The Effects of Continuous Compared to Accumulated Exercise on Health: A Meta-Analytic Review

Most of the trials employed walking interventions (n = 12, [6, 25, 26, 33, 35‐37, 39‐42, 44]). Two trials each chose indoor cycling [31, 46] and jogging [32, 45] as their exercise mode, whereas one trial each used low-impact aerobics to music [38] and a variety of aerobic exercise modes including walking, jogging, cycling, cross-country skiing, rowing, and stair-climbing machines [34]. Only one trial [43] used resistance training (via resistance equipment) as the mode of exercise.

All exercise intensities set in the eligible trials would be considered moderate. Twelve of the trials set intensity relative to a percentage of the participant’s maximal heart rate (MHR) or heart rate reserve (HRR) (i.e. MHR minus resting heart rate). Of these trials, seven [25, 31, 32, 34, 35, 37, 45] used a directly measured MHR in their calculation of a relative target heart rate, whereas the remaining five [26, 33, 38, 39, 42] used age-predicted MHR (i.e. 220 minus age). The percentage heart rate employed by the 12 trials ranged between 60 and 80% of MHR or HRR. Two trials each prescribed exercise intensity based on the rate of perceived exertion (RPE)—one used a modified 0–10 Borg scale [41] and the other used the 6–20 Borg scale [38]. Two trials prescribed intensity based on a percentage of VO_2max. Eguchi et al. [46] instructed participants to cycle at a power output (W) corresponding to 50% of their directly measured VO_2max, whereas Chung et al. [40] asked participants to walk on a treadmill at an intensity corresponding to 83% of estimated VO_2max so that participants expended 200 kcal in 30 min of walking. One trial [43] set exercise intensity according to the number of repetitions of a resistance exercise performed at a certain percentage of 1 repetition maximum (1 RM). Two trials did not mention specific exercise intensity but prescribed ‘brisk’ walking [36, 42], whereas another trial reported only that the walking was ‘moderate’ intensity.

Only four trials used intensity to apply progression to their intervention [34, 38, 43, 46], and of these, only two provided specific details (Schachter et al. [38], from 40–50% to 65–75% HRR and 10–11 to 13–14 RPE; Quinn et al. [34], from 50–60% to 70–80% HRR). Eight trials, however, applied progression by increasing exercise time per week through increasing days/week, minutes/bout, or bouts/day over the intervention period [25, 26, 31, 37, 38, 41, 42, 44]. Shiau et al. [43] increased the load (% 1 RM) participants lifted in each exercise session to provide progression. Only one trial [25] did not provide details about the specific progression applied.

In the 22 accumulated exercise interventions included in the 19 eligible trials, the most common bout duration was 10 min (n = 15). Of these 15 interventions, one prescribed two bouts/day [36], 11 set three bouts/day [31‐33, 35, 37, 39‐41, 43, 45, 46], and three prescribed four bouts/day [42, 44]. In all but four of these interventions, frequency was set at 5 days/week (Chung et al. [40], 3 days/week; Eguchi et al. [46],  3 days/week; Shiau et al. [43], 3 days/week; Murtagh et al. [36], 3 days/week). Three trials consisted of two bouts of 15 min/day performed on 3–5 [38], 4 [34], and 5 [31] days/week. In the remaining interventions, one asked participants to accumulate 40 min/day of exercise in three bouts/day on 5 days/week [26]; another intervention directed participants to accrue 30 min of exercise in at least 10-min bouts (two to three bouts/day) daily [6]; similarly, in one of Coleman et al.’s [41] intervention groups, participants were told to perform bouts of at least 5 min to amass 30 min of exercise on 6 days/week; and finally another intervention had participants complete two bouts of exercise of a sufficient duration to expend 150 kcal per bout 5 days/week [25]. Shiau et al. [43] asked participants to perform one set each of nine resistance exercises, with 30-s recovery between sets, in three sessions (~ 10 min per session performed at 8 am, 5 pm and 9 pm) on 3 days/week.

The most common continuous exercise prescription was 30 min of exercise on 5 days/week (150 min/week total; n = 7 [31‐33, 35, 37, 39, 45]), followed by 40 min on 5 days/week (200 min/week total; n = 3 [26, 42, 44]). The largest amount of weekly exercise prescribed to participants was 7 days of 30 min (210 min/week total, [6]), whereas the least prescribed was 3 days of 20 min/week (60 min total [36]). One trial [25] did not prescribe exercise based on time; instead they gave participants an energy expenditure target of 300 kcal (1256 kJ) per exercise session. In another trial [43], participants performed three sets each of nine resistance exercises (30-s and 60-s rest between sets and exercises, respectively) at 5 pm on 3 days/week.

Only three trials consisted of completely supervised exercise sessions [31, 32, 40], whereas three trials included a mixture of supervised and unsupervised sessions [25, 35, 36]. It was unclear in one trial [43] whether participants were supervised. The exercise sessions were unsupervised in all other trials. Two of the supervised interventions took place in a university exercise facility, whereas the other one [40] was performed in a public health centre. The mixed supervision interventions occurred in a university gym [36] or campus [35], or in both an indoor track and outdoors [25]. For the unsupervised interventions, most were home-based (i.e. participant’s own house—including its surrounding areas, such as nearby streets, parks, etc.). Participants in one trial performed their exercise sessions in a military college fitness facility [43]. Eight trials consisted of home-based outdoor walking [6, 26, 33, 37, 39, 41, 42, 44]. One trial asked participants to exercise at home by following instructions on an aerobics video [38], whereas another trial gave treadmills to participants in one of the groups, so they could walk at home [42]. In a trial consisting of Japanese workers, each worker had access to a cycle ergometer placed within 5 min of their workplace [46]. DeBusk et al. [45] allowed participants to complete their prescribed jogging either at home or at work, whereas Quinn et al. [34] permitted participants to complete the exercise sessions using a variety of aerobic modalities in presumably a variety of different settings. All interventions were delivered in an individual format, except for two trials that involved a group exercise intervention [31, 43].

In addition to an exercise intervention, five trials included dietary modification, in the form of hypocaloric diets. The calorie-restricted diets in these trials consisted of 500 kcal less than participants’ Harris Benedict equation-derived daily energy intake [26], 300 kcal/day less than ‘usual intake’ [40], total calorie intake of 80% of resting energy expenditure [31], a target calorie intake of 1200–1500 kcal/day with fat limited to 20% of total intake [44], and 1500 kcal target intake for participants weighing at least 90 kg, and 1200 kcal for those weighing less than 90 kg [42]. The calorie restrictions were applied to all conditions in each trial. One additional trial [43] provided participants with nutrition instruction (recognition and recording of food categories and portions) from a nutritionist.

Six trials (35%) also provided an additional education or behaviour change component to their exercise arms. These components took the form of educational classes held at the beginning of the trials [26, 41], weekly or monthly meetings [38, 41, 42, 44], and phone calls during the study period [38, 39].

Of the nine (47%) trials that included a control group, seven trials asked control participants to maintain their usual activity routine throughout the study [26, 31, 33, 37‐39] [25, 40]. One trial did not report what advice was given to controls [35]. Three of the control groups could be described as attention controls, given that study personnel contacted participants during the study period [25, 37, 38].

Only one trial [38] was considered as having a low risk of selection bias, because the authors adequately generated their randomised sequence with a random component and adequately concealed allocation to the intervention so that participants and investigators could not foresee assignment to the trial conditions. Another five trials (26%) used a random component to assign participants to study conditions, but four of these trials [33, 35, 36, 43] were at a high risk of selection bias, because participants or investigators might have foreseen assignment to the study groups; the other trial [46] was at an unclear risk of bias due to insufficient information being provided.

Two trials used a non-random component to generate the sequence [31, 32] and participants or investigators could foresee group allocation, and were thus judged to be at a high risk of selection bias. One further trial [6] was at a high risk of selection bias because one of the authors oversaw randomisation and would have foreseen group assignment. We judged ten trials (53%) to have an unclear risk of selection bias, because they lacked descriptions of both the generation of the random sequence and the allocation concealment method.

All eligible trials included were at high risk for performance bias because it is not possible to blind the trial personnel and participants to exercise interventions. Similarly, all trials were considered at a high risk of detection bias because they either failed to blind outcome assessors or gave no information on blinding of all outcome assessors—which we judged as lack of blinding. However, two trials [33, 35] did employ outcome assessors blinded to participants’ group assignment to measure blood pressure. Outcome assessors were blinded to group allocation in only one of 19 trials [32] that assessed anthropometric measures.

Of the 1080 randomised participants who participated in the eligible trials, 17% dropped out before the end of the intervention period. Drop-outs were slightly higher in the accumulated exercise bouts conditions (20%) compared with continuous exercise (16%) and control (14%).

Only four trials (21%) were at a low risk of attrition bias; for three trials this was because they retained all participants to the end of the intervention period [32, 40, 46], whereas the other trial reported just one drop-out during the intervention [41]. One trial was considered at an unclear risk of attrition bias because although an intention-to-treat analysis was reported, the authors provided no information about how missing data were handled [25]. All other trials were considered at a high risk of attrition bias due to either high attrition (≥ 20% [6, 26, 33‐38, 40, 42, 47]) or inappropriate handling of missing data [39, 42, 44, 45].

We judged all 19 trials as having an unclear risk for reporting bias, as no study protocol, design paper, or trial registration was available; the information was therefore insufficient to judge this item for eligible trials.

Eighteen trials (95%) were at low risk of bias owing to adequate group similarity at baseline, and one was at high risk of bias, because the baseline BMI differed across groups [26].

In 12 trials (637 participants) that reported the percentage of total prescribed exercise sessions or time completed [25, 26, 33‐36, 38, 39, 42, 44‐46], the mean percentage adherence was lower in the accumulated exercise condition compared with continuous exercise (78 ± 17% vs. 83 ± 15%). Participants adhered adequately to the exercise intervention in 11 (61%) of the trials; however, in five trials (28%), adherence to the exercise intervention was so low that we judged it to cause a high risk of bias [6, 31, 38, 42, 46]. In one of these trials [6], participants met their daily goal of 30 min on only 2.3 days and 2.8 days per week for accumulated and continuous exercise, respectively, whereas another [31] reported that participants exercised on an average of 3.7 and 3.9 days of a target of 5 days a week for the respective conditions. Finally, three trials were judged to be at an unclear risk of bias because they provided no adherence data [37, 40, 43].

Meta-analysis was possible for comparisons between accumulated and continuous exercise for total minutes of exercise, percentage prescribed sessions completed, average days per week of exercise, self-reported exercise (overall, minutes per week, and minutes per day), objectively measured exercise (overall and pedometer measured only), heart rate (overall and average heart rate in bpm), and RPE. Of these outcomes, we found that continuous exercise groups completed a statistically higher percentage of prescribed sessions (MD − 3.88%, 95% CI − 6.92 to − 0.84; I² = 46%; eight studies, 384 participants), but the accumulated exercise group achieved statistically greater amounts of exercise when it was objectively measured (SMD 0.25, 95% CI 0.01–0.49, I² = 25%; six studies, 523 participants).

Subgroup analyses revealed that the effect on the percentage of prescribed sessions completed in favour of continuous exercise was only statistically significant for trials employing an exercise dose of 150 min (MD − 3.07, 95% CI − 4.47 to − 1.68, I² = 0%; three studies, 85 participants), and not those which prescribed fewer or greater than 150 min. However, the statistically higher objectively measured exercise in accumulated exercise groups was evident only in trials that prescribed an exercise dose of > 150 min (SMD 0.33, 95% CI 0.04–0.61, I² = 31%; four studies, 420 participants). No accumulated exercise versus control meta-analysis was possible due to too few trials reporting exercise levels in the control groups.

We found no statistical differences between accumulated and continuous exercise for any cardiorespiratory fitness outcomes (VO_2max, relative VO_2max, exercise economy, and test duration/distance) in our analyses of change from baseline scores to post-intervention values.

However, there was a moderate favourable effect on post-intervention VO_2max values with accumulated exercise compared with control (SMD 0.52, 95% CI 0.24–0.81, I² = 6%; four studies, 223 participants). Similarly, we found statistically higher relative VO_2max post-intervention values (MD 2.32 ml/kg/min, 95% CI 1.10–3.54, I² = 4%; three studies, 197 participants) and statistical improvements in relative VO_2max from baseline to post-intervention (MD 2.78 ml/kg/min, 95% CI 2.51–3.05, I² = 0%; two studies, 110 participants) with accumulated exercise versus control. However, there was no difference between accumulated exercise and controls when exercise economy and test duration or distance outcomes were pooled.

No statistical differences were observed in any subgroup analysis performed by exercise dose (< 150 min, 150 min, and > 150 min) in comparisons between accumulated exercise and either continuous exercise and control.

Additional cardiorespiratory fitness outcomes not included in the meta-analysis were as follows: Murtagh et al. [36] observed statistical pre- to post-intervention reductions in heart rate at stages 2 and 3 of a treadmill test (p < 0.05) and mean RPE (p < 0.05) in both accumulated and continuous exercise groups, but not in the control group. However, the authors reported no differences between groups for changes in mean VO₂ during a submaximal treadmill test pre- to post-intervention. Another study [34] found that after 12 weeks of training, walking economy (% VO_2max and % heart rate maximum while walking at 101.8 m/min) improved statistically (p < 0.05) in an accumulated exercise group, but not in a continuous exercise group. Statistical pre- to post-intervention reductions in SBP and DBP at 101.8 m/min walking speed were observed in the accumulated exercise group (p < 0.05), whereas only reductions in DBP were reported in the continuous exercise group (p < 0.05). Murphy and Hardman [35] found statistically increased VO₂ at 2 mmol/l in both accumulated and continuous exercise groups relative to controls (both p < 0.05), but no between exercise group differences. Another study [45] found that accumulated and continuous exercise similarly statistically decreased peak (both p < 0.01) and submaximal exercise (both p < 0.001) heart rate, but found no statistical changes in blood pressure during submaximal or maximal exercise in either group. Finally, one study [44] reported no pre- to post-intervention improvements in predicted VO₂ at heart rates of 110 and 125 bpm for both accumulated and continuous exercise groups (both p < 0.001), but no statistical between-group differences were found.

In meta-analyses of resting heart rate (post-intervention) and SBP and DBP (post-intervention and change from baseline) values, we found no statistical differences between accumulated and continuous exercise.

Compared with control, accumulated exercise was associated with statistically lower post-intervention DBP values (MD − 4.83 mmHg, 95% CI − 7.83 to − 1.84, I² = 26%; four studies, 161 participants), but no statistically different effects on SBP (post-intervention or change scores).

In comparisons between accumulated and continuous exercise and accumulated exercise and control, we found no statistical differences for any outcome by exercise dose.

We found a small but statistical reduction in body mass from baseline to post-intervention in favour of accumulated exercise compared with continuous exercise (MD − 0.92 kg, 95% CI − 1.59 to −0.25, I² = 0%; five studies, 211 participants). Participants in the accumulated exercise groups, however, did not have statistically lower body mass post-intervention than the continuous exercise groups. No differences between accumulated and continuous exercise were found for any other anthropometric or body composition outcome.

Compared with control, accumulated exercise statistically reduced baseline to post-intervention values for body mass (MD − 1.94 kg, 95% CI − 3.42 to − 0.47, I² = 82%; four studies, 97 participants), BMI (− 0.97 kg/m², 95% CI − 1.70 to − 0.24, I² = 79%; two studies, 48 participants), waist circumference (− 2.62 cm, 95% CI − 4.67 to − 0.56, I² = 67%; two studies, 44 participants), and sum of skinfolds (− 6.39 mm, 95% CI − 8.25 to − 4.53, I² = 0%; three studies, 70 participants). The removal of the most extreme value did not reduce heterogeneity in the body mass analysis above. A combined analysis of post-intervention and change scores revealed a statistical but small reduction in body fat percentage with accumulated versus continuous exercise (− 0.92%, 95% CI − 1.78 to − 0.07, I² = 0%; four studies, 147 participants). No statistical differences between accumulated exercise and control were observed for post-intervention body mass, BMI, and waist or hip circumference values.

In subgroup analyses by exercise dose, accumulating > 150 min of weekly exercise in multiple bouts per day resulted in statistical small effects on body fat percentage (combined post-intervention and change from baseline values; MD − 0.87%, 95% CI − 1.71 to − 0.04, I² = 0%; three studies, 166 participants) compared with 150 min of exercise amassed via single continuous bouts per day. Similarly, compared with control, accumulating 150 min/week of exercise in multiple bouts per day resulted in statistically lower body mass (post-intervention; − 3.01 kg, 95% CI − 4.34 to − 1.68, I² = 73%; two studies, 52 participants) and sum of skinfolds (post-intervention; − 6.46 mm, 95% CI − 8.38 to − 4.54, I² = 0%; two studies, 48 participants).

Additional anthropometric and body composition outcomes not included in the meta-analysis were as follows: Two studies found no statistical differences in body mass post-accumulated or continuous exercise [34, 45]. Quinn et al. [34] also reported no statistical difference between exercise groups for body fat or lean mass measured via six-site skinfold measurement and waist and hip circumferences. Another study [37] reported statistical pre- to post-intervention decreases (− 6.7%, p < 0.05) in body fat percentage in the continuous exercise group, but not in the accumulated exercise or control groups. Schmidt et al. [31] observed statistical within-group reductions in sum of circumferences (hip, waist, thigh, and upper arm) measures (p < 0.01) in accumulated and continuous exercise groups; however, no between-group differences were found. Finally, Jakicic et al. [42] reported no differences for bone mineral content between exercise groups.

Only seven trials (three with control groups [29, 36, 40]) reported blood biomarker data [29, 32‐34, 36, 40, 46]. We found small statistical baseline to post-intervention reductions in LDL cholesterol with accumulated versus continuous exercise (MD − 0.39 mmol/l, 95% CI − 0.73 to − 0.06, I² = 23%; two studies, 41 participants). No differences were observed for any other blood biomarker (total cholesterol, HDL cholesterol, triglycerides, fasting blood glucose, and fasting insulin). Compared with control, we found no statistical effects of accumulated exercise on any blood biomarker (total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides, plasma glucose). No subgroup analyses by exercise dose were possible for any blood biomarker because too few trials were available.

Additional blood biomarker outcomes not included in the meta-analysis were as follows: One study [34] showed modest statistical improvements in HDL cholesterol (p < 0.05) with accumulated exercise only, but no changes in any other lipid values following either accumulated or continuous exercise. Another study [32] found no between-group differences in changes in any lipid outcome. Chung et al. [40] observed a statistical interaction effect between time and group (p < 0.01) for the atherogenic index [(total cholesterol − HDL cholesterol)/HDL cholesterol], with contrast analysis revealing statistical increases in the control group, but not in the two exercise conditions. Eguchi et al. [46] found no differences in oxidative stress (plasma thiobarbituric acid reactive substances) between accumulated and continuous exercise groups. Finally, another study [29] found statistically lower 2-h glucose concentrations (p < 0.05) in both accumulated and continuous exercise groups compared with control, but no differences between exercise groups. In the same study, no statistical differences in 2-h insulin were observed between exercise groups or control.

In meta-analyses involving only 41 participants in two studies [33, 37] that measured mood [both via profile of mood states (POMS)], continuous exercise resulted in statistically lower depression and anxiety subscale scores (SMD 0.93, 95% CI 0.15, 1.71, I² = 27% and 0.68, 95% CI 0.03–1.32, I² = 0%). However, no statistical differences in vigour subscale scores were found. Only one of these studies included a control group, so accumulated exercise versus control comparisons were not possible. Similarly, both studies prescribed an exercise dose of 150 min (both 5 days of 3 × 10-min bouts), so no subgroup analysis by exercise dose was possible. Only Osei-Tutu and Campagna [37] included the POMS total mood disturbance score and observed statistical decreases in both accumulated and continuous exercise groups over the 8-week intervention, but not in a control group. Conversely, only Murphy et al. [33] reported anger, confusion, and fatigue POMS subscale scores, and found no statistical differences in the effects of accumulated versus continuous exercise.

Three trials [6, 33, 38] compared the effects of accumulated and continuous exercise on self-efficacy-related outcomes. We did not combine these in a meta-analysis because of differences between the constructs assessed. Schachter et al. [38] found statistical improvements in self-efficacy for managing pain, managing other symptoms, and performing functional tasks among women with fibromyalgia who exercised in continuous bouts compared with controls (p = 0.034) at mid-intervention, and greater improvements in self-efficacy among accumulated bout exercisers compared with control (p = 0.001) at the end of the 12-week intervention. However, the effect on self-efficacy was similar between the two exercise patterns. Murphy et al. [33] found no changes in self-efficacy for walking among participants walking in accumulated or continuous exercise bouts, and increases in self-efficacy for cycling, jogging, and stair climbing among accumulated bout walkers (p < 0.05) only. The third trial [6] reported similar decreases in self-efficacy related to achieving physical activity recommendations among those assigned to accumulated and continuous bouts (d = 0.40 in both cases).

Two trials [25, 38] compared the effects of accumulated and continuous exercise on pain-related outcomes using three self-report inventories (Fibromyalgia Impact Questionnaire, Body Pain Diagram, Arthritis Impact Measurement Scales 2) and physician-rated pain scores. In one trial [38], those assigned to the control group demonstrated improvements in pain (p = 0.046), whereas no differences were found within or between accumulated and continuous exercise bout groups. In the second trial, Asikainen et al. [25] measured self-reported exercise-related pain and injuries at the end of a 15-week intervention. Although 35% of exercise participants reported exercise-related pain, only 17% reported that the pain was sufficient to temporarily interrupt their participation. The authors [25] also observed that participants in the accumulated exercise bouts group reported statistically fewer lower-limb problems compared with continuous exercise group participants (p = 0.021).

No statistical differences were found for daily energy intake (kcal/day) or percentage of daily energy intake from fat between accumulated and continuous exercise, accumulated exercise and control, or any subgroup analysis by exercise dose. In one study [42] that did not report data, there were no statistical differences in pre- to post-intervention energy intake or macronutrient composition. Another study, by Alizadeh et al. [26], also reported no differences in changes in the percentage of energy from carbohydrate, fat, and protein post-accumulated versus continuous exercise.

Asikainen et al. [30] reported that the proportion of participants reaching maximum points on the one-leg squat test for lower-extremity muscle strength increased statistically in both accumulated and continuous exercise groups versus control (odds ratio 4.6 and 4.1 in accumulated and continuous groups, respectively, vs. control, p = 0.008). In the same trial, however, no between-group differences were observed in the proportion of participants reaching the maximum score on the one-leg standing balance test. Finally, walking time on the UKK 2-km Walk Test increased statistically in both exercise groups when compared with the control group (p < 0.001). Similarly, Shiau et al. [43] reported statistical improvements (p < 0.05) in maximal strength (via 1 RM bench press), anaerobic performance (via 30-s Wingate test), and blood lactate response to anaerobic exercise (30-s Wingate) after a 12 week accumulated (three bouts of one set of each exercise per session for 3 days/week) or continuous (three sets of each resistance exercise per session for 3 days/week) resistance training intervention, but no statistical between-group differences. DeBusk et al. [45] reported no statistical differences between accumulated and continuous groups for participant-reported sweating during exercise bouts, and overall enjoyment and convenience of exercise bouts. Finally, Shiau et al. [43] reported no statistical between-group differences in daily physical activity assessed via IPAQ.