Below are notes I took from reading the paper: How much food tracking during a digital weight management program is enough to produce clinically significant weight loss?

It’s known that weight loss can be facilitated by lifestyle interventions.

One of the strongest predictors of how much weight loss will occur is how often participants track the food and drink they consume each day, usually with a view to remaining within a set calorie goal.

However, tracking your consumption in detail is effortful. Most people do not find it possible to do so over the longer term. One approach to improving adherence would be to reduce the amount of effort required, for example by reducing how often people are asked to track food or reducing the amount of their food they are asked to track.

It is not yet clear how much food tracking is needed to get to certain weight loss milestones. Studies so far suggest that “abbreviated” food tracking approaches do correlate with weight loss, with the possible exception of tracking by taking a photo of the food.

Using data from subscribers to a weight loss program over a 6 month intervention period, the study aims to:

  1. establish whether the abbreviated method of food tracking facilitated by by the program remains associated with weight loss.
  2. establish thresholds of food tracking that are associated with certain weight loss milestones

Aim 1 was tested with both a nonparametric and linear regression methods.

There was a significant correlation between the percentage of days participants tracked food with the percentage of their original weight that the participant lost after 6 months, r = 0.4, p < 0.001. A 1% increase in the number of days tracking food associated with a 0.08% greater in weight loss.

A ROC curve analysis looked at showed that food tracking was highly predictive of the clinically significant 3%, 5% and 10% weight loss thresholds after 6 months. The optimum predictivity threshold was defined by the sum of sensitivity and specificity.

  • For predicting achieving at least a 3% weight loss, the optimum threshold of food tracking corresponded to 29% of intervention days.
  • For 5% weight loss this was 39% of days tracked,.
  • 67% of days was the optimum threshold for predicting the 10% weight loss threshold.

Aim 2 added a time series clustering analysis method, dynamic time warping, to look for similarities between member food tracking behaviour.

This produced 3 clusters based on the number of days participants had tracked food in each week over the intervention period. The clusters differed by levels of food tracking, weight loss, participant age and marital status.

This study could not quantify the mechanisms why participants showed different food tracking behaviours. For instance, perhaps some had higher motivation or fewer barriers to food tracking and/or weight loss.

Limitations include:

  • Participants were biased towards being female, with college degrees and higher incomes.
  • This approach cannot fully establish that food tracking causes weight loss. There could be confounders.
  • Whilst the optimum thresholds for predicting 3%, 5% or 10% of weight loss were established, there was no step-change in cutoff that perfectly predicted weight loss success. We should be cautious about interpreting the specific thresholds here.
  • Food tracking was defined as tracking at least 1 food item on a given day; other definitions could be applied.

In conclusion, this study shows that participants can achieve clinically significant weight loss after 6 months even without perfect adherence to the recommended daily food tracking. In fact, no-one in this study tracked every day. Given that perfect adherence doesn’t seem to be possible for most people, and isn’t necessary for achieving significant weight loss, we should question whether it should be the default prescription for weight loss interventions.