On Tuesday 1st April I thought I’d spotted the April Fool joke for the day when I woke up to headlines such as “Seven-a-day fruit and veg ‘saves lives’” and “Diet that adds years to life.” Sorry BBC – no lives will be saved – we’re all going to die. We’re also not going to live for years longer by having 7 fruit & veg a day.
The first thing that I noticed was that the abstract (in italics below) declared that everything except smoking had been adjusted for. Table 1 then showed that 39% of the 0-1 portion a day people smoked vs. 10% of the 7+ portions a day people.
“Methods Cox regression was used to estimate HRs and 95% CI for an association between fruit and vegetable consumption and all-cause, cancer and cardiovascular mortality, adjusting for age, sex, social class, education, BMI, alcohol consumption and physical activity, in 65 226 participants…”
I emailed the named person on the report who described leaving this out as a “terrible oversight.” Jolly good of her to get back to me so quickly, as Dr Longo still hasn’t. I’d already done a couple of interviews before the reply, during which I said that the abstract didn’t record having adjusted for smoking. There is a huge lifestyle thing going on here, however and I don’t know how you can adjust for an entire lifestyle.
The study did a similar thing to the Longo animal protein study – they didn’t divide the subjects studied into equal sized groups – quintiles. They divided them into uneven groups of those whom (when asked what they ate yesterday – yes, I know – usual huge caveats) replied:
0-<1 portion of fruit & veg (15.2%);
1-<3 portions of fruit & veg (30.9%);
3-<5 portions of fruit & veg (29%);
5-<7 portions of fruit & veg (15.7%); and
7+ portions of fruit & veg (9.3%).
By allocating people unevenly to researcher defined groups (0-<1; 1-<3; 3-<5 etc) rather than to quintiles, an artificial baseline and opposite extreme can be created. I use the analogy – it’s like doing a study about height and using basketball players as the benchmark.
All the headlines then came from comparing the people who didn’t even have 1 portion of fruit or veg the day before, with the people who had 7+ portions. Let’s see Table 1 for some further comparisons of these two extreme groups, which don’t even account for 25% of all people in the study. Let’s call them baddies (0-<1) and goodies (7+) and then let’s play the relative comparison game:
– Smoking is four times higher in baddies;
– Inactivity is four times higher in baddies;
– Baddies are almost twice as likely to be in manual jobs;
– Baddies are more than twice as likely to have no qualifications. The goodies are four times more likely to be university educated.
This isn’t just about fruit and veg – it’s about a whole lifestyle of advantaged people and disadvantaged people. Do researchers really think that we could take a Glaswegian man or woman – with a life expectancy of 71.6 and 78 respectively – and give them the life expectancy of the men and women in Kensington and Chelsea – 85.1 and 89.8 respectively – with extra portions of fruit and vegetables? May I suggest that the fruit and veg intake is a marker of a healthy lifestyle and not the maker of one?
The April 2010 publication from the European Prospective Investigation into Cancer (EPIC study) did allocate people to quintiles (even groups). It reviewed almost 142,605 men and 335,873 women people across 23 different European centres over 8.7 years and concluded:
“Conclusions A very small inverse association between intake of total fruits and vegetables and cancer risk was observed in this study. Given the small magnitude of the observed associations, caution should be applied in their interpretation. “
The November report from the EPIC study noted: “The possibility that fruit and vegetables may help to reduce the risk of cancer has been studied for over 30 years, but no protective effects have been firmly established.”
Then today’s study, with unevenly allocated groups claimed “A robust inverse association…” Um…
From Association to Relative Risk
The press release said that “Eating at least seven daily portions was linked to a 42% lower risk of death from all causes and from cancer and heart disease/stroke of 25% and 31%, respectively, after excluding deaths within the first year of the monitoring period.”
Why would deaths within the first year be excluded? Exclusions must be explained and have a valid rationale. ‘The numbers looked better when we did this’ is not a valid rationale. Table 2 gives us the Hazard Ratios, which provide the headlines:
|Portions of fruit & veg consumed in the previous day HRs|
|Model 2 (*)||1||0.88||0.76||0.7||0.67|
|Model 2 (*) – excluding deaths within a year||1||0.86||0.71||0.64||0.58|
Model 2 is said to have adjusted for sex, age-group, smoking, social class, BMI, education, physical activity and alcohol intake.
The “42% lower risk of death” comes from 1 minus 0.58 in this bizarre “excluding deaths within a year” manoeuvre. The line above this, which doesn’t make this exclusion, would give a relative risk of 33% (1 minus 0.67). This table is merely saying – when we picked this extreme benchmark group of people who just don’t eat fruit and veg and gave these guys an incident rate of 1, the ‘goodies’ had an incident rate of 0.67. So, for every 1 baddie death observed, we observed 0.67 goodie deaths. Already not so headline grabbing.
Then look at the actual deaths rate. There were 4,399 deaths from a population of 65,226 people. The average length of follow-up was 7.7 years. That means the death rate was 6.7% during the whole study. Or 0.88% in any one year of the study. That’s 8.8 people per 1,000 who died of any cause. (The subjects were all over 35 – no upper limit was given).
To achieve the relative difference and maintain the overall death rate average of 0.88%, the death rate in the baddie group would have been approximately 10 in 1,000 and the death rate in the goodie group would have been approximately 6-7 in 1,000 (take 33% or 42% – the absolute difference is still tiny).
Can you imagine the following as the headline this morning?
“Study finds people who ate 7+ fruits and veg yesterday are associated with 6-7 deaths per 1,000 people, while people who didn’t even eat 1 portion notched up 10 deaths per 1,000 people.”
As I’ve said in previous study dissections – that wouldn’t make the headlines.