Sustaining regular exercise during weight loss maintenance: The role of consistent exercise timing

Abstract

Background:

Exercising at a consistent versus variable time of day cross-sectionally relates to greater moderate-to-vigorous physical activity (MVPA) among weight loss maintainers. This study evaluated the relationships between exercise timing and both MVPA levels and habit strength, as well as stability in exercise timing, over one year among maintainers in the National Weight Control Registry.

Methods:

Participants (n=709) completed questionnaires assessing exercise timing, MVPA, and exercise automaticity (a measure of habit) at baseline and one-year follow-up. At each assessment, participants were labeled temporally consistent exercisers if >50% of their exercise sessions/week occurred in one time window: early morning, late morning, afternoon, or evening. Participants exercising consistently during the same window at both assessments were labeled as having stable patterns.

Results:

Temporally consistent exercise at baseline, regardless of its specific time, related to greater MVPA over time (p’s<.05). Approximately half of temporally consistent exercisers at baseline exhibited stable patterns. Early morning exercise and greater exercise automaticity at baseline predicted stable patterns (p’s<.005). Temporally consistent exercise, especially during the early morning, related to greater automaticity across time (p’s<.01).

Conclusions:

Consistent exercise timing may help maintainers accrue more MVPA. Consistent early morning exercise was most strongly related to exercise automaticity and routine stability.

Regular exercise to achieve recommended levels of moderate-to-vigorous physical activity (MVPA)¹ reduces risk and severity of numerous diseases and can improve aspects of daily functioning like energy and mood.^2–4 Among those with overweight or obesity who have intentionally lost weight, high levels of MVPA also appear critical for protecting against weight regain.^5,6 However, many individuals attempting weight loss do not increase their MVPA to recommended levels, and MVPA levels tend to decline over time following intentional weight loss.^7,8 Because weight regain can negate or attenuate the health benefits of weight loss, it is imperative to identify factors that can be leveraged to facilitate regular MVPA long-term.

The National Weight Control Registry (NWCR) is a registry of more than 10,000 adults who have maintained a weight loss of ≥30 lbs. for ≥1 year. Previous research shows that NWCR members on average engage in high levels of MVPA (e.g., 42 min/day).^9,10 Thus, the NWCR can serve as an important resource for learning about factors that may be targeted to promote and sustain high levels of MVPA among individuals attempting weight loss maintenance.

One factor that may promote regular MVPA is consistency in the time of day that exercise is performed (“temporal consistency”). Temporal consistency may support high MVPA levels for several reasons, including simplified planning (i.e., by reserving a standing time day-to-day for MVPA) and strengthened habit formation (i.e., by creating learned associations between a certain time of the day and MVPA, such that the time becomes a cue that impulsively triggers MVPA with minimal cognitive effort or intention).^11–13 In partial support of the hypothesized benefits of temporal consistency, one prior study among NWCR members found that temporally consistent exercisers reported greater MVPA than temporally variable exercisers.¹⁴ However, this study was limited by a cross-sectional design and it is thus not clear whether temporally consistent exercisers continue to exhibit greater MVPA over time. Previous research also suggests that morning exercise, in particular, may confer benefits for weight loss independent of total MVPA levels.^15,16 Exercising in the morning could also provide certain advantages when trying to establish a consistent exercise routine (e.g., fewer scheduling conflicts to interfere with regular MVPA).^11,17,18 However, few studies have examined exercise timing’s association with processes that may be important for building a consistent exercise routine, such as increased exercise automaticity.

Thus, among a sample of physically active NWCR members who completed two consecutive yearly surveys (first survey = “baseline,” second survey = “one-year follow-up”), this study aimed to examine: whether temporally consistent exercisers, whom were found to engage in more MVPA than temporally variable exercisers in previous cross-sectional work, remained more active at follow-up; how often baseline temporally consistent exercisers continued to exhibit the same exercise timing patterns one year later; and the relationship between temporal consistency and exercise automaticity (an indicator of habit strength).¹⁹ The latter aim evaluated two questions: (1) Does temporal exercise consistency relate to greater exercise automaticity at baseline and continued greater automaticity at follow-up? and (2) Among those with temporally consistent exercise, does greater automaticity at baseline predict greater stability in exercise routines over time? In addressing these questions, we focused on whether temporally consistent morning exercisers had the greatest MVPA levels, the greatest stability in their routines, and greater exercise automaticity over time.

Our overarching hypotheses were that temporally consistent exercise would relate to higher MVPA levels, greater stability in exercise routines, and greater exercise automaticity. We also hypothesized that these relationships would be strongest among temporally consistent exercisers who performed exercise in the early morning compared to other times of the day.

Methods

Participants.

Participants were volunteers enrolled in the NWCR. Eligibility criteria for inclusion in the NWCR include age ≥18 years and ≥30 lbs. (13.6 kg) weight loss maintained for ≥1 year. Participants are asked to provide documentation of their weight loss at the time of enrollment (e.g., medical records, photographs). Additional eligibility criteria for the present study included reporting ≥2 days/week of MVPA on a validated physical activity questionnaire²⁵ and providing valid responses on the exercise timing measure at baseline (both required to characterize initial exercise timing patterns), and completion of questionnaires at follow-up.

Procedure.

Individuals are recruited for the NWCR on an ongoing basis through the NWCR website. Participation in the NWCR is voluntary and no compensation is provided. Individuals who are admitted to the NWCR are sent an initial questionnaire and annual follow-up questionnaires online. Participants in the present study were those who: (1) met the aforementioned eligibility criteria, and (b) had an opportunity to and completed two consecutive questionnaires approximately one year apart between July 2018, when the relevant measures (described below) were added to the NWCR battery, and July 2020, when data were downloaded for analysis. Procedures were approved by The Miriam Hospital’s Institutional Review Board. Participants provided informed consent before enrolling in the NWCR.

Measures.

Exercise timing.

Exercise timing was assessed with a questionnaire designed by our research team. Participants read the statement “When I exercise, it tends to be at the same time of day,” and responded with “True,” “False,” or “I don’t exercise” [ineligible]. Participants who responded “True” then provided the start time of their longest exercise session, if any, for each day within a typical week. Participants were instructed to report only on exercise sessions that were ≥10 minutes in duration, made them breathe harder than normal, and were meant to improve health (i.e., exercise rather than incidental physical activity). Participants’ exercise timing patterns were classified with a previously used framework,¹⁴ which is similar to other categorical approaches for characterizing exercise timing.^15,17,20 Participants were classified as temporally variable exercisers if they stated “false” in response to the above statement (“When I exercise, it tend to be at the same time of day”), or if they stated “true” but then reported an exercise timing pattern that did not exhibit a predominant time of day of exercise, as described below. Participants were classified as temporally consistent exercisers if >50% of their exercise sessions across the week (2/2, 2/3, 3/4, etc.) started within a single time window, as follows: early morning: 4:00–8:59 am; late morning: 9:00–11:59 am; afternoon: 12:00–4:59 pm; evening: 5:00 pm-3:49 am. Participants’ exercise timing patterns were classified using this framework at both assessments. Baseline temporally consistent exercisers who continued to exercise consistently at the same time (e.g., afternoon-afternoon) at follow-up were coded as remaining stable in timing. To be considered stable in timing, participants were required to be exercising consistently at the same time of day at follow-up rather than to simply be classified as a temporally consistent exerciser to avoid classifying shifting routines (e.g., afternoon-early morning) as stable.

MVPA Levels.

The Paffenbarger Physical Activity Questionnaire²¹ assessed MVPA at both assessments. Weekly MVPA minutes was determined by summing the number of minutes spent walking in bouts of ≥10 minutes for the purpose of exercise and the number of minutes spent engaging in other fitness, sport, or recreational activities of a moderate-to-vigorous intensity each week.²¹ MVPA days/week were determined by asking participants, “In general, at least once per week, do you engage in regular activity similar to brisk walking, jogging, bicycling, etc., long enough to work up a sweat, get your heart thumping, or get out of breath?” Participants who responded “yes” then reported the number of days/week they engage in activity of this intensity. “No” responses were coded as zero days (note: ≥2 days at baseline required for inclusion).

Exercise Automaticity.

A four-item version of the Self-Report Behavioral Automaticity Index¹⁹ assessed exercise automaticity at both assessments. As in prior work,¹⁴ participants were presented the stem, “Engaging in regular exercise (i.e., being physically active for the purpose of improving your health for at least 10 consecutive minutes) is something…” and then responded to four items assessing exercise automaticity (e.g., “…I do automatically). Response options were 1 (never) to 4 (always), and responses were averaged. Cronbach α=0.88 and baseline and 0.86 at follow-up, indicating good internal consistency.

Sociodemographic Characteristics, BMI, and Weight History.

A questionnaire assessed participants’ current height and weight (for BMI), birth date (for age), gender, racial and ethnic identities, and employment status (“Are you currently working for pay full or part time?”: yes/no). Participants also reported their highest and lowest weight (to determine maximum weight loss), and month/year since first losing and maintaining ≥30 lbs. (used to determine weight loss maintenance duration at NWCR enrollment).

Statistical analysis.

Data were analyzed in IBM’s SPSS version 25 in 2020–2021. Sociodemographic characteristics and exercise timing patterns were characterized with descriptive statistics (e.g., frequencies, means). Potential differences in sociodemographic characteristics between exercise timing groups (consistent vs. variable or early morning, late morning, afternoon, or evening—all as classified at baseline) were examined with Chi-square or ANOVA tests. Tests of significance were two-tailed with α=.05; a Bonferroni correction was used for post-hoc tests in models where differences were observed by specific exercise time among temporally consistent exercisers. Differences in MVPA days/week and exercise automaticity between exercise timing groups from baseline to follow-up were evaluated using repeated-measures ANOVA/ANCOVA. Between-person effects pertaining to differences in MVPA or automaticity across time for the exercise timing groups were of primary interest. If significant between-person effects were observed (indicating mean differences between timing groups across time), simple main effects were conducted to compare differences between timing groups by assessment.

Due to MVPA min/wk having a highly positively skewed distribution, Kruskal-Wallis tests or Quade’s rank ANCOVA (if controlling for relevant sociodemographic factors) were used to assess differences in MVPA min/wk between timing groups separately for baseline and follow-up timepoints. The independent variable was exercise timing classification at baseline (consistent vs. variable or early morning, late morning, afternoon, or evening) and the dependent variable was MVPA min/wk at baseline or follow-up.

The number of baseline temporally consistent exercisers who exhibited stable exercise timing patterns at follow-up was characterized with frequencies and percentages. Potential differences in stability by the specific time of exercise were evaluated with a Chi-square test of independence that assessed the relationship between the specific time of temporally consistent exercise at baseline and continued exercise timing consistency (yes/no) at follow-up. Binary logistic regression was used to examine baseline automaticity as a predictor of stable timing patterns at follow-up among baseline temporally consistent exercisers.

Due to inclusion criteria requiring participants to provide data on follow-up questionnaires, no participants were missing data on the Paffenbarger or exercise timing questionnaire at follow-up. However, due to an error with saving data from online surveys, approximately 70 participants were missing sociodemographic or BMI data and approximately 100 participants were missing weight history data (see Table 1 note). Eighteen participants were also missing exercise automaticity data at follow-up and were excluded from those analyses.

Table 1.

Participant characteristics and MVPA levels for the full sample and exercise consistency groups

	Full sample (n = 709)	Comparisons by temporal exercise consistency status
		Consistent (n = 479)	Variable (n = 230)	F or X2	p
Participant Characteristics
Age, M (SD)	53.5 (13.9)	53.0 (13.6)	54.4 (14.5)	1.31	.254
Baseline BMI, M (SD)	26.5 (4.6)	26.5 (4.6)	26.5 (4.4)	0.009	.926
Gender, n (%)				1.61	.204
Female	410 (64.4%)	270 (65.9%)	161 (70.9%)
Male	227 (35.6%)	140 (34.1%)	66 (29.1%)
Race, n (%)				0.37	.541
White	599 (94.0%)	407 (94.4%)	192 (93.2%)
African American/ Black	17 (2.7%)	10 (2.3%)	7 (3.4%)
Asian American	6 (0.9%)	5 (1.2%)	1 (0.5%)
Other	15 (2.4%)	9 (2.1%)	6 (2.9%)
Ethnicity, n (%)				0.08	.785
Non-Hispanic	623 (97.8%)	422 (97.9%)	201 (97.6%)
Hispanic	14 (2.2%)	9 (2.1%)	5 (2.4%)
Employment Status, n (%)				0.69	.406
Employment	486 (76.3%)	333 (77.3%)	153 (74.3%)
Not employed	151 (23.7%)	98 (22.7%)	53 (25.7%)
Maximum Weight Loss (kg; M[SD])	38.2 (16.9)	37.9 (16.1)	38.9 (18.7)	0.40	.530
Time Weight Loss Maintained (years; M[SD])	4.3 (5.7)	4.1 (5.3)	4.7 (6.4)	1.71	.191
MVPA Levels
Baseline MVPA levels
Min/wk, mdn (Q1, Q3)	330.0 (195.0, 530.0)	345.0 (200.0, 570.0)	300.0 (178.8, 482.5)	6.12	.013
Days/wk, M (SD)	4.7 (1.6)	4.8 (1.6)	4.4 (1.6)	7.39	.007
One-year follow-up MVPA levels
Min/wk, mdn (Q1, Q3)	315.0 (165.0, 504.5)	325.00 (180.0, 545.0)	285.0 (150.0, 450.0)	4.41	.036
Days/wk, M (SD)	4.9 (2.1)	4.3 (2.1)	3.9 (2.2)	5.04	.025

Note. Percentages are for valid data. For temporal exercise consistency comparisons, race was dichotomized as White vs. non-White. The following number of participants were missing data: age – n=73; BMI = 4; gender, race, ethnicity, and work status – n = 72; maximum weight loss – n =103. Abbreviations: BMI = body mass index, MVPA = moderate-to-vigorous physical activity.

As described in Measures, temporal exercise consistency was defined using a threshold of >50% of weekly exercise sessions occurring during a single time window. As a sensitivity analysis, we also conducted analyses using a >70% threshold. We observed the same pattern of results. To remain consistent with prior research,^14,15,20 we report results using the >50% threshold. Given that surveys were sent to participants across the year, we also examined the proportion of surveys completed per season (Winter: Dec-Feb, Spring: March-May, Summer: June-Aug, Fall: Sept-Nov) and whether temporal consistency/inconsistency status or the specific time of day of temporally consistent exercise differed by season. Although more baseline (35.0%) and follow-up (34.3%) surveys were completed in the winter relative to other seasons (X²’s > 37.0; p’s<.001; Spring: 21.2% [baseline] and 25.1% [follow-up]; Summer: 22.5% [baseline] and 20.6% [follow-up]; and Fall: 21.3% [baseline] and 19.9% [follow-up]), the proportion of individuals classified as temporally consistent vs. inconsistent exercisers did not differ based on season, nor did the proportion of individuals exercising consistently at different times of the day (p’s>.39).

Results

Sample Characteristics.

Figure 1 displays participant flow. As shown, a total of 709 participants met eligibility criteria. Analyses comparing the 709 participants retained for analyses to the 694 individuals excluded showed that retained participants had greater average weight loss duration (4.3 vs. 3.3 years; p=.009) and that a greater proportion of the retained vs. excluded sample was male (35.6% vs. 27.7%; p=.003); no differences were observed for age, race, ethnicity, gender, employment status, or maximum weight loss (p’s>.05). Among the 1,049 individuals who endorsed adequate MVPA at baseline and were considered for inclusion, the 709 participants retained for analyses reported more MVPA min/week (median=300.0 min/wk) than the 340 who were excluded (median=277.5 min/wk; the majority of whom were excluded for not completing a follow-up survey; p=.005); differences were not observed for MVPA days/wk or exercise automaticity (p’s>.05). Of the 709 included participants, 255 (36%) contributed data to the previous cross-sectional PA timing analyses published by our group that assessed relations between PA timing and MVPA levels at only one timepoint¹⁴; hence, a majority of participant data for cross-sectional analyses was unique, as was all prospective data. The mean time between survey completion was 364.1 days (SD=26.5 days).

Figure 1.

Outline of flow to determine study sample.

Table 1 shows that most participants identified as female, White, non-Hispanic, and were working full- or part-time. Participants were on average middle-aged and had a BMI of approximately 26 kg/m². Participants had, on average, lost 38.2 (SD=16.9) kg and maintained weight loss for 4.3 (SD=5.7) years.

Differences Based on Temporally Consistent vs. Temporally Variable Exercise

Approximately two-thirds (67.6%) of participants were temporally consistent exercisers at baseline, while 32.4% were temporally variable exercisers. Temporally consistent and variable exercisers did not differ on sociodemographic characteristics or BMI. A repeated-measures ANOVA showed that temporally consistent exercisers reported more MVPA days/wk on average across time than temporally variable exercisers (average difference of 0.3 days/wk; between-person effect: F=7.928, p=.005) and that MVPA days/wk on average decreased from baseline to follow-up and to a similar magnitude across groups (average decrease of 0.5 days/wk; within-person effect: F=37.882, p<.001: interaction: F=0.062, p=.803). Table 1 displays means, standard deviations, and simple main effects for MVPA days/wk for each timing group by assessment point. Table 1 also displays results from Kruskal-Wallis analyses showing that, similar to findings for MVPA days, baseline temporally consistent exercisers reported more MVPA min/wk at baseline and sustained greater min/wk at follow-up. Approximately half (n=240, 50.1%) of baseline temporally consistent exercisers continued to exercise consistently at the same time at follow-up (i.e., had a stable routine).

A repeated-measures ANOVA showed that temporally consistent exercisers reported greater exercise automaticity scores across time (average difference of 0.32 in scores; between-subject effect: F=27.593, p<.003), and that, on average, there was non-significant change in scores from baseline to follow-up (average change of 0.001; main effect of time: F=.001, p=.960; interaction: F=1.80, p=.180). Table 2 displays means, standard deviations, and simple main effects for automaticity scores for each timing group by assessment point. Additionally, among baseline temporally consistent exercisers, baseline automaticity positively predicted the likelihood of remaining stable in exercise timing at follow-up, B=0.413, SE=0.114, Wald X²=13.011, p<.001, OR=1.511, 95% CI [1.207, 1.891].

Table 2.

Exercise automaticity based on exercise timing patterns at baseline

Comparison	Timepoint when automaticity measured	Comparison groups	M / EMM	SD / SE	F	p
Temporal consistency vs. variability	Baseline	Consistent	2.85	.83	27.50	<.001
		Variable	2.50	.85
	Follow-up	Consistent	2.81	.82	16.98	<.001
		Variable	2.53	.82
Specific time of day of temporally consistent exercisea	Baseline	Early morning	2.99	.06	3.17	.024
		Late morning	2.88	.10
		Afternoon	2.75	.12
		Evening	2.71	.07
	Follow-up	Early morning	2.96	.06	4.03	.008
		Late morning	2.82	.10
		Afternoon	2.64	.12
		Evening	2.67	.07

Note. In all models, participants were categorized into comparison groups based on exercise timing at baseline. Means and standard deviations are presented for models evaluating differences between temporally consistent and temporally variable exercisers. Estimated means and standard errors are presented for models evaluating differences by the specific time of day of temporally consistent exercise, as models controlled for age and employment status.

Differences Based on Specific Time of Exercise among Temporally Consistent Exercisers

Table 3 shows that the most common time of day of exercise among temporally consistent exercisers at baseline was early morning (42.8%), followed by evening (29.4%). Exercise timing groups differed on age, with evening exercisers being younger than all other groups and early morning exercisers being younger than late morning exercisers. Specific time of exercise related to employment status, with fewer late morning exercisers being employed. A two-way ANCOVA controlling for age and employment status showed that the exercise timing groups did not differ on MVPA days/wk on average over time (mean differences of <0.5 days/wk between groups; between-person effect: F=2.278, p=.079) and that all exercise timing groups on average demonstrated decreases in MVPA days/wk that were of a similar magnitude (average decrease of 0.5 days/wk; within-person effect: F=4.072, p=.044; interaction: F=0.816, p=.486). Table 3 displays estimated MVPA days/wk for each group by assessment controlling for age and employment status. As shown in Table 3, no differences in MVPA min/wk were observed. The proportion of baseline temporally consistent exercisers who exhibited stable timing patterns at follow-up differed based on exercise time, X²(3, N=449)=23.518, p<.001, with more early morning exercisers (n=126, 61.5%; p<.001) and fewer afternoon exercisers (n=15, 28.8%; p=.001) remaining consistent compared to other timing groups (late morning: n=38 [46.9%] and evening: n=60 [42.6%]).

Table 3.

Participant characteristics and MVPA levels by specific time of day of temporally consistent exercise

	Early morning (n = 205)	Late morning (n = 81)	Afternoon (n = 52)	Evening (n = 141)	F or X2	p
Participant Characteristics
Age, M (SD)	53.6 (12.8)	59.1 (14.3)	55.2 (13.5)	48.5 (13.0)	10.50	<.001
Baseline BMI, M (SD)	26.1 (4.0)	26.3 (4.4)	26.9 (4.9)	27.0 (5.5)	1.27	.284
Gender, n (%)					0.59	.899
Female	115 (62.2%)	43 (64.2%)	28 (58.3%)	84 (64.1%)
Male	70 (37.8%)	24 (35.8%)	20 (41.7%)	47 (35.9%)
Race, n (%)					2.15	.543
White	176 (95.1%)	65 (97.0%)	44 (91.7%)	122 (93.1%)
African American/Black	3 (1.6%)	1 (1.5%)	2 (4.2%)	4 (3.1%)
Asian American	2 (1.1%)	1 (1.5%)	0 (0%)	2 (1.5%)
Other	4 (2.2%)	0 (0%)	2 (4.2%)	3 (2.3%)
Ethnicity, n (%)					6.96	.073
Non-Hispanic	180 (97.3%)	67 (100.0%)	45 (93.8%)	130 (99.2%)
Hispanic	5 (2.7%)	0 (0%)	3 (6.3%)	1 (0.8%)
Employment Status, n (%)					25.82	<.001
Employed	151 (81.6%)	37 (55.2%)	34 (70.8%)	111 (84.7%)
Not employed	34 (18.4%)	30 (44.8%)	14 (29.2%)	20 (15.3%)
Maximum Weight Loss (kg; M[SD])	39.2 (16.9)	38.8 (19.2)	38.5 (12.6)	35.6 (13.9)	1.29	.276
Time Weight Loss Maintained (years; M[SD])	3.6 (4.0)	3.9 (5.4)	5.3 (8.5)	4.4 (5.5)	1.25	.292
MVPA Levels
Baseline MVPA levels
Min/wk, mdn (Q1, Q3)	355.0 (237.5, 600.0)	325.0 (177.5, 605.0)	320.0 (141.3, 476.3)	330.0 (177.5, 540.0)	2.04	.108
Days/wk, EMM (SE)	5.0 (.12)	5.0 (.20)	4.6 (.23)	4.5 (.14)	--	--
Follow-up MVPA levels
Min/wk, mdn (Q1, Q3)	340.0 (200.0, 562.5)	350.0 (170.0, 630.0)	300.0 (165.0, 431.3)	285.0 (150.0, 505.0)	2.25	.082
Days/wk, EMM (SE)	4.5 (.16)	4.3 (.27)	4.3 (.31)	4.0 (.19)	--	--

Note. Quade’s ANCOVA was used to assess differences in MVPA min/wk by timing group when controlling for age and employment status; however, raw medians and quartile values are reported for interpretability. Post-hoc tests: age -- early morning < late morning, p=.022; evening < other groups, p’s<.05; employment -- fewer late morning exercisers were employed, p<.001, Abbreviations: BMI = body mass index, MVPA = moderate-to-vigorous physical activity.

A repeated-measures ANCOVA controlling for age and employment status showed that, among baseline temporally consistent exercisers, the specific time of exercise related to automaticity scores across time (between-person effect: F=4.445, p=.004), with early morning exercisers reporting greater automaticity scores (about 0.3 points higher) than evening exercisers (p=.005; other comparisons: p<.13; see estimated means, standard errors, and simple main effects in Table 2). Automaticity scores also on average decreased across time and to a similar extent across groups (average decrease of 0.06 points; within-person effect: F=4.343, p=.038; interaction: F=0.148, p=.931).

Discussion

Regular MVPA is important for weight loss maintenance and overall health yet can be challenging to achieve. This study evaluated the potential importance of exercise timing—both temporal consistency in exercise performance and the specific time of day that exercise is consistently performed—to MVPA levels, exercise automaticity, and stability in exercise routines over a one-year period among physically active successful weight loss maintainers. There were three main findings. First, extending prior research,¹⁴ we found that temporally consistent exercisers engaged in greater MVPA at baseline and continued to exhibit greater MVPA at follow-up, while the specific time of day of temporally consistent exercise was unrelated to MVPA levels. Although we cannot draw causal conclusions from our data and experimental testing is needed to determine whether temporal consistency can be leveraged to increase MVPA, it may be that having a consistent exercise time—regardless of when it is—helps individuals to protect time for MVPA and build an exercise habit.²² Partial support for this latter possibility comes from our analysis of the relationship between exercise timing and exercise automaticity, as discussed below. Alternatively, it may be that more active individuals happen to demonstrate greater temporal consistency, or that additional variables (e.g., motivation) partially underlie the relationship between temporal consistency and MVPA levels. Randomized trials are needed to clarify the nature of this relationship. Nonetheless, the present findings highlight temporal consistency as a novel correlate of MVPA and suggest encouraging individuals to exercise at a similar time day-to-day could assist with increasing MVPA.

A second major finding of this study is that exercise timing patterns were quite variable over time, although this differed based on the specific time of day of exercise and degree of exercise automaticity. Only half of temporally consistent exercisers at baseline continued to exercise consistently at the same time of day at follow-up. This finding aligns with prior research suggesting that exercise behavior in general fluctuates over time,²³ although this was somewhat surprising in the present sample, given that NWCR members on average report high levels of consistency across a range of weight control behaviors.^24,25 However, as hypothesized and as predicted from a habit formation perspective, temporally consistent exercisers were more likely to continue exercising at the same time at follow-up (i.e., demonstrate a stable routine) if they reported greater exercise automaticity at baseline. Additionally, temporally consistent exercisers reported greater exercise automaticity at baseline and continued to report higher levels of automaticity at follow-up. Thus, it appears that temporal consistency may both help to establish and be sustained by greater exercise habit strength, consistent with prior research.^12,13,26 While future studies are needed in which individuals are randomized to different exercise timing prescriptions and habit strength is measured to explicate these relationships, these preliminary findings suggest that encouraging temporal consistency in exercise could also help to promote habit formation and establish more stable exercise routines. This has implications for health, as self-reported habit has consistently been shown to relate to exercise behavior, even when accounting for factors like motivation, and habit is believed to be key for behavioral maintenance.^27,28 Thus, factors like temporal consistency that promote exercise habit formation should help to facilitate continued engagement in regular MVPA over the long-term, positioning individuals to continue experiencing the health benefits of MVPA.

The third major finding was that, as hypothesized, temporally consistent exercise in the early morning related to both greater stability in exercise timing and greater automaticity. Approximately 62% of early morning exercisers at baseline continued to exercise in the early morning at follow-up, and early morning exercisers reported greater exercise automaticity than evening exercisers at baseline and continued to report greater automaticity at follow-up. However, our hypothesis that consistent early morning MVPA would relate to the greatest amount of MVPA was not supported. Thus, while consistent early morning exercise, in particular, does not appear to relate to greater MVPA compared to temporally consistent exercise performed at other times of the day, early morning exercise relates to more stable exercise routines and a stronger exercise habit. Consistent performance of greater MVPA over time is associated with better weight loss maintenance in the NWCR²⁹ and more robust health benefits of MVPA,³⁰ and research suggests a stronger exercise habit should help to promote maintenance of exercise behavior over time by reducing individuals’ reliance on goals and effortful decision-making to continue engaging in MVPA.^27,31 Limited prior research has sought to understand how NWCR members engage in high amounts of MVPA long-term. Our results suggest that consistently exercising in the early morning—which was the most common time of temporally consistent exercise—may be one strategy NWCR members use to build a strong exercise habit and sustain MVPA levels. Accordingly, important questions for future research are whether encouraging individuals to exercise in the early morning can assist insufficiently active individuals in developing stable exercise patterns and habit, and whether stronger habits and stable patterns predict long-term maintenance of MVPA.

Given that there were differences in age and employment status by time of day of exercise, future studies should examine how lifestyle and sociodemographic factors impact feasible times for temporally consistent exercise and the potential effects of prescribed exercise timing for MVPA and other health outcomes. Regarding the feasibility of exercising at a particular time, a few prior studies have examined the effects of exercise in the morning vs. other times of the day on a variety of outcomes (e.g., weight change) using in-laboratory, supervised exercise protocols.^16,32,33 While these studies are limited by small samples sizes and low external validity due to the tightly controlled nature of the exercise settings, high adherence rates to prescribed morning exercise have been observed (e.g., exercise session adherence rates of 94% in the morning vs. 87% for the evening).³² Additional research with larger and more diverse samples is needed. Regarding potential moderators of the effects of exercising at a particular time of day, one recent study that examined cardiorespiratory fitness and cardiovascular disease risk in relation to exercise timing among adults with type II diabetes observed differences in findings based on sex.²⁰ Specifically, men who exercised consistently in the midday had the lowest cardiorespiratory fitness, while women who had variable exercise timing had the lowest fitness. In contrast, for cardiovascular disease risk, men who exercised consistently in the morning had the highest risk, while exercise timing was unrelated to risk among women. These findings highlight the importance of considering individual differences when assessing exercise timing’s relation to outcomes of interest. Additionally, particularly as other work has suggested that consistent morning MVPA may relate to greater weight loss among those with overweight or obesity independent of overall MVPA levels,^15,16 it will be important for future studies to consider the comprehensive, multifactorial effects of exercise timing on health when formulating clinical recommendations.

Notable strengths of the current study include its prospective design with one-year follow-up and evaluation of the present research questions among physically active weight loss maintainers, which can offer insights into novel strategies for fostering high levels of MVPA and preventing weight regain. This study also extends our previous work in a larger sample with minimal overlap, further strengthening the reliability of findings on the relation of exercise timing to MVPA levels. There are also study limitations. The observational design precludes conclusions about causal relationships. Additionally, we were most interested in whether temporally consistent exercisers continued to report greater MVPA and automaticity over time rather than in demonstrating the independent predictive effect of temporally consistent exercise above and beyond initial levels of MVPA and automaticity. Thus, it is possible that temporal consistency serves as a marker of weight loss maintainers who are more active in general but does not predict future activity and exercise automaticity once accounting for initial levels of MVPA and automaticity; experimental testing will be critical to elucidate these relationships. Importantly, while the NWCR provides a unique resource to evaluate behaviors related to weight loss maintenance, generalizability of the findings is limited by a homogenous sample, especially regarding racial and ethnic identity, and—per analyses comparing retained participants to those who were excluded—our sample may have been biased toward men and those with greater weight loss duration and greater activity levels. Thus, it is critical that our findings be replicated in more diverse samples. Although we used validated and widely used questionnaires whenever possible (i.e., to assess MVPA levels and exercise automaticity), all variables of interest were self-reported by participants and may thus be subject to reporting bias. Replication is needed in which both MVPA levels and timing patterns are measured using accelerometers. Additionally, while the timing classification criteria we used are most consistent with approaches used in the literature to date, there may be alternative ways to characterize exercising timing patterns (e.g., classifying all exercise episodes starting within a 2 hour period day-to-day as temporally consistent), which should be further explored in future studies. Lastly, as many individuals in the present sample were highly active, future research is needed that examines the potential importance of exercise timing among samples with a broader range of physical activity levels, including lower average levels of MVPA. It is also important to note that, while temporally consistent exercisers reported more MVPA and stronger exercise habits, temporally inconsistent exercisers were also highly active and had been successful with weight loss maintenance. Thus, future research should examine for whom consistent exercise timing and exercise at a specific time of day (e.g., morning vs. evening) is most beneficial.

Conclusions

In conclusion, findings indicate that temporal exercise consistency relates to greater MVPA levels and stronger exercise habit across time among weight loss maintainers. Consistent early morning exercise, in particular, predicted more stable exercise routines across the one-year study period and was related to the highest levels of exercise automaticity. While future experimental research is needed to test the causal effects of prescribed exercise timing on MVPA behavior and habit formation, the present findings suggest that regularly exercising at the same time of day could help to increase or sustain high levels of MVPA and make exercise more habitual, with early morning exercise potentially being most beneficial for establishing a stable exercise routine and strong exercise habit.