and had observed that my dad, and in fact all my other relatives who died of stroke, were heavy smokers.
In the meantime I try to limit my salt consumption, thinking it might help and likely wouldn’t hurt. But summer’s coming and I’m already craving BLT sandwiches made with homegrown tomatoes, and watermelon sprinkled liberally with salt…