Track Calories Without Obsessing

The risk is real, it is concentrated, and it is not evenly spread#

You can log food for years at no psychological cost, and you can also log food in a way that quietly becomes the problem — and both are documented in the same literature, about the same app. Among 105 adults who had recently left residential or day-patient eating-disorder treatment, 74.3% had used MyFitnessPal; of those, 73.1% said it had at least somewhat contributed to their disorder and 30.3% said it contributed very much¹. Hand the same app to 200 undergraduate women screened as low-risk, randomize half of them to log for a month, and eating-disorder scores do not move: β = −0.04 (95% CI −0.11 to 0.02) on the EDE-QS³.

Neither number is the "real" one, and the distance between them is the practical guidance. Whether tracking hurts depends far more on who is doing it, and on what rule they attach to the number, than on the act of logging. So the guardrails worth having are not "log less." They are: know whether you are in the group the evidence actually flags, keep the rules around the number graded rather than absolute, and read the log at the resolution it can support — which, given that even a careful day carries a margin of roughly ±20%, is far coarser than a four-digit total implies.

Where the studies split, and what separates them#

The cross-sectional literature is consistent, and it is not reassuring. Among 493 college students, those using calorie trackers showed higher eating concern and higher dietary restraint after controlling for BMI, and fitness tracking was independently associated with eating-disorder symptomatology even after adjusting for gender and past-month bingeing and purging². In Levinson's clinical sample, perceived app contribution correlated with restraint (r = .25), shape concern (r = .34) and weight concern (r = .36).

The experimental literature points the other way. Hahn's participants were not half-hearted about it — they logged 89.1% of days — so the null is not a compliance artifact. And in an eight-week study where 68 women used MyFitnessPal while repeatedly reporting their state, tracking frequency was associated with weight and shape concerns at the trait level, but daily tracking did not predict next-day scores on any outcome; within a day, tracking at one timepoint was followed by lower reported restraint at the next⁴.

That pattern — a between-person association with no within-person effect — is what selection looks like. People already preoccupied with weight and shape are more likely to reach for a tracker, which produces the correlation without the tracker having produced the preoccupation. A systematic review of 27 studies reached the same standoff and declined to resolve it: the cross-sectional association was reasonably consistent, it "was not replicated in experimental research," and the direction of the relationship cannot currently be established⁵.

The between-person association is there and the day-to-day effect is not. That is the signature of a tool people in difficulty reach for — not proof that it never puts them there.

Which is exactly why the null should not be spent too freely. Hahn's sample was screened for low baseline risk, and her own conclusion is that higher-risk groups need studying because self-monitoring "may be contraindicated" among them. Moody's review found the evidence base was almost entirely white female undergraduates with a BMI under 30, with clinical eating-disorder populations barely represented — while noting that among participants who did carry a diagnosis, 66% reported increased app use. The reassurance is real and it is narrow.

The scary prevalence figures come from a broken instrument#

If you have read that a quarter of health-conscious people have orthorexia, that number almost certainly came from the ORTO-15 questionnaire, and it should be set aside. The scale has low internal reliability and no first- or second-order one-factor structure, which means its items must not be summed — doing it anyway produces a total score with no interpretable meaning. Its content is worse than its statistics: some items never mention healthy eating at all, and others score ordinary behaviour as pathology, including being willing to spend more money on healthier food. Applied at scale it returned a prevalence of 27.5%, which would make orthorexia more common than every established eating disorder put together⁷.

This does not make the underlying worry imaginary. Levinson's patients are not a measurement artifact. It means the self-check you run on yourself should not be borrowed from a questionnaire that flags label-reading as a symptom — because a scale that catches everyone catches nothing, and it makes a genuinely worrying pattern harder to see rather than easier.

Rigid and flexible restraint is the distinction with data behind it#

The most useful split in this literature is not tracking versus not tracking. It is how a target is held. Two subscales built and validated across a 54,517-person weight-reduction program with seven-day food diaries and a random population sample of 1,838 adults separated rigid control — all-or-nothing rules, forbidden foods, a line you either hold or break — from flexible control, meaning graded adjustment: a bit less today, a bit more tomorrow. Rigid control went with higher disinhibition, higher BMI, and more frequent and more severe binge episodes. Flexible control went with lower disinhibition, lower BMI, fewer and milder binges, and a higher probability of successful weight reduction over the program's year⁶.

Hold that finding loosely in one respect: it is observational, drawn from people who chose a commercial weight-loss program plus a survey sample, so flexible control may partly be a marker of an already-easier relationship with food rather than its cause. What survives the caveat is the direction. The same calorie target can be held two ways, and the two ways do not behave alike.

That translates into concrete settings on a food log:

Decision	Rigid version	Flexible version
The target	1,800 kcal exactly, every single day	a weekly band you land inside most days
A day over	the day is blown; punish tomorrow	one point in a seven-point average
A missed log	the streak is broken	a gap; the week still reads
Precision	weigh everything	weigh the repeats and the oil, eyeball the rest
Foods	allowed and forbidden lists	nothing forbidden; frequency varies
Duration	indefinite, by default	a defined stretch, then a deliberate stop

Two of those rows have their own evidence elsewhere: reading a band rather than a daily verdict is the daily-versus-weekly review question, and the precision row is settled more decisively than most people expect in how precise tracking actually needs to be.

Most of the reported damage sits in features, not in counting#

When 24 university women with eating-disorder histories — 83% in recovery, with a mean of 30 months of app use — were surveyed, talked through real apps aloud, and interviewed, what they named was rarely the calorie number itself⁸. It was the furniture around it. The end-of-day projection was read as an instruction rather than a warning: one participant described the forecast as telling her that "if you keep on under-eating, you're going to be 98 pounds, which is exactly what you want." A total rendered in red carried a verdict — "that red number would scare me a lot" — while a green one set up a reward loop. Streaks and reminders drove compulsive daily logging and anxiety about stopping. Barcode scanning pulled the diet toward pre-packaged, scannable foods and away from anything unmeasurable. Gamification recast undereating as winning: "trying to beat the calories."

That is a small qualitative sample of women who were already ill, so it describes mechanisms rather than incidence. But notice that streaks, projections and colour-coded verdicts are product decisions, not consequences of counting — a distinction worth stating plainly by a blog published by a company that builds a tracker. Most of them can be switched off, and the ones that can't are a reason to change tools rather than to stop measuring.

Signals worth noticing, and how to loosen#

The useful self-check is behavioural, not a questionnaire score. Is the log recording what you eat, or deciding it? Does a meal you cannot log — a friend's cooking, an unfamiliar restaurant — produce disproportionate distress or get avoided? Are you steering toward barcoded food because it is countable? Does the streak matter more to you than the trend? Does the day's total set your mood? Any one of those is a signal to loosen the rules rather than to log harder.

Loosening has specific moves. Read the week instead of the day. Log the meals that teach you something and skip the ones you already know. Put an end date on the stretch rather than leaving it open-ended — and when the teaching is done, step down to a lighter proxy for the deficit or hand part of the job back to appetite. Above all, read the total as the band it always was; a figure you cannot precisely fail is a figure that cannot become a verdict.

One limit deserves to be stated without softening. If you have a history of an eating disorder, the reassuring trial evidence does not cover you: Hahn's study excluded elevated-risk participants by design, and the systematic review found clinical populations essentially unstudied. That is a gap in the evidence rather than a clean bill of health, and it is a conversation to have with a clinician who knows your history — not one to settle from an article.

FAQ#

Does calorie counting cause eating disorders?#

The evidence does not support that as a general claim. Cross-sectional studies consistently find tracker users score higher on disordered-eating measures, but experimental work has not reproduced it, and a review of 27 studies concluded the direction of the relationship cannot yet be established⁵. In a randomized month of logging among low-risk women, symptom scores were unchanged³. For people already unwell, the picture is different — most patients in one clinical sample said the app made things worse.

Is it a bad sign if I feel anxious when I can't log a meal?#

It is worth taking seriously. Women with eating-disorder histories specifically described streaks and reminders producing compulsive logging and anxiety about stopping⁸. Occasional irritation at an unloggable dinner is ordinary; consistently avoiding meals you cannot measure, or feeling the day is ruined by a missing entry, is the all-or-nothing pattern that tracked worse outcomes in the restraint research.

Can someone with a history of an eating disorder use a calorie app safely?#

The research cannot answer that, and the absence should be read as caution rather than permission. Trials showing no harm deliberately recruited low-risk participants, and reviews find clinical populations barely studied⁵, while retrospective clinical reports are strongly negative¹. This is a decision to make with a clinician who knows your history, not one to make from a prevalence statistic.

Sources#

This article was researched and drafted with AI assistance and reviewed for accuracy by the BurnWeek team. It is general information, not medical advice. How we research and correct our articles →