The decision this page changes. Whether someone trying to lose weight should step on a scale regularly, how often, and whether a scale on its own, without a programme, does anything. The lever is cheap and easy to stop, so the questions are direction, size and which claim the evidence actually supports. The phrase self-weighing works runs together three different comparisons with three different answers, and the decision depends on which one applies. (inferred from Madigan et al., 2015)
The evidence base
One systematic review and random-effects meta-analysis of RCTs (Madigan 2015; searched to August 2014): 24 trials, 21 pooled, adults with overweight or obesity in weight-loss trials. Trials had 23 to 415 participants (median 110) and 15 of 24 ran in the USA; four enrolled only women and the rest were 26 to 91% women (median 75%). Weight was measured objectively in all but one trial. Missing weights were standardised to baseline weight carried forward (BOCF): «We used BOCF because this mitigates bias that may arise because participants that do less well may be reluctant to be followed up.» One trial was an exception: «One trial used a conservative method of imputation that was similar to this imputation, by adding 0.5 kg to the last weight observed carried forwards. The trial was included within the analysis as presented [21].» (Madigan et al., 2015)
What it cannot speak to: maintenance after loss. The review excluded maintenance trials by design: «Trials were excluded if they enrolled people after weight loss where the prime aim of the intervention was to prevent weight regain or trials that enrolled people that aimed to help prevent gradual weight gain.» (Madigan et al., 2015) Eleven full texts were excluded on these grounds. So weighing helps you keep it off is not tested here; see the maintenance section below.
Appraisal. Risk of bias was assessed with the Cochrane tool; most trials gave too little
information to judge sequence generation or allocation concealment (low risk in 11 and 7 trials
respectively; 4 high risk on concealment), 6 were high risk on attrition (of 23 reporting trials, though the counts sum to 24; the Methods and Results also give different attrition thresholds —), and selective
reporting could be checked in only 4. The review gives no GRADE or other certainty rating and does
not assess publication bias. [searched: GRADE / certainty / publication bias / funnel — zero across the source]
(Madigan et al., 2015)
A second review with a different scope (Zheng 2014). A narrative systematic review of 17 studies:
«On the basis of the remaining 13 RCTs, we further identified 4 relevant studies using the Cited Reference
Search strategy, bringing our total to 17 studies.» Its inclusion rule was RCTs or studies ancillary to
RCTs, and its tables state randomisation for only some of them
(Zheng et al., 2014); several are read
below as associations or quasi-randomised comparisons (table below). There is no reported pooled value. Of the 17, 12 are on weight loss or
maintenance and 5 on preventing weight gain. 15 ran in the USA and 2 in the UK; samples ran from 32 to 3768 people,
61-100% women and 62-91% White. It took in maintenance and gain-prevention studies, and psychological outcomes are one of its two
aims. It reports no risk-of-bias or
certainty appraisal and no search end date; the only appraisal language is that sample homogeneity
«also influenced the level of evidence». Weighing frequency was self-reported in 14 of the 17.
(Zheng et al., 2014)
Madigan excluded maintenance and gain-prevention trials; it also set out to assess psychological
harm, and reports it in an adverse-events section. (Madigan et al., 2015)
[searched: quality / bias / GRADE / PRISMA / Jadad / level of evidence — only quality of life, a reference title and the one limitations sentence; srcgrep across the 1-chunk denominator]
The two reviews share five trials (Gokee-LaRose 2009, Lally 2008, Linde 2011, Steinberg 2013, VanWormer 2009) of Zheng’s 17 and Madigan’s 24; the review author lists do not overlap, though Madigan’s group ran one Zheng-included maintenance trial, and Madigan cites Zheng and searched its reference list. Where the reviews agree on those five, the agreement is the same data read twice, so it cannot count as type E. (Madigan et al., 2015; inferred from Zheng et al., 2014)
Three comparisons, three answers
All figures are between-group mean differences in weight at programme end (intervention minus control). The text and Table 3 give no pooled control-arm weight change, so no absolute baseline sits beside them. (Madigan et al., 2015)
| Comparison | Control arm | Trials (n) | MD kg (95% CI) | I² | 95% prediction interval | Timepoint |
|---|---|---|---|---|---|---|
| Scale alone | no intervention | 1 (183) | -0.5 (-1.3 to 0.3) | — | — | end of a 3-month intervention |
| Self-weighing plus other self-regulation parts, added to a programme | the same programme | 4 (274) | -1.7 (-2.6 to -0.8) | 0% | -7.5 to 4.1 | programme end, 10 weeks to 7.2 months |
| A programme that includes self-weighing | no or minimal intervention | 15 (2490) | -3.4 (-4.2 to -2.6) | 82% | -6.9 to 0.1 | programme end |
Scale alone: insufficient evidence, not no effect. One trial, run by the review’s authors, tested advice to self-weigh without a programme and found -0.5 kg (-1.3 to 0.3) after three months. The authors’ own reading moves in two steps: «How- ever, unlike Michie and colleagues, we found that self- monitoring alone was ineffective for weight loss. However, only one study investigated this and the estimate was imprecise enough to encompass effects that would be worthwhile.» (Madigan et al., 2015)
Added to a programme: a direction, with the scale’s share not isolated. Four trials gave both arms the same programme and added self-weighing to one arm. The review’s Results text describes an extra part in all four, though its sentence on three of them cites only two trials: one added training in using the scale like a blood glucose monitor plus weight feedback, and «The other three trials gave participants the option to record their diet and physical activity [38, 39].» The Methods had planned a split of this set: «Within this group, there were two subgroups: trials where self-weighing was the only addition to the behavioural programme and trials where several self-regulatory in- terventions including self-weighing were added to the behavioural programme.» Neither the Results nor Fig. 2 reports that split (Fig. 2 pools the four as one set, and Table 2 codes all four as the same programme plus «self-monitoring/ self-weighing techniques»), so whether any of the four added the scale alone is unclear. The authors’ summary: «There was evi- dence that adding advice to self-weigh to a behavioural programme improves its effectiveness, but only four trials have assessed this, and the estimate of effect was impre- cise and clouded by the use of other self-regulatory ele- ments.» Heterogeneity was 0% (Fig. 2 prints tau² = 0.00) across the four, but the printed prediction interval runs from -7.5 to 4.1 kg; an interval that wide is hard to square with a tau² of zero, so the width may be a reporting error, though the authors’ reading is: «in some interventions participants would lose a considerable amount of weight but in others interventions participants may gain weight.» Three of the four asked for daily weighing. In the forest plot (Fig. 2) only one trial’s interval excludes zero (Wing 2010, -1.9, -3.0 to -0.8); the other three cross it. (Madigan et al., 2015) (Madigan et al., 2015)
How much the comparison arm weighed in one of the four is unclear (Zheng 2014). Gokee-LaRose 2009 is in Madigan’s add-on set, coded «Daily», at a 10-week timepoint. Zheng’s row for it describes the arms as: «During the 10 weekly group meetings, BSR were given digital memory scales and were instructed to SW and use color zones, while SBT were not. After group sessions, SBT were encouraged to SW weekly at home but were reminded to place primary emphasis on their behaviors, not on their weight.» The same row then prints «At 10 wk, 95% BSR vs. 94% SBT SW daily.» and «At 20 wk, 70.6% BSR SW daily vs. 17.6% SBT SW weekly; no SBT Ps SW daily.» Madigan defines the exposure by instruction: «Self-weighing was defined as participants being asked to weigh themselves rather than being weighed as part of a programme.» It excluded «n = 9 both groups self-weighing» trials but kept this one. (Zheng et al., 2014) (Madigan et al., 2015) The row does not settle what the comparison arm was asked to do during the 10 weeks. After group sessions can mean after each weekly session (the arm was asked to weigh weekly throughout) or after the 10-week programme (it was not asked during it). Madigan keeping the trial fits the second reading but does not prove it. The 94% daily figure fits neither instruction cleanly and sits against the row’s 20-week reading; it may be a misprint, or that arm weighed more often than asked. Zheng’s adherence text gives 90% of participants weighing daily at the end of the 10-week programme, which matches neither the 95% nor the 94% figure exactly and so does not decide between the readings. So this trial’s 10-week contrast was the self-regulation package against either weekly weighing, daily weighing, or no instruction to weigh. Either way the sign of the -1.7 kg is unchanged; the primary paper would settle it.
A programme that includes it, against minimal care: the programme’s effect. The -3.4 kg compares a whole multi-component programme that includes self-weighing with no or minimal intervention. The authors’ reading of it: «There was strong evidence that behavioural weight loss programmes that incorporate self-weighing are more effective than minimal interventions.» Their scope statement covers all the pooled numbers, though elsewhere they call the mean differences «the net effect of the self-weighing inter- vention itself» and conclude «Self-weighing as part of a multicomponent programme is effective in facilitating weight loss». The scope statement: «Our aim is not to produce a definitive estimate of the effect of self- weighing on weight loss at a particular point in time; rather it is to find evidence that self-weighing as a tech- nique is effective.» The next two sentences say how to read the numbers: «As length of follow-up is the same in both intervention and control groups, any differences are down either to random variation or differences in effectiveness of self-weighing. Thus the estimates we produce should not be read as estimates of the effect of self-weighing, but as valid evidence that self-weighing can be effective in these contexts.» They add: «It was impossible to separately code those techniques used to promote self-weighing and those that related to other components of the intervention.» (Madigan et al., 2015) (inferred from Madigan et al., 2015) The -3.4 kg and the -1.7 kg come from different trial sets, so the -3.4 kg cannot be split into a programme share and a scale share. The closest estimate of adding self-weighing is the add-on comparison, -1.7 kg (-2.6 to -0.8), and it too carries co-delivered parts, so it is not the scale’s effect alone either; the authors’ reading above (evidence that self-weighing can be effective in these contexts, not an estimate of the effect of self-weighing) applies to it too. Read -1.7 kg as the held magnitude for adding self-weighing, with its co-delivered parts, to a programme like these, not as the effect of the scale by itself; and in one of the four trials how often the comparison arm weighed is unclear (above).
What the second review adds — associations inside trials, not effects
Zheng’s inclusion rule asks for RCTs, but most of its weight findings are not randomised contrasts. The common form compares people who weighed more often with people who weighed less often inside one trial or cohort, which is observational. The review’s design labels have to be read off its tables (inferred from Zheng et al., 2014):
| Design of the weight result | Studies | What they found |
|---|---|---|
| Within-study frequency association | Sherwood 2006 (1801), Jeffery 1999 (809), VanWormer 2012 (1747), VanWormer 2009 (100; 46% vs 8% reaching 5% loss, «≥ weekly» vs «≤ weekly» as printed), Welsh 2009 (73), Gokee-LaRose 2009 (40), Funk 2010 (348, one arm), Wing 2006 (314) | more weighing went with more loss, less regain or less gain |
| Within-study comparison, null | Kong 2012 (123 analysed) | no significant difference in adjusted % weight change, daily or more vs less than daily |
| Between-group trial, weight reported (among these rows Zheng says randomized only for French; VanWormer, Steinberg, Lally and Linde are RCTs in Madigan) | VanWormer 2009 (100), Steinberg 2013 (91), French 2011 (N = 90; households randomized), Lally 2008 (104), Linde 2011 (68) | -7.5 vs +1.3 lb at 6 mo, none at 12 mo (retention 13% and 38%); -6.55% vs -0.35% at 6 mo (a four-part programme); no significant BMI difference; Lally: intervention arms lost more than control at 8 wk; Linde: -0.72 ± 3.7 vs -0.22 ± 0.23 kg at 6 mo, no P printed |
| Other rows | Keep It Off 2013 (see maintenance; its row says randomized), Gokee-LaRose 2010 (52) | Keep It Off: weighing frequency only; Gokee-LaRose 2010: both arms told to weigh daily |
| Weighing isolated as the intervention | Levitsky 2006 (32 and 41; allocation not stated), Madigan 2013 (3768; quasi-randomised) | see gain prevention and maintenance below |
The review’s own text is stronger than its tables: «All of the 17 studies reviewed reported that more frequent self- weighing was associated with greater weight loss (27,43- 45,48,49,52), less weight regain (36,41,47,51,54) and better weight gain prevention (55-59).» Kong (ref 49) is in that list, while its row reads «No significant difference in adjusted mean percent WC was observed in the ‘‘daily or more’’ vs. the ‘‘less than daily’’ group (49).» The next sentence qualifies it: «However, the effect sizes varied across the studies (Figure 2).» Figure 2 plots per-study correlations and draws an unlabelled summary diamond at about -0.1, with no value, method or heterogeneity printed (direct PDF read); in it the intervals for Kong, Lally and Linde 2011 cross zero. Lally’s row is a second frequency null («no difference between weekly and monthly SW subgroups»), and the review’s own text says so a few sentences later: «Lally et al. reported no differences between weekly and monthly weighing (43)». French 2011 (ref 57, in the gain-prevention list) does not test the sentence’s quantity: its row reports only a between-household null on BMI and no frequency-weight association. VanWormer 2009 fits the sentence: its row prints a frequency association (46% vs 8%, no timepoint given; the text credits the weekly association to the 12-month paper, ref 45) beside a 12-month null that does not say which contrast it refers to. On Kong’s and Lally’s rows the all 17 sentence overstates; the other rows mostly fit it or do not address it. (Zheng et al., 2014)
(inferred from Zheng et al., 2014) For the associations,
reverse causation is the first alternative: people who are losing weight may weigh more because the
number is good news, and people regaining may stop. Madigan names the same problem for observational
maintenance data (section below); Zheng states no causality caveat
[searched: causal / reverse — zero across the 1-chunk denominator]. So the associations show the
behaviour and the outcome travel together; they do not size what the scale adds. On the weight-loss
question the randomised estimates on this page stay Madigan’s.
How often to weigh — daily vs weekly is a between-trial contrast
Within the 15 programme-vs-minimal trials, seven asked for at least daily weighing (-3.6 kg, -5.4 to -1.8; I² 91%) and eight for less than daily, seven of them weekly (-3.3 kg, -4.0 to -2.5; I² 65%); subgroup difference p = 0.57. These are different trials with different programmes, not a randomised comparison of frequencies: «Our analyses address- ing theoretical constructs were analyses across trials and therefore observational, as no trial directly addressed these issues.» The authors conclude: «There was no evidence that daily weighing led to greater weight loss than weekly weighing and it ap- pears that both may be effective when combined with multi-component interventions.» (Madigan et al., 2015)
No difference shown is not equivalence. The daily subgroup’s interval is wide (-5.4 to -1.8) and its prediction interval (-10.2 to 3.0) crosses zero, while the less-than-daily prediction interval (-4.6 to -1.0) does not. In the add-on set, weekly weighing rests on one trial (Allen); the other three asked for daily. Adherence was measured in only three trials, all daily per the text (44%, 50%, 95%), so whether people keep up daily weighing as well as weekly is unknown. The frequency coding is not consistent inside Madigan: the text lists VanWormer [21] among the at-least-daily trials («[18, 21, 35–37, 43, 44]») while Table 2 codes it «Weekly», and Zheng describes Lally’s arms as weekly vs monthly where Table 2 codes «Daily» (below), so the daily and less-than-daily subgroups may not be cleanly split. A within-trial randomised daily-vs-weekly trial was in progress when the review was written. (Madigan et al., 2015) — closes when a pooled or single randomised daily-vs-weekly contrast inside the same programme is held.
Zheng’s frequency evidence does not settle it either. Inside single studies, Welsh found daily weighers lost more than those weighing weekly or less, and in VanWormer 2012 daily and weekly weighers lost about 1.8 and 0.9 kg, «which was more than Ps who SW monthly»; both compare people who chose their frequency, so they carry the reverse-causation problem above. Where arms were told to weigh at different frequencies, Lally 2008 (104 people, 8 weeks) found «no difference between weekly and monthly SW subgroups»; the row does not say whether those arms were randomised [an RCT in Madigan], and Madigan’s table codes the same trial’s weighing frequency as «Daily». Zheng’s authors: «Although at this time, the evidence cannot not support endorsement of an ideal self- weighing frequency or duration for use of this strategy, the recent work provides more evidence that daily weighing is an acceptable strategy for weight loss intervention.» (cannot not as printed.) Gokee-LaRose 2009 also gave one arm scales, an instruction to weigh and a self-regulation package during the 10 weeks (Madigan codes it «Daily»), and encouraged the other to weigh weekly at home after group sessions (timing unclear, above); its printed 10-week adherence is unclear (above). (Zheng et al., 2014) (Madigan et al., 2015)
People weigh less often as time passes, at least early on. In Gokee-LaRose 2009 about 90% weighed daily at the end of the 10-week programme, falling significantly by 20 weeks; «This pattern of initial high level of adherence to self-weighing followed by a gradual decrease was reported by others (48,52).» A second study by the same group: 91% and 100% of the two arms daily at 8 weeks, 61% and 90% at 16 weeks. The Discussion: «Although five studies reported short-term (<9 months) patterns of self-weighing with a gradual decline over time, no studies have reported the pattern of self-weighing over an extended time period» (the PDF’s example is 12 months or more). Single readings at a year or later do exist in the tables: in Keep It Off, about 60% of the guided arm and 43% of the self-directed arm reported weighing daily at 24 months, and the text says participants weighed daily on 54% of days (the row: 1029 of 1932 surveys, 53.8%); in Kong, 88% weighed at least weekly at 12 months. (Zheng et al., 2014) (inferred from Zheng et al., 2014) Under adherence is part of the effect: early decline is shown, the longer-term course is described only by point readings inside supported trials, and the guided-vs-self-directed gap in Keep It Off suggests support helps keep the habit. That favours picking the frequency you will keep up, but it is not a shown dose difference.
Accountability — between-trial, and mostly about weight lost
Trials where participants «knew that they were accountable to a therap- ist/researcher» (14 trials, -3.6 kg, -4.6 to -2.7; I² 83%, prediction interval -7.5 to 0.3) beat the two without (-2.3 kg; Table 3 prints -3.2 to -1.5, the text -3.1 to -1.5; I² 0%, no prediction interval printed), p = 0.03. Two trials with strong accountability, where the therapist contacted anyone who did not weigh, found -5.6 and -8.8 kg. The authors add that elsewhere the accountability «was more closely related to weight lost and not the act of weighing», and the comparison is between trials, with only two trials on one side. (Madigan et al., 2015) (inferred from Madigan et al., 2015) Read it as a hypothesis that someone seeing the numbers helps, not as an established modifier: between-trial subgroups are held as hypotheses (CLAUDE.md, the over-personalization rule).
One within-trial subgroup, held as a hypothesis. «Pacanowski reported a subgroup analysis from a trial of self-weighing that found people with in- ternal weight locus of control and males lost more weight [44]. However, indicating that on average some groups respond better does not necessarily make these predictors useful screening tools to exclude people from self-weighing.» (Madigan et al., 2015) (inferred from Madigan et al., 2015) A single trial’s subgroup, reported secondhand, not known to be prespecified: a candidate modifier, not a reason to advise anyone differently.
Maintenance and durability — what is and is not held
(Madigan et al., 2015) Three trials followed people beyond the last contact: -8.0 kg (-12.5 to -3.5) about 18 months after, -0.3 kg six months after a single-session programme, and -7.5 kg (-11.3 to -3.7) nine months after; pooled across 185 people, -5.5 kg (printed interval -11.4 to 4.7), I² 86%. That interval is lopsided around -5.5, its lower bound repeats the second trial’s printed lower bound, and the same Table 3 row prints «0.04» in the column that holds the effect P for the other rows, which fits an interval excluding zero; so the interval, or the P, is probably misprinted. The three come from different comparison types (an add-on trial, a single-session programme, a programme vs minimal care). (inferred from Madigan et al., 2015) So the pool does not estimate the durability of any one of the three comparisons above. Two of the three single trials have intervals that exclude zero, but each is one small trial of a different comparison, the pool mixes comparison types with I² 86%, and its printed interval and P disagree; durability is thinly evidenced, not established.
The maintenance evidence the review cites is observational, and the authors lean toward it. They suggest self-weighing may suit the long term because it is easy to keep up, and cite the National Weight Control Registry: «Additionally because self-weighing is less cumbersome than recording diet and physical activity it might be a behaviour that can be continued and therefore help weight control in the longer term. The National Weight Control Registry has found that those who are successful at preventing weight regain, after weight loss, weigh themselves on a regular basis [46].» On a separate prospective study (ref 48; the quote’s closing marker cites a different paper) they add a caveat: «Higher weighing frequency was associated with greater weight loss and less weight regain at 24 months follow-up. However, greater motivation to maintain weight or success in achieving weight mainten- ance may motivate people to weigh themselves frequently, which makes observational data difficult to interpret [49].» (Madigan et al., 2015) (inferred from Madigan et al., 2015) The authors’ caveat names two alternatives, motivation (a confounder) and success prompting the weighing (reverse causation); both would apply to the registry association too, though they do not apply it there. The direction they suggest is plausible; no randomised maintenance estimate is held. Self-weighing for maintenance is therefore insufficient evidence in the held fabric, not no effect -> Weight-Loss Maintenance and Metabolic Adaptation. (Superseded in part 2026-10-29: one quasi-randomised estimate is now held secondhand, below; the state moves from insufficient evidence to one low-certainty estimate.)
Updated 2026-10-29 (Zheng 2014): one quasi-randomised estimate, held secondhand. Zheng includes the Madigan group’s maintenance trial, one of only two studies the review describes as isolating self-weighing in delivery (the intervention still bundled the offer of free weighing scales, calls and a record card). People who had recently lost weight were offered free weighing scales, two phone calls and encouragement to weigh weekly and record it on a card (3290 people), against no maintenance intervention (478). Over 12 months (3 months of intervention, 9 of follow-up) the intervention arm regained 1.23 ± 5.8 kg and the control arm 1.83 kg (printed «1.83-5.5 kg»); «Adjusting for covariates resulted in a mean difference of 0.68 kg (95% CI: 0.12-1.24) at 12-mo follow-up.» The arms differed at baseline (86.7% vs 68.0% women; BMI 35.1 vs 33.7). (Zheng et al., 2014) The other maintenance findings in Zheng are associations inside trials: daily weighing went with a lower risk of regaining 2.3 kg or more in STOP Regain (314) and weight entries with less regain in Funk (348, one arm), and less frequent weighing with gain in Sherwood 2006 (1801); Keep It Off (419) reports only that guided participants were more likely to weigh daily (OR 1.97, 1.31 to 2.99), with no weight result in the row. (Zheng et al., 2014) (inferred from Zheng et al., 2014) So maintenance moves from no held trial estimate to one: about 0.7 kg less regain over a year, against a control-arm regain of about 1.8 kg, from a quasi-randomised trial with unequal arms, reported through a review that did not appraise it. That is a small, imprecise estimate from a quasi-randomised trial; the primary trial and a pooled estimate are not held. — the same open token as on Weight-Loss Maintenance and Metabolic Adaptation.
Preventing weight gain — two small isolated tests and a null programme trial
Zheng’s five gain-prevention studies are a different stratum from the rest of this page: the aim is to avoid gain rather than to lose, and none is an obesity-treatment sample (mean BMI 26.7 to 29.6 where given; Levitsky’s first-year college women weighed about 62 kg; 31% of VanWormer 2012’s worksite staff had a BMI under 25), though the abstract describes the review’s participants as «adults seeking weight loss treatment». Levitsky 2006 ran two one-semester studies in first-year college women (mean weight about 62 kg; 32 and 41 people; allocation not stated; retention 68.8% vs 93.8% in the first study and 94.1% vs 66.7% in the second, intervention vs control) and is the other study that isolated the scale: daily weighing with e-mailed weights gained 0.1 ± 0.99 vs 3.1 ± 0.51 kg (P=0.001), and in the second study lost 0.82 ± 0.56 vs gained 2.0 ± 0.65 kg (P<0.01). French 2011 randomised households (90 participants) to a programme with daily weighing; at follow-up 71% vs 36% weighed weekly or more, with no significant BMI difference. Jeffery 1999 (809, 3 years) and VanWormer 2012 (1747 at six worksites, 2 years) are associations: daily weighers in Jeffery lost weight while every other category gained. (Zheng et al., 2014) (inferred from Zheng et al., 2014) The direction favours weighing in two very small studies whose allocation is not described and whose arms lost different shares of participants, and the one household-randomised trial was null on BMI. That is insufficient evidence, not no effect, and it does not carry to the weight-loss comparisons above.
The scale monitors the outcome, not a behaviour
Madigan separates self-weighing from the self-monitoring that behaviour-change reviews found effective on its own: «Additionally self-weighing is different to the behaviours investigated by Michie and colleagues, as self- weighing is monitoring the outcome rather than the be- haviour.» Their proposed mechanism is that people must reflect on the reading and then change eating and activity, which a scale alone may not prompt. (Madigan et al., 2015) (inferred from Madigan et al., 2015) This is why a scale is not the same tool as an activity tracker, which monitors the behaviour itself -> Wearable Activity Trackers and Physical Activity.
Open question — psychological harm
Whether regular weighing causes psychological harm (body dissatisfaction, low mood, disordered eating) is not settled by this review. Most included trials did not report adverse effects; the three that measured psychological outcomes reported no between-group differences, and the authors’ own caveat is that «few trials assessed whether self-weighing led to psychological problems». Two of those three trials added non-randomised analyses that also found none: Steinberg found no difference in body dissatisfaction or depressive symptoms between intervention participants instructed to weigh daily who did and did not lose weight, and «Gokee LaRose and colleagues found no relation- ship between change of frequency of self-weighing and disordered eating [41].» (Madigan et al., 2015) Outside psychology, «Three trials reported there were no serious adverse events related to the intervention in either the self- weighing or control group [21, 26, 32]. One trial de- tected five serious adverse events possibly related to the intervention but not specifically self-weighing.» (Madigan et al., 2015) No excess detected in three small trials (two with non-randomised follow-up analyses) is not shown safe; the harm side needs its own review and is held open here.
Updated 2026-10-29 — a review with psychological outcomes as one of its two aims (Zheng 2014). «Five studies reported an association between self-weighing and psy- chological outcomes. Investigators found no evidence that there were adverse effects to frequent self-weighing (27,37,44,49,53).» Two papers asked participants to rate daily weighing on an 8-point satisfaction scale and found it was perceived positively: Steinberg 2013 (the same trial as ref 53) and Gokee-LaRose 2010. (Zheng et al., 2014)
| Study (n) | Population | Design of the psychological result | Direction as printed |
|---|---|---|---|
| Welsh 2009 (73) | adults in a phone-based weight-loss trial, 89% women | within-study association | change in weighing not associated with body satisfaction; non-significant for eating disorder |
| Wing 2007, STOP Regain (314) | adults maintaining a loss, about 81% women | within-study association | more weighing with more restraint, less disinhibition, fewer depressive symptoms; daily weighers less likely to report 4 or more binge episodes a month |
| Gokee-LaRose 2009 (40) | young adults, mean age 29, 87.5% women | 2-group trial [ref 44 is titled a randomized controlled trial]; the row does not say between- or within-group | no negative effects; positive changes in body image dissatisfaction, binge eating, depressive symptoms, eating disorder |
| Kong 2012 | postmenopausal women with overweight or obesity per the reference titles (the row prints 79% female) | no design readable (no psychological result in the row) | cited in the list; its table row gives no psychological result |
| Steinberg 2014 (91) | adults in a weight-loss RCT | between-group (group x time) | lower body dissatisfaction (P=0.007) and greater dietary restraint (P=0.001) in the daily-weighing arm |
(Zheng et al., 2014)
The review names constructs, not instruments
[searched: questionnaire / inventory / Beck / instrument — no psychological instrument named across the 1-chunk denominator].
Parameter table — the two reviews on harm.
| Parameter | Madigan 2015 | Zheng 2014 | Same quantity? |
|---|---|---|---|
| Studies with a psychological outcome | 3 trials (Steinberg, Gokee-LaRose 2009, Madigan 2014) | 5 studies (Welsh, Wing 2007, Gokee-LaRose 2009, Kong, Steinberg) + satisfaction ratings in Gokee-LaRose 2010 and Steinberg 2013 (the same trial) | no — different inclusion rules (Madigan: weight-loss RCTs; Zheng: also maintenance and observational analyses inside trials); 2 studies shared, 6 distinct across both |
| Contrast | between groups, plus two non-randomised analyses | two within-study associations, one between-group, one unclear (Gokee-LaRose 2009), one with no psychological result (Kong) | no |
| Population | adults in weight-loss RCTs | treatment-seeking adults with overweight or obesity | yes, broadly |
| Instruments | «by questionnaire», constructs listed | constructs only | yes (neither names them) |
| Appraisal | Cochrane risk-of-bias tool for the trial set | none | no |
| Verdict | «There were no adverse effects of the trials reported, however few trials assessed whether self-weighing led to psychological problems. Those that did, found no evi- dence of negative consequences.» | «effective and safe» | no — both report no harm found; only Zheng goes on to safe (see below) |
(Madigan et al., 2015) (Zheng et al., 2014) (Madigan et al., 2015; inferred from Zheng et al., 2014) The counts differ because the inclusion rules differ, so three vs five is not a disagreement (not-joined check ii). Across both reviews about six distinct studies looked for harm, all small, all in adults with overweight or obesity trying to lose weight or keep it off (one of them, Madigan 2014, without a programme: «participants were asked about their mood and how they felt about their body at three months follow-up and there was no difference be- tween groups»); three are papers whose titles put the harm question (Welsh 2009, Wing 2007, Steinberg 2014), each a secondary analysis of a weight trial.
One shared trial is reported differently by the two reviews. Madigan says Steinberg’s group «found no sig- nificant differences» between groups across body dissatisfaction, anorectic cognitions, depressive symptoms, restraint, disinhibition, hunger and binge eating. Zheng’s row reports «Int group reporting lower body dissatisfaction and greater dietary restraint compared to controls» (P=0.007 and P=0.001), a difference where Madigan reports none. Neither reading shows more body dissatisfaction with daily weighing, but they cannot both be right about the between-group result; and whether greater dietary restraint is benign is itself part of the eating-disorder question. — the primary paper decides. Gokee-LaRose 2009 is not filed as a clash: Madigan reports improvement in both groups with no difference between them, and Zheng’s row reports «positive changes» without saying whether they are between- or within-group, so the two may describe different quantities. (Madigan et al., 2015) (Zheng et al., 2014)
Zheng’s safe goes further than its evidence. The authors: «It should be noted that these studies concerned only treatment-seeking overweight or obese adults (27,37,44,49,53), as opposed to studies that focused on detecting overweight or obese from normal weight individuals (62), or managing weight among adolescents (63). Thus, self-weighing can be an effective and safe self-monitoring strategy that should be implemented in weight- control programs without concern for negative psychological consequences.» The early harm study the introduction cites («daily self-weighing generated negative mood (e.g., depression, body dissatisfaction) (29)») was excluded at extraction because it used weighing only for recording weight, and the studies of normal-weight people and of adolescents and young adults (ref 63 is titled Self-weighing behaviors in young adults: tipping the scale toward unhealthy eating behaviors?) are contrasted with the included ones, not reviewed. (Zheng et al., 2014) (inferred from Zheng et al., 2014) The held evidence supports no excess harm detected in treatment-seeking adults with overweight or obesity, from about half between-group trial results; the rest are within-study associations that reverse causation could also produce, and one (Kong) whose row gives no psychological result (people who are losing weight may both weigh more and feel better). It does not support safe, and it does not cover the groups outside its included psychological studies: adolescents and normal-weight people (which Zheng names), and people with or at risk of an eating disorder (no included study is described as sampling them). Levitsky’s 18-year-old college women and VanWormer 2012’s worksites are in the review, but with no psychological outcome. Madigan also enrolled only adults and cites an adolescent study only as background for the concern. Harm in those groups is insufficient evidence, not absence of harm. — closes when a review that includes those samples is held.
Labelled default — what to do
(inferred from Madigan et al., 2015; Zheng et al., 2014) The action is cheap, reversible and low-cost, so it gets a plain default with its basis stated (CLAUDE.md, firmness scales with stakes; Acting on a Direction Without a Magnitude):
-
If you are in a weight-loss programme, weigh yourself regularly — daily or weekly, whichever you will keep doing. Basis: four RCTs adding self-weighing to a programme, each with a co-delivered part per the review’s Results text (an option to record diet and activity in three, self-regulation training with weight feedback in one; the Methods’ planned self-weighing-only subgroup is never reported): -1.7 kg (-2.6 to -0.8) at programme end within about 7 months, printed prediction interval -7.5 to 4.1 (flagged above as possibly misreported); no heterogeneity was detected across the four trials (I² 0%). The scale’s own share is not isolated.
-
Frequency is your choice. Basis: a between-trial comparison across programme trials (p = 0.57); in the add-on set weekly rests on one trial; no randomised daily-vs-weekly test is held. Zheng’s within-study associations favour more frequent weighing but compare people who chose their frequency; weighing tends to drop off in the first months, and the longer course is known only from point readings inside supported trials.
-
Someone seeing the numbers may help (a coach, a programme, a person). Basis: between-trial, two trials on one side; a hypothesis.
-
Without a programme, a scale on its own is untested for weight loss beyond one trial. Trying it costs little; pair it with a plan for eating and activity, since the authors’ proposed mechanism is that the reading must lead to a change in behaviour. Basis: one 3-month trial, -0.5 kg (-1.3 to 0.3), which the authors call imprecise enough to include a worthwhile effect: insufficient evidence, not proof of no effect; the pairing is the page’s inference from the mechanism, not a tested combination.
-
After a loss, keep weighing, at least weekly. Basis: one quasi-randomised maintenance trial reported through Zheng, whose package (the offer of free weighing scales, two phone calls, a record card) encouraged weekly weighing, so the estimate belongs to the package, not the scale alone (more often is supported only by associations), about 0.7 kg less regain over 12 months (0.68, 0.12 to 1.24) against a control regain of about 1.8 kg (how the outcome weight was measured is not stated), plus associations inside trials; small, and the primary trial is not held.
-
Stop if it distresses you. The harm question is open, and stopping costs little. Basis: about six small studies across both reviews, all in adults with overweight or obesity trying to lose or keep off weight (three between-group trial results, two within-study associations, and Kong, whose row gives no psychological result), detected no excess harm; one shared trial is reported differently by the two reviews. Neither review’s psychological evidence covers adolescents, people with or at risk of an eating disorder, or normal-weight people not trying to lose weight; both cite adolescent or normal-weight studies only as background (not held here), and no included study is described as sampling people with or at risk of an eating disorder, so for these groups the held evidence does not speak to harm.
-
Trying not to gain: weighing is cheap to try; the effect is not established. Basis: two very small studies in first-year college women favour daily weighing, but their allocation is not described and their arms kept different shares of participants; the one household-randomised trial (90 people) was null on BMI. Insufficient evidence, not no effect. The only psychological measure in the gain-prevention studies is acceptability: in Gokee-LaRose 2010 «Daily SW was perceived positively in both groups at 8 wk with no group differences.» Zheng reports no measure of body dissatisfaction, mood or disordered eating for them, so the harm evidence above does not reach this group; stop if it distresses you applies. (inferred from Zheng et al., 2014)
Where it ranks. A small refinement inside a programme, not a big rock. The closest estimate of adding it is -1.7 kg (-2.6 to -0.8) on a weight surrogate, at programme end within about 7 months, with co-delivered parts, a printed prediction interval crossing zero (its width flagged above as possibly misreported) and thinly evidenced durability -> Layer 1 - Ranking Interventions for a Stratum. For keeping weight off, the one held trial estimate, for a weekly-weighing package (the offer of free weighing scales, two calls, a record card), is about 0.7 kg less regain over a year; it measures a different outcome (regain, not loss), comparator and design than the -1.7 kg, so the two are not ranked against each other. (inferred from Madigan et al., 2015; Zheng et al., 2014)
Guidance context. At the time, «Currently, widely used behavioural programmes in the UK advise
their participants against weighing themselves.» That advice ran against the direction of this
review’s add-on finding; no current guidance position on self-weighing is held. (An earlier edit,
2026-11-06, struck that clause as superseded by the 2013 AHA/ACC/TOS guideline. Revised 2026-11-09:
the guideline, now held directly below, is a dated US position with its trial search ending in 2009,
not a current one, so the clause stands.)
(Madigan et al., 2015)
Zheng reports, secondhand, that the US position went the other way: «Self-weighing has also been
recommended as part of the AHA/ACC/TOS Guidelines for Managing Overweight and Obesity in Adults from
the National Heart, Lung, and Blood Institute (24).» The guideline itself (Jensen 2014) is not held, so
its wording and grading are unread. (superseded 2026-11-06, below) (Zheng et al., 2014)
(inferred from Madigan et al., 2015) the comparison of the advice with
the review’s own direction is the page’s.
What the 2013 AHA/ACC/TOS guideline actually says (read directly, 2026-11-06). Zheng’s relay (recommended as part of the guideline) is accurate. In the guideline, weighing is never its own evidence statement or its own graded recommendation; it appears in six places, each as one part of something larger. (Corrected 2026-11-09: the list first counted five and missed Box 8, below.)
- As a programme component (evidence statement, graded High). CQ4 ES1 describes what effective high-intensity programmes contain: «Behavior therapy: Comprehensive lifestyle interventions usually provide a structured behavior change program that includes regular self-monitoring of food intake, physical activity, and weight. These same behaviors are recommended to maintain lost weight, with the addition of frequent (ie, weekly or more often) monitoring of body weight.» The High grade attaches to the description of the whole package, not to weighing’s own contribution. (Jensen et al., 2014)
- As a component of remote programmes (evidence statement, Moderate). CQ4 ES3 describes effective electronically delivered programmes as ones «which include frequent self-monitoring of weight, food intake, and physical activity» with personalised feedback; again the grade is for the package. (Jensen et al., 2014)
- Inside the maintenance recommendation (Grade A, Strong). Rec 4g: «For weight loss maintenance, prescribe face-to-face or telephone-delivered weight loss maintenance programs that provide regular contact (monthly or more frequently) with a trained interventionist† who helps participants engage in high levels of physical activity (ie, 200–300 min/wk), monitor body weight regularly (ie, weekly or more frequently), and consume a reduced-calorie diet (needed to maintain lower body weight).» (Jensen et al., 2014)
- In the treatment algorithm, for people not trying to lose weight. For overweight adults without raised cardiovascular risk, and for people formerly overweight, the algorithm text (Box 7, avoid weight gain) says to «advise patients to frequently measure their own weight and to avoid weight gain by adjusting their food intake if they start to gain more than a few pounds.» No evidence question on preventing gain sits behind it. (Jensen et al., 2014)
- In the readiness assessment (Box 8), named as a programme component. «The clinician can ask, ‘How prepared are you to make changes in your diet, to be more physically active, and to use behavior change strategies such as recording your weight and food intake?’ These are the components of a comprehensive lifestyle intervention.» It is a question to ask before starting, not a graded statement. (Jensen et al., 2014)
- In the maintenance algorithm text (Box 15), framed as an association. «Strategies such as frequent self-weighing (at least weekly), consumption of a reduced-calorie diet, and high levels of physical activity (>200 min/wk) are associated with better weight maintenance over time.» This is the only place the guideline uses the term self-weighing, and it is expert algorithm text, not a graded statement. Its wording is associational, which fits the caveat Madigan raises about observational weighing data (motivation as a confounder, success prompting the weighing). (Jensen et al., 2014)
What follows for this page’s questions. (inferred from Jensen et al., 2014; Madigan et al., 2015; Zheng et al., 2014)
- For maintenance the guideline sets a floor, weekly or more, and does not choose between daily and weekly. During active loss it says «regular self-monitoring» (ES1) or «frequent self-monitoring of weight» (ES3), with no frequency; for gain prevention Box 7 says «frequently». So it neither supports nor contradicts this page’s daily-vs-weekly comparison; it never asked that question.
- Its grade cannot be read as an effect size for weighing. Weighing is bundled with a monthly-contact programme, 200-300 min/wk of activity and a reduced-calorie diet, the same co-delivery problem that limits the reviews above. The closest held numbers are phase-specific: -1.7 kg (-2.6 to -0.8) for adding self-weighing (with other self-regulation parts) to a programme during loss, and about 0.7 kg less regain over a year (0.68, 0.12 to 1.24) for the maintenance phase the Grade A recommendation covers. That second figure is small and imprecise: one quasi-randomised trial with unequal arms, held secondhand through a review that did not appraise it (above).
- A US guideline vs UK programme practice, at the time. The US guideline wrote weekly-or-more weighing into a Grade A maintenance recommendation while, per Madigan, UK behavioural programmes advised against it. This page does not judge which was right.
The guideline’s maintenance evidence and programme-intensity statements are on Lifestyle Weight-Loss Programme Intensity and Delivery.
Limits
- Weight is a surrogate. Weight is the only efficacy outcome pooled; three trials measured psychological outcomes and none pooled a patient-important outcome. (Madigan et al., 2015)
- Short. Programme-end figures; the add-on trials ran 10 weeks to 7.2 months. (Madigan et al., 2015)
- Mostly US trials, mostly women (15 of 24 in the USA; four women-only trials, median 75% women in the rest; 11 internet-based or mixed). (Madigan et al., 2015)
- No certainty grading and no publication-bias assessment; trial counts per comparison are small (1, 4, 15). (Madigan et al., 2015)
Single source. The page rests on one review; a second review of the same lever is the next source, not yet held.Superseded 2026-10-29 (Zheng 2014): two reviews now, sharing five trials, and the second has no quality appraisal and no reported pooled value, so its weight-loss findings add associations, not a second effect estimate. (inferred from Madigan et al., 2015; Zheng et al., 2014)- Exposure self-reported. In Zheng, 14 of 17 studies measured weighing frequency by self-report. (Zheng et al., 2014)
- One stratum. Mostly women in both reviews (Madigan: above); Zheng’s samples were also predominantly White and middle-aged, which its authors call the main limitation. (Zheng et al., 2014)