Description of Studies
A total of 5183 individual studies were identified through the initial search process after the removal of duplicates. Thirty-one studies underwent full-text review, and 9 out of the 31 studies were excluded. Studies were excluded according to the following exclusion criteria: (1) did not meet intervention or comparator criteria (n = 3, e.g., combined continuous and accumulated exercise, accumulated exercise in a single session, such as HIIT); (2) only the abstract was available (n = 3); (3) did not meet outcome criteria (n = 2, e.g., did not provide any of the outcomes of interest, including glucose-, insulin-, or TG-related measures); and (4) duplicates (n = 1). Three studies were identified from citation searching, and two studies were included from the updated search. Therefore, a total of 27 studies were included. The results of the systematic search are presented in Fig. 1. Below we provide a summary of the key characteristics (participants, study design, intervention, and outcome details) of these eligible studies (see Tables 1, 2 for an overview of short-term and long-term intervention study characteristics, respectively).
Study Designs
Of the 27 studies that met our inclusion criteria, four studies [14, 15, 30, 31] with 267 participants were long-term (≥ 2 weeks) intervention studies, with a duration of 2–15 weeks. All 23 short-term studies and two long-term intervention studies were randomized crossover design, whereas two long-term intervention studies were parallel randomized controlled trials [30, 31]. There were 13 out of the 23 short-term studies with a total of 118 participants utilized 1-day designs and examined the effect of exercise within the same day, with durations ranging from 4-h to 24-h, while ten studies with 250 participants utilized multi-day designs and examined the effect of exercise on the second-morning responses, including one study [13] that examined both the same-day and second-morning effect.
Participants
A total of 27 studies with 635 participants were included in the meta-analysis. Three of the long-term studies [14, 15, 31], a total of 133 participants, and three of the short-term studies [32,33,34], a total of 48 participants, recruited individuals with type 2 diabetes. In addition, one short-term study [16] included adults with insulin resistance (n = 9); all other studies were conducted in non-diabetic populations. Most of the participants were physically inactive or sedentary. Five [34,35,36,37,38] and seven [31, 33, 39,40,41,42,43] studies included only females and males, respectively. The average age of participants ranged from 22.6 [44] to 71.0 years [36]; the body mass index (BMI) ranged from 21.0 [20] to 34.0 kg/m2 [45]. Sample sizes ranged from 8 [41] to 134 [38].
Interventions
Accumulated exercise in the included studies could be characterized into two types: (1) 2–3 bouts of exercise (e.g., 10–15 min per bout) timed around meals; and (2) activity spread across frequent, brief bouts (≥ 5 bouts, e.g., 1–6 min per bout) throughout the day (known as PA breaks). Thirteen studies adopted PA breaks, while 14 studies (including four long-term intervention studies) adopted 2–3 bouts of accumulated exercise. For PA breaks, most studies (n = 12) used frequent brief bouts with 1–5 min per bout every 15–60 min in a 6.5–12 h period, except for one study [35] that included 5 bouts of 6 min cycling within 240 min. Moreover, except for four studies [35, 43, 44, 46] involving high-intensity exercise, most PA breaks (n = 9) involved low-moderate intensity exercise. For 2–3 bouts of exercise, most studies adopted 3 bouts of 10–15 min exercise (n = 11); three studies adopted 2*20 min [33], 3*30 min [39], and 2*24 min [30] accumulated exercise. Except for one study [16] involving high intensity exercise, all 2–3 bouts of exercise involved low-moderate intensity exercises. Most short bouts of exercise were before or after each main meal, with an interval of 4–5-h (n = 9), except for two studies with an interval of 20 min [20, 47]; one study conducted 2 bouts of 20 min walking before and after 40 min of lunch [33], while two long-term studies did not specify exercise timing or intervals [30, 31]. For energy-matched continuous exercise, except for a study involving one session of high-intensity interval exercise [44], all other studies used one bout of 30–90 min low-moderate intensity exercise. Two studies [48, 49] were excluded due to the unmatched energy expenditure.
Outcomes
Nineteen short-term studies, a total of 295 participants, reported PPG AUC indices. Of these studies, 12 comprising 210 participants, reported the same-day effect, while the remaining seven reported the second-morning effect. Four studies recruited individuals with type 2 diabetes, while the remaining 15 studies recruited non-diabetic populations. Moreover, 12 studies, comprising 214 participants, compared the effects of PA breaks with those of continuous exercise on PPG AUC, while the remaining seven studies, a total of 81 participants, compared 2–3 bouts of accumulated exercise with continuous exercise. Four studies included high-intensity exercise; five and three studies recruited only men and women, respectively. Six short-term studies, comprising 78 participants, reported 24-h glucose indices (24-h glucose AUC or 24-h mean glucose) measured using continuous glucose monitoring. Five short-term studies reported second-morning fasting glucose [13, 20, 41,42,43], while all four long-term intervention studies reported PPG AUC and fasting glucose.
Postprandial insulin AUC was reported in 12 studies, comprising 202 participants, of which, six studies examined the same-day effect. Fasting insulin levels were reported in seven studies with second-morning effects and two long-term intervention studies [31, 38].
Moreover, postprandial TG was reported in 13 studies, a total of 227 participants, of which, eight examined the same-day effect. Two and four long- and short-term intervention studies, respectively, reported fasting TG.
Intervention Effects
Glucose Measures
When considering short-term effects, some studies were designed to examine the same-day effects of the exercise intervention, while others examined these effects the following day via second-morning responses. For same-day effects, accumulated, compared to continuous, exercise showed a significant lowering effect on PPG AUC, with an SMD of − 0.36 (95% CI: [− 0.56, − 0.17], P = 0.0002, I2 = 1%) (Fig. 2). Subgroup analysis indicated that only PA breaks as accumulated exercise (SMD − 0.36 [95% CI: (− 0.64, − 0.08)], P = 0.01, I2 = 30%) reduced PPG AUC, compared to continuous exercise, while 2–3 bouts of accumulated exercise did not (SMD − 0.32 [95% CI: (− 0.74, 0.10)], P = 0.14, I2 = 0%) (Fig. 2). Moreover, accumulated, compared to continuous, exercise had a greater effect on PPG AUC in studies with nondiabetic populations (SMD − 0.36 [95% CI: (− 0.62, − 0.10)], P = 0.007, I2 = 16%) but not in studies with diabetic populations (Fig. 3), and in studies with low- to moderate-intensity exercise (SMD − 0.38 [95% CI: (− 0.59, − 0.17)], P = 0.0005, I2 = 0%), but not in studies with high-intensity exercise (Fig. 4). All studies measuring 24-h glucose (24-h mean glucose or 24-h glucose AUC) focused on same-day effects. Furthermore, 24-h glucose measures showed no significant differences between accumulated and continuous exercise (SMD − 0.21 [95% CI: (− 0.52, − 0.09)], P = 0.17, I2 = 0%) (Additional file 1: Fig. S1).
For second-morning effect on PPG AUC, no significant differences were observed between accumulated and continuous exercise (SMD 0.04 [95% CI: (− 0.29, 0.36)], P = 0.83, I2 = 0%) (Fig. 5). Moreover, no differences in exercise bout-based subgroup analysis were observed (Fig. 5). As less than two studies focused on high-intensity exercise or diabetic populations, no exercise intensity- and population-based subgroup analyses were performed. No differences were observed between accumulated and continuous exercise effects on second-morning fasting glucose (SMD − 0.07 [95% CI: (− 0.19, 0.06)], P = 0.29, I2 = 14%) (Additional file 1: Fig. S2).
For long-term intervention studies, no differences in PPG AUC (SMD − 0.55 [95% CI: (− 1.47, 0.37)], P = 0.24, I2 = 93%) (Additional file 1: Fig. S3) and fasting glucose (SMD − 0.46 [95% CI: (− 1.30, 0.37)], P = 0.28, I2 = 92%) (Additional file 1: Fig. S4) were observed between exercise conditions.
Insulin Measures
For short-term effects, no differences were observed between accumulated and continuous exercise effects on either same-day postprandial insulin AUC (SMD − 0.20 [95% CI: (− 0.44, 0.04)], P = 0.10, I2 = 0%) (Fig. 6) or second-morning postprandial (SMD − 0.29 [95% CI: (− 0.74, 0.15)], P = 0.20, I2 = 28%) (Additional file 1: Fig. S5) and fasting insulin (SMD − 0.06 [95% CI: (− 0.37, 0.24)], P = 0.69, I2 = 0%) (Additional file 1: Fig. S6).
Only two long-term intervention studies measured insulin levels, with no differences observed on either postprandial insulin AUC (SMD 0.18 [95% CI: (− 0.21, 0.56)], P = 0.36, I2 = 0%) or fasting insulin (SMD 0.07 [95% CI: (− 0.55, 0.69)], P = 0.82, I2 = 43%) outcomes.
TG Measures
For short-term effects, no differences were observed for same-day postprandial TG AUC (SMD 0.17 [95% CI: (− 0.34, 0.39)], P = 0.11, I2 = 0%) (Fig. 7), second-morning postprandial TG AUC (SMD 0.11 [95% CI: (− 0.24, 0.47)], P = 0.53, I2 = 0%) (Additional file 1: Fig. S7), and second-morning fasting TG (SMD − 0.08 [95% CI: (− 0.51, 0.35)], P = 0.73, I2 = 0%) (Additional file 1: Fig. S8). The pooled effects from two long-term intervention studies showed no difference in fasting TG (SMD 0.22 [95% CI: (− 0.16, 0.61)], P = 0.25, I2 = 0%).
Risk of Bias
Overall, most of the studies were with some concerns, excluding three studies with a high risk of bias [15, 31, 39] and five with a low risk of bias [13, 14, 44, 50, 51] (Fig. 8). Except for two studies that did not specify whether they were randomized trials [35, 39], all other studies were randomized trials. Of these, only seven reported randomization details [13, 14, 38, 44, 46, 50, 51]. Moreover, only five studies clearly reported that participants were blinded until arriving at the laboratory to complete the trials [13, 14, 44, 50, 51]. Although the study design does not enable participants or research staff to be blinded from the intervention and the measurement of the outcomes, the measured outcomes are difficult for either the participant or researcher to influence. Most studies did not report whether there were any missing data and how the missing data were handled. Two studies [15, 31] were considered as having a high risk of bias and one study [52] as with some concerns due to missing outcome data. One study [31] was defined as a “per-protocol” effect protocol, as one participant was excluded for not adhering to the exercise intensity protocol. Only one study was with some concerns for baseline imbalances due to differences in baseline BMI between groups [31]. Most crossover studies had at least a 3-day wash-out period between trials to avoid the carry-over effect; one study [32] had only a 1-day wash-out period.
Sensitivity Analyses and Publication Bias
Sensitivity analyses, in which studies with a high risk of bias were removed, did not substantially change the results. A series of sensitivity analyses were performed by removing each of the studies. This showed that when only one study [50] was removed, the subgroup analyses favoring the effects of accumulated exercise with PA breaks (Fig. 2), or in non-diabetic populations (Fig. 3), on same-day PPG AUC were no longer significant. Only the same-day effect on PPG AUC had at least 10 studies, a necessary requirement for conducting a publication bias assessment; funnel plots showed no indication of publication bias (Additional file 1: Fig. S9).