Skip to content

When you choose to publish with PLOS, your research makes an impact. Make your work accessible to all, without restrictions, and accelerate scientific discovery with options like preprints and published peer review that make your work more Open.

PLOS BLOGS Absolutely Maybe

Am I Going to Need a Smaller Plate? In Which I Juggle a New Weight Loss Trial & Old Systematic Reviews


Cartoon of a sad big plate and a dancing small plate


It just seems logical, doesn’t it? Super-sizing your food can super-size you. So when I used to see claims that a bigger plate = a bigger you, it made sense.

But then a new trial rolled in, straight into my wheelhouse. It’s got everything! Old, conflicting systematic reviews! Research integrity issues! And personal interest. I’d like to lose some weight. I don’t want to have to buy all new tableware, but hey, if it’s that easy, then I guess I should think about it. I rolled up my sleeves and got stuck in.

So here are the characters in this story. First, the new trial. By Daina Kosīte and colleagues, it’s decent-sized for this kind of study – 134 adults. They were randomized to large or small plates for a self-service lunch in a purpose-built laboratory that’s meant to be home-like, but with concealed video cameras the participants are told about. They didn’t know it was how much they would eat that was being studied, though:

To conceal the true purpose of the study, at the time of recruitment, participants were informed that the study was examining the impact of time of the day on a range of mental processes, and that they had been allocated to a lunchtime session.

They had half an hour for lunch, in a “lounge room” with a TV going, and a heated food trolley.

This trial had lots of the pillars of good science: it was fully pre-registered, with a data analysis plan posted early on, too. Randomization was by an external statistician, and the data analysis was, too. The people preparing and serving meals had nothing to do with the study.

The result?

There was no clear evidence of a difference in consumption between the two groups… There was no evidence of impact on meal micro-structure, with the exception of more food being left on the plate when larger plates were used.

Next up: the systematic reviews. The trial authors refer to 3 and the reviews’ conclusions aren’t all in line – I didn’t go looking if there are others.


Cartoon of dueling meta-analysts


The first, by Robinson & co, is just over 5 years old, with a search strategy in early 2014. Given that there are newer ones, let’s leave it aside. This was the conclusion, though:

Evidence to date does not show that dishware size has a consistent effect on food intake, so recommendations surrounding the use of smaller plates/dishware to improve public health may be premature.

The second is a Cochrane review with meta-analysis, by Hollands & co, which is about all sorts of portion size questions, including plate size. This review is pretty old. It has studies up to January 2015, although they haven’t incorporated the most recent ones they found. They don’t think those would change their conclusions.

This review meets my criteria for a systematic review worth taking a closer look at. The authors found a “small to moderate effect” of dinnerware size on food consumption by adults. There’s authorship overlap with the new Kosīte trial: 2 are co-authors of both the trial and the Cochrane review.

The third is by Holden & co, also with meta-analysis. They say it’s about plate size, but when they say “plate”, they mean any kind of container. This review didn’t get past most of my knockout criteria, and there were all sorts of problems. I can’t tell how old it was, because they didn’t report when they did their search. But it was published in 2016. And they concluded: “varying the size of the container holding food… has a substantial effect on amount self-served and/or consumed”.

And the last character in this story brings in the research integrity complicator. It’s Brian Wansink, the prominent nutrition researcher who had to leave Cornell University this year after investigation of his body of work left a trail of retractions – 15 by late last year – and even more questions in its wake. It’s the case uncovered by the so-called “data thugs”, Nick Brown, Tim van der Zee, James Heathers, and Jordan Anaya. That’s a story that’s really worth reading, if you don’t know it.

Now we have 2 big questions: what impact would the new trial have on those systematic reviews, and how does the meltdown of Wansink’s work affect them?

When you plough through and pick out the studies on plates and bowls alone, the 2 systematic reviews actually end up covering quite similar territory. By my count, the Cochrane review includes 12 studies specifically on dinnerware alone, with 764 people (including 129 children). The studies were small, but there was one with 220 people at a congress meal (although it was only a conference abstract). None of them were in everyday life.

Of the 9 studies with risk of bias assessed, all were “unclear risk” except one that was at high risk of bias. That’s a weak evidence base. I used their criteria on the new Kosīte trial, trying to do it consistently with theirs, and I judged it to be at low risk of bias. It’s a set of studies with very mixed results, and I think the new big trial pushes it clearly into “no strong evidence of benefit” territory.

The Cochrane review has 1 study that Holden doesn’t have – the one with the 220 people. Meanwhile, the Holden review has 6 studies that Cochrane doesn’t have, but I couldn’t find one of them in their references, so I don’t know what it’s about (Chang 2007). Two of them were excluded by Cochrane because of their less rigorous study type (Yip 2013 and Libotte 2014, both of which found no impact).

This one might fit the question, though: Robinson 2015, 61 people eating popcorn while watching TV in a lab (no impact of bowl size). The fifth one definitely wouldn’t fit: it didn’t involve food consumption (Sharp 2012). And the sixth is one of the retracted Wansink studies, which is probably another misfit here anyway (2008).

Which brings us to the impact of Wansink studies. In the Cochrane review, 42% of the participants were in studies authored by Wansink, including 1 that van der Zee includes in his Wansink dossier as having “serious issues with (reporting) accuracy/veracity”. (It’s the ice cream one, from 2006.) And then there’s this red flag: for one of the included studies, the Cochrane authors report, “Study authors contacted for missing data but data no longer available”. Not being able to produce the data is the reason for some of the Wansink retractions.

Where does all this leave us? The systematic reviews end up being rather misleading, I think, in part because they have lumped too many disparate interventions together and obscured the weakness of the evidence base. The “claiming benefit” side of the ledger is compromised, too. And now we have a piece of strong evidence that doesn’t show a benefit. None of it gets directly at the question of weight loss and whether it’s worth buying a new dinner set.

Count me down as endorsing the Kosīte trial authors’ conclusion:

This study suggests that previous meta-analyses of a low-quality body of evidence may have considerably overestimated the effects of plate size on consumption. However, the possibility of a clinically significant effect – in either direction – cannot be excluded. Well-conducted trials of tableware size in real-world field settings are now needed…




Disclosures: I know a couple of the authors involved in the new trial and the Cochrane review, and I am on the advisory group for the UK Reproducibility Network, of which one is a leader (Marcus Munafò). I have not discussed the study or this post with them before publication. Other than being overweight and using plates, I have no interest, commercial or otherwise, in this topic.


The cartoons are my own (CC BY-NC-ND license). (More cartoons at Statistically Funny and on Tumblr.)



Leave a Reply

Your email address will not be published. Required fields are marked *

Add your ORCID here. (e.g. 0000-0002-7299-680X)

Back to top