Introduction
Globally, Type 2 Diabetes (T2D) is a cause of a major disease burden [
1], with obesity and sedentary lifestyle as key risk factors in development and progression of T2D [
2]. Despite individual-centered public health efforts (i.e. lifestyle interventions), T2D prevalence in high-income economies such as those in western Europe is at estimated 8.5% and still increasing [
1,
3,
4], with those persons in lower socio-economic status (SES) especially affected [
5‐
7]. Consequences of T2D can be serious, encompassing physical and psychological aspects of health, thus adversely impacting on individuals’ Quality of Life (QoL) [
8,
9]. In T2D-prevention, lifestyle interventions supporting weight-loss and weight-loss maintenance have potential to improve health related outcomes [
10‐
13], and, consequently, QoL [
14,
15].
Despite the potential benefits of lifestyle interventions in T2D-prevention, premature intervention cessation and stress are leading to sub-optimal intervention benefits [
13,
16‐
18]. Therefore, it is not enough to identify those individuals, who are benefiting from interventions (i.e. “successful achievers”) [
18], but also to identify those pathways that influence success of weight-loss and weight-maintenance interventions in T2D-prevention [
19,
20].
Premature intervention cessation is the result of a complex interaction between intervention inputs, individuals, and context variables [
19,
21,
22]. The question is known as “whiches conundrum” asked by King [
23]:
“Which intervention, for which people, under which circumstances?” While personality traits (e.g. neuroticism, extraversion) do not appear to be associated with intervention cessation in T2D-prevention [
22], factors such as higher baseline body mass index (BMI), younger age, employment or study, hesitancy about the efficiency of lifestyle changes, have been associated with intervention cessation [
16,
24‐
29]. There is no consensus about the influence of factors such as low mood on intervention cessation [
22,
30]. This may likely be due to the fact, that associations of mood and treatment cessation are moderated by other variables.
Identification of pathways to successful weight-loss maintenance is challenging owing to the interconnectivity and interchangeability between factors, in where factors can have both direct and indirect influence on the outcomes [
20,
31,
32]. Precisely because of the partly contradictory findings [
33,
34], an improved understanding of the interactions between different factors would enable more targeted lifestyle interventions supporting weight-loss and weight-loss maintenance for individuals with prediabetes. While a considerable body of literature has examined factors associated with successful weight-loss maintenance [
34,
35], there is limited evidence from large scale studies examining complex pathways associated with intervention cessation and chronic stress in T2D-prevention.
In this study, health-related QoL, social support, use of primary care, and mood at the start of a lifestyle intervention were examined. It was hypothesized that these variables function influence intervention cessation and chronic stress, both of which are associated with less favorable weight-loss and weight-loss maintenance outcomes [
13,
33,
36]. Furthermore, it was examined whether sex and SES moderated relationships between health-related QoL, social support, use of primary care, and mood as predictors of intervention cessation and chronic stress.
Higher risk of adverse consequences from T2D has been associated, for example, with sex (men), non-Caucasian ethnicity, and lower SES, which, in turn, are associated with lower likelihood of enrolment and higher likelihood of intervention cessation [
37‐
41]. Sex has been indicated as a potential moderator for variables such as mood, chronic stress, and eating restraint during weight-loss and weight maintenance [
18,
42]. Although social support has been associated with positive weight-loss outcomes especially among women, overall role of social support in T2D-prevention is less well understood [
29]. Further, while men are less likely to participate and attend regularly in lifestyle interventions, role of participant’s sex in moderating relationships between different predictor variables and intervention cessation and chronic stress during weight-loss and weight loss maintenance is less clear [
26,
27,
43].
Lower SES (measured as income) and higher chronic stress have been associated with worse weight-loss outcomes for both men and women and with overall lower QoL [
18,
44‐
47], as well as with intervention cessation [
18,
45]. In addition, associations between weight and SES (measured as level of education and income), appear, at least partially, mediated by sex [
5,
48,
49]. For women, higher SES has been associated with lower likelihood of obesity but similar association has not been observed between men in lower and higher SES [
50].
Beyond premature intervention cessation, chronic stress has been identified as one of the key factors hindering weight-loss and weight-loss maintenance efforts in lifestyle interventions [
18,
36]. Stress, a physiological and/or psychological response to perceived internal or external stressors can be seen as adaptive (short-term) or harmful (long-term) [
51]. Stressors, i.e. events or conditions that lead to physical or psychological stress, can be dependent on individual’s life situation, SES, and ethnic group membership [
51‐
53]. Especially chronic stress influences individuals’ biological systems negatively, which, consecutively, can have negative influences on daily functioning, cognitive capacities, and health [
52]. Stress is also associated with poorer health behaviors, including poorer dietary choices and physical inactivity [
36,
54], thus increasing the risk of weight-gain [
32,
55]. Subsequently, higher perceived chronic stress can be counterproductive in lifestyle modification interventions aiming for weight-loss and weight-loss maintenance [
18,
32]. Lower SES has been associated with higher stress and increased weight-gain especially among men [
31].
In these secondary analyses, data were analyzed from a group-based T2D-prevention intervention PREVIEW (PREVention of diabetes through lifestyle Intervention and population studies in Europe and around the World) with weight-loss and weight-loss maintenance phases [
13]. Here at the baseline of the PREVIEW intervention, it was examined whether the QoL, social support, primary care utilization, and mood at the PREVIEW intervention baseline were associated, firstly, with intervention cessation and, secondly, with chronic stress. We hypothesized that lower QoL and lower social support, as well as higher primary care use and mood disturbances would predict higher likelihood of intervention cessation and higher chronic stress. Thirdly, we examined if QoL, social support, primary care utilization, and mood were independent predictors of intervention cessation and chronic stress over and above baseline BMI. We hypothesized that QoL, social support, primary care utilization, and mood would predict intervention cessation independently from BMI. Fourthly, we examined whether sex and SES moderated relationships between predictor variables (i.e. QoL, social support, primary care utilization, mood) and intervention cessation and chronic stress. Here, degree of education was used as an indicator of SES [
56,
57].
Results
Participants and predictor variables characteristics are shown in Tables
1 and
2. Chi-square tests between points of cessation, stress, and social-demographic characteristics were performed and summarized in Table
3.
Table 1
Baseline participants demographic characteristics for all participants and separated by point of cessation and chronic stress at the beginning of the PREVIEW intervention
Age (m ± sd) | 51.6 ± 11.6 | 47.6 ± 12.5 | 48.4 ± 12.1 | 51.1 ± 11.2 | 54.8 ± 10.1 | 55.5 ± 10.4 | 52.9 ± 11.0 | 49.2 ± 11.7 | 47.8 ± 11.8 |
Sex (%) |
Men | 720 (32.4%) | 96 (26.5%) | 155 (32.5%) | 126 (30.1%) | 343 (35.7%) | 224 (42.7%) | 240 (34.6%) | 155 (27.4%) | 101 (23.1%) |
Women | 1500 (67.6%) | 266 (73.5%) | 322 (67.5%) | 293 (69.9%) | 619 (64.3%) | 300 (57.3%) | 454 (65.4%) | 410 (72.6%) | 336 (76.9%) |
BMI (m ± sd) | 35.4 ± 6.6 | 36.4 ± 7.9 | 37.0 ± 6.9 | 36.8 ± 6.6 | 33.5 ± 5.3 | 34.3 ± 5.4 | 34.6 ± 5.9 | 35.8 ± 6.8 | 37.2 ± 8.0 |
Degree of education (%) |
Up to secondary education | 373 (16.8%) | 75 (20.7%) | 79 (16.6%) | 86 (20.5%) | 133 (13.8%) | 74 (14.1%) | 131 (18.9%) | 78 (13.8%) | 90 (20.6%) |
Secondary vocational education | 389 (17.5%) | 61 (16.9%) | 100 (21.0%) | 64 (15.3%) | 164 (17.0%) | 101 (19.3%) | 106 (15.3%) | 105 (18.6%) | 77 (17.6%) |
Higher vocational education | 367 (16.5%) | 46 (12.7%) | 65 (13.6%) | 70 (16.7%) | 186 (19.3%) | 109 (20.8%) | 123 (17.7%) | 82 (14.5%) | 53 (12.1%) |
University education | 835 (37.6%) | 134 (37.0%) | 176 (36.9%) | 161 (38.4%) | 364 (37.8%) | 176 (33.6%) | 255 (36.7%) | 242 (42.8%) | 162 (37.1%) |
Other | 256 (11.5%) | 46 (12.7%) | 57 (11.9%) | 38 (9.1%) | 115 (12.0%) | 64 (12.2%) | 79 (11.4%) | 58 (10.3%) | 55 (12.6%) |
Ethnicity |
Caucasian | 1944 (87.6%) | 279 (77.1%) | 429 (89.9%) | 345 (82.3%) | 891 (92.6%) | 484 (92.4%) | 622 (89.6%) | 485 (85.8%) | 353 (80.8%) |
Other | 276 (12.4%) | 83 (22.9%) | 48 (10.1%) | 74 (17.7%) | 71 (7.4%) | 40 (7.6%) | 72 (10.4%) | 80 (14.2%) | 84 (19.2%) |
Marital status |
Married or Civil Partnership | 1502 (67.7%) | 212 (58.6%) | 309 (64.8%) | 267 (63.7%) | 714 (74.2%) | 366 (69.8%) | 484 (69.7%) | 385 (68.1%) | 267 (61.1%) |
Divorced, widowed, separated | 354 (15.9%) | 73 (20.2%) | 74 (15.5%) | 73 (17.4%) | 134 (13.9%) | 99 (18.9%) | 105 (15.1%) | 76 (13.5%) | 74 (16.9%) |
Single or other | 364 (16.4%) | 77 (21.3%) | 94 (19.7%) | 79 (18.9%) | 114 (11.9%) | 59 (11.3%) | 105 (15.1%) | 104 (18.4%) | 96 (22.0%) |
Household – Living with others |
One adult | 447 (20.1%) | 84 (23.2%) | 98 (20.5%) | 88 (21.0%) | 177 (18.4%) | 109 (20.8%) | 129 (18.6%) | 109 (19.3%) | 100 (22.9%) |
Two adults | 922 (41.5%) | 120 (33.1%) | 182 (38.2%) | 167 (39.9%) | 453 (47.1%) | 261 (49.8%) | 310 (44.7%) | 218 (38.6%) | 133 (30.4%) |
Three or more adults | 340 (15.3%) | 51 (14.1%) | 76 (15.9%) | 64 (15.3%) | 149 (15.5%) | 68 (13.0%) | 113 (16.3%) | 85 (15.0%) | 74 (16.9%) |
One adult and at least one child | 48 (2.2%) | 18 (5.0%) | 14 (2.9%) | 11 (2.6%) | 5 (0.5%) | - | 16 (2.3%) | 20 (3.5%) | 11 (2.5%) |
Two adults and at least one child | 397 (17.9%) | 73 (20.2%) | 90 (18.9%) | 76 (18.1%) | 158 (16.4%) | 74 (14.1%) | 111 (16.0%) | 117 (20.7%) | 95 (21.7%) |
Three or more adults at least one child | 66 (3.0%) | 16 (4.4%) | 17 (3.6%) | 13 (3.1%) | 20 (2.1%) | 11 (2.1%) | 15 (2.2%) | 16 (2.8%) | 24 (5.5%) |
Employment |
In Study or Employment (Regardless of hours) | 1429 (64.4%) | 226 (62.4%) | 315 (66.0%) | 298 (71.1%) | 590 (61.3%) | 317 (60.5%) | 463 (66.7%) | 392 (69.4%) | 257 (58.8%) |
Not economically active (e.g. carer, unemployed, off sick) | 230 (10.4%) | 52 (14.4%) | 56 (11.7%) | 39 (9.3%) | 83 (8.6%) | 32 (6.1%) | 60 (8.6%) | 63 (11.2%) | 75 (17.2%) |
Retired | 375 (16.9%) | 47 (13.0%) | 62 (13.0%) | 57 (13.6%) | 209 (21.7%) | 126 (24.0%) | 125 (18.0%) | 71 (12.6%) | 53 (12.1%) |
Other | 186 (8.4%) | 37 (10.2%) | 44 (9.2%) | 25 (6.0%) | 80 (8.3%) | 49 (9.4%) | 46 (6.6%) | 39 (6.9%) | 52 (11.9%) |
Table 2
Descriptive statics for the variables of QoL, mood, social support, and healthcare utilisation for all participants and separated by point of cessation and chronic stress at the beginning of the PREVIEW intervention
Social-cognitive variables |
QoL Physical health (4 – 20) | 15.1 ± 2.5 | 14.4 ± 2.6 | 14.7 ± 2.6 | 15.0 ± 2.5 | 15.6 ± 2.2 | 16.6 ± 1.9 | 15.8 ± 1.9 | 14.7 ± 2.1 | 12.8 ± 2.6 |
QoL Psychological (4 – 20) | 14.0 ± 2.3 | 13.5 ± 2.4 | 13.7 ± 2.4 | 13.9 ± 2.3 | 14.4 ± 2.1 | 15.6 ± 1.7 | 14.6 ± 1.7 | 13.5 ± 1.9 | 11.9 ± 2.3 |
QoL Social relationships (4 – 20) | 14.6 ± 2.8 | 14.1 ± 3.0 | 14.4 ± 2.9 | 14.4 ± 2.8 | 14.9 ± 2.6 | 15.7 ± 2.4 | 15.0 ± 2.4 | 14.2 ± 2.6 | 12.9 ± 3.2 |
QoL Environment (4 – 20) | 15.2 ± 2.3 | 14.2 ± 2.3 | 14.7 ± 2.3 | 15.3 ± 2.2 | 15.9 ± 2.0 | 16.7 ± 1.7 | 15.7 ± 1.8 | 14.6 ± 2.0 | 13.4 ± 2.3 |
Chronic stress (0 – 40) | 14.2 ± 6.3 | 16.5 ± 6.4 | 14.9 ± 6.3 | 14.2 ± 6.1 | 13.0 ± 6.0 | 26.7 ± 16.6 | 38.7 ± 18.0 | 54.3 ± 22.5 | 85.4 ± 33.9 |
Mood states (0 – 228) | 49.0 ± 30.7 | 57.8 ± 34.1 | 52.4 ± 31.1 | 50.4 ± 31.9 | 43.5 ± 27.3 | 6.3 ± 1.0 | 6.3 ± 1.0 | 6.3 ± 1.0 | 6.1 ± 1.1 |
Social support—family encouragement diet (1 – 5) | 2.7 ± 1.2 | 3.0 ± 1.2 | 2.8 ± 1.2 | 2.7 ± 1.1 | 2.6 ± 1.1 | 1.9 ± 0.9 | 2.1 ± 0.9 | 2.2 ± 1.0 | 2.4 ± 1.1 |
Social support—family discouragement diet (1 – 5) | 2.1 ± 1.0 | 2.4 ± 1.1 | 2.2 ± 1.0 | 2.2 ± 1.0 | 1.9 ± 0.9 | 2.0 ± 0.8 | 2.1 ± 0.8 | 2.3 ± 0.9 | 2.4 ± 0.9 |
Social support—friends’ encouragement diet (1 – 5) | 2.2 ± 0.9 | 2.4 ± 0.9 | 2.2 ± 0.9 | 2.2 ± 0.9 | 2.1 ± 0.8 | 1.8 ± 0.7 | 2.0 ± 0.7 | 2.1 ± 0.8 | 2.2 ± 0.9 |
Social support—friends’ discouragement diet (1 – 5) | 2.0 ± 0.8 | 2.2 ± 0.9 | 2.1 ± 0.8 | 2.0 ± 0.8 | 1.9 ± 0.7 | 2.0 ± 0.9 | 2.1 ± 0.9 | 2.0 ± 0.8 | 2.1 ± 1.0 |
Social support—family participation PA (1 – 5) | 2.0 ± 0.9 | 2.1 ± 0.9 | 1.9 ± 0.8 | 2.0 ± 0.9 | 2.1 ± 0.9 | 1.8 ± 0.8 | 1.9 ± 0.8 | 1.9 ± 0.8 | 1.9 ± 0.9 |
Social support—friends’ participation PA (1 – 5) | 1.9 ± 0.8 | 2.0 ± 0.9 | 1.8 ± 0.8 | 1.9 ± 0.8 | 1.8 ± 0.8 | 16.6 ± 1.9 | 15.8 ± 1.9 | 14.7 ± 2.1 | 12.8 ± 2.6 |
Primary healthcare utilisation during last 3 months |
Contact with GPa or NPa |
No contact | 995 (44.8%) | 165 (45.6%) | 231 (48.4%) | 176 (42.0%) | 423 (44.0%) | 266 (50.8%) | 320 (46.1%) | 230 (40.7%) | 179 (41.0%) |
Once | 566 (25.5%) | 89 (24.6%) | 115 (24.1%) | 105 (25.1%) | 257 (26.7%) | 140 (26.7%) | 180 (25.9%) | 153 (27.1%) | 93 (21.3%) |
Twice | 358 (16.1%) | 54 (14.9%) | 82 (17.2%) | 75 (17.9%) | 147 (15.3%) | 69 (13.2%) | 113 (16.3%) | 104 (18.4%) | 72 (16.5%) |
Three or more times | 301 (13.6%) | 54 (14.9%) | 49 (10.3%) | 63 (15.0%) | 135 (14.0%) | 49 (9.4%) | 81 (11.7%) | 78 (13.8%) | 93 (21.3%) |
Prescription renewal |
Did not renew prescription | 1669 (75.2%) | 277 (76.5%) | 374 (78.4%) | 314 (74.9%) | 704 (73.2%) | 392 (74.8%) | 525 (75.6%) | 441 (78.1%) | 311 (71.2%) |
Renewed prescription | 551 (24.8%) | 85 (23.5%) | 103 (21.6%) | 105 (25.1%) | 258 (26.8%) | 132 (25.2%) | 169 (24.4%) | 124 (21.9%) | 126 (28.8%) |
Diet or PA advice from healthcare professionals |
Did not receive advice | 1823 (82.1%) | 309 (85.4%) | 395 (82.8%) | 333 (79.5%) | 786 (81.7%) | 449 (85.7%) | 561 (80.8%) | 455 (80.5%) | 358 (81.9%) |
Received advice | 397 (17.9%) | 53 (14.6%) | 82 (17.2%) | 86 (20.5%) | 176 (18.3%) | 75 (14.3%) | 133 (19.2%) | 110 (19.5%) | 79 (18.1%) |
Referral to or contact with another specialist healthcare professional (any reason) |
No referral or contact | 1738 (78.3%) | 279 (77.1%) | 375 (78.6%) | 342 (81.6%) | 742 (77.1%) | 440 (84.0%) | 562 (81.0%) | 428 (75.8%) | 308 (70.5%) |
Referral or contact | 482 (21.7%) | 83 (22.9%) | 102 (21.4%) | 77 (18.4%) | 220 (22.9%) | 84 (16.0%) | 132 (19.0%) | 137 (24.2%) | 129 (29.5%) |
Money spend on PA activities |
No spend | 1112 (50.1%) | 201 (55.5%) | 286 (60.0%) | 211 (50.4%) | 414 (43.0%) | 242 (46.2%) | 319 (46.0%) | 298 (52.7%) | 253 (57.9%) |
Spend | 1108 (49.9%) | 161 (44.5%) | 191 (40.0%) | 208 (49.6%) | 548 (57.0%) | 282 (53.8%) | 375 (54.0%) | 267 (47.3%) | 184 (42.1%) |
Taking medication or supplements with or without prescription |
No medication or supplements | 885 (39.9%) | 180 (49.7%) | 221 (46.3%) | 156 (37.2%) | 328 (34.1%) | 192 (36.6%) | 276 (39.8%) | 220 (38.9%) | 197 (45.1%) |
Medication or supplements | 1335 (60.1%) | 182 (50.3%) | 256 (53.7%) | 263 (62.8%) | 634 (65.9%) | 332 (63.4%) | 418 (60.2%) | 345 (61.1%) | 240 (54.9%) |
Table 3
Baseline group comparisons between demographic variables for cessation and chronic stress
Ethnicity | Cessation χ2(3) = 72.17, p < .001 Chronic stress χ2 (3) = 33.84, p < .001 |
Other (i.e. non-Caucasian) | | 5.66 | - | 3.04 | -4.44 | -3.12 | - | - | 4.03 |
Marital status | Cessation χ2 (6) = 41.31, p < .001 Chronic stress χ2 (3) = 28.01, p < .001 |
Married or Civil Partnership | | - | - | - | 2.47 | - | - | - | - |
Divorced, widowed, separated | | - | - | - | - | - | - | - | - |
Single or other | | 2.29 | - | - | -3.48 | -2.90 | - | - | 2.88 |
Household Living with others | Cessation χ2 (15) = 53.85, p < .001 Chronic stress χ2 (15) = 68.91, p < .001 |
Two adults | | -2.47 | - | - | 2.67 | 2.94 | - | - | -3.60 |
One adult, at least one child | | 3.63 | - | - | -3.46 | -3.07 | - | - | - |
Two adults, at least one child | | - | - | - | - | - | - | - | - |
At least three adults, one child | | - | - | - | - | - | - | - | 3.05 |
Employment | Cessation χ2 (9) = 42.81, p < .001 Chronic stress χ2 (9) = 76.68, p < .001 |
In Study or Employment (Regardless of hours) | | 2.37 | - | - | - | - | - | - | - |
Not economically active (e.g. carer, unemployed, off sick) | | - | - | - | - | -3.02 | - | - | 4.42 |
Retired | | - | - | - | 3.65 | 3.98 | | -2.50 | -2.42 |
Other | | - | - | - | - | - | - | - | 2.54 |
Non-Caucasian ethnicity, being single, and living in a household with at least one child were associated with lower likelihood of achieving the ≥ 8% weight-loss (cessation group 1) and high chronic stress (stress group 4). Higher likelihood of completing the PREVIEW intervention (cessation group 4) was associated with being married and living in a two-adult household. Retired participants were more likely not only to complete the intervention (cessation group 4), but also to report low chronic stress (stress group 4).
Predictors of cessation
Multinomial logistic regression with cessation as an outcome variable and “age”, “degree of education”, “primary healthcare utilization”, “social support”, “moods”, and “QoL” as predictor variables indicated overall model significance (χ2 (66) = 347.8, p < 0.001) with good data fit (Pearson χ2 (6591) = 6651.9, p = 0.27). The following emerged as significant predictor variables; “QoL environment” (χ2 (3) = 52.7, p < 0.001), “family discouragement for diet” (χ2 (3) = 11.6, p = 0.009), “money spend on PA activities” (χ2 (3) = 15.2, p = 0.002), and “taking medication or supplements” (χ2 (3) = 18.4, p < 0.001).
Participants in very early (group 1) and early (group 2) cessation groups reported lower environmental QoL, while experiencing lower family support for diet changes. Participants in both, very early (group 1) and early (group 2) cessation groups, were also more likely to report not taking medication or supplements, while only participants in early cessation group (group 2) were less likely to spend money on PA (e. g. fitness offers) than those who completed the weight-maintenance phase. No variables were significantly associated with late cessation (group 3). Results with parameter estimates for variables associated with the different group memberships are summarized in Table
4.
Table 4
Parameter estimates for significant predictor variables for cessation compared to the PREVIEW study completers
VARIABLES ASSOCIATED WITH CESSATION |
Group 1 Very early cessation (Did not achieve weight-loss target) |
QoL Environment | -.25 | .04 | 42.74 | 1 | p < .001 | .78 |
Social Support—Family discouragement diet | .30 | .10 | 9.71 | 1 | p = .002 | 1.35 |
Healthcare utilisation – Not taking medication or supplements | .50 | .14 | 12.63 | 1 | p < .001 | 1.66 |
Group 2 Early cessation (Drop-out by early PREMIT maintenance stage) |
QoL Environment | -.16 | .04 | 21.28 | 1 | p < .001 | .85 |
No spend on PA activities during last 3 months | .47 | .12 | 14.53 | 1 | p < .001 | 1.60 |
Healthcare utilisation – Not taking medication or supplements | .42 | .13 | 10.71 | 1 | p = .001 | 1.51 |
SEX AS MODERATOR OF CESSATION |
Group 1 Very early cessation (Did not achieve weight-loss target) |
QoL Environment - |
Men | -.36 | .08 | 20.73 | 1 | p < .001 | .70 |
Women | -.23 | .05 | 26.76 | 1 | p < .001 | .79 |
Between levels of up to secondary and secondary vocational z-score = 3.92 significant | | | | | | |
Family discouragement diet - |
Women | .31 | .12 | 7.16 | 1 | p = .007 | 1.37 |
Not taking medication or supplements - |
Women | .48 | .17 | 8.21 | 1 | p = .004 | 1.62 |
Group 2 Early cessation (Drop-out by early PREMIT maintenance stage) |
Not taking medication or supplements |
Men | .8 | .20 | 11.5 | 1 | p = .001 | 2.10 |
QoL Environment - |
Women | -.18 | .04 | 17.68 | 1 | p < .001 | .84 |
Family discouragement diet – |
Women | .35 | .11 | 10.15 | 1 | p = .001 | 1.43 |
Group 3 Late Cessation (Drop out after early PREMIT maintenance stage) |
Family discouragement diet |
Women | .30 | .11 | 6.80 | 1 | p = .009 | 1.34 |
DEGREE OF EDUCATION AS MODERATOR OF CESSATION |
Group 1 Very early cessation (Did not achieve weight-loss target) |
QoL Environment - |
Up to secondary education | -.27 | .10 | 7.77 | 1 | p = .001 | .77 |
Secondary vocational education | -.42 | .10 | 16.32 | 1 | p < .001 | .66 |
University education | -.24 | .06 | 14.91 | 1 | p < .001 | .79 |
Between levels no significant differences |
Family discouragement diet –Family discouragement diet – | | | | | | |
University education | .57 | .16 | 12.68 | 1 | p < .001 | 1.78 |
Group 2 Early cessation (Drop-out by early PREMIT maintenance stage) |
QoL Environment - |
University education | -.25 | .06 | 19.48 | 1 | p < .001 | .78 |
Social Support—Family discouragement diet University education | .44 | .15 | 8.67 | 1 | p = .003 | 1.55 |
Group 3 Late cessation (Drop out after early PREMIT maintenance stage) |
Family discouragement diet – |
University education | .58 | .15 | 14.00 | 1 | p < .001 | 1.78 |
Cessation – Independence of the significant predictors from BMI
Two logistic regressions were calculated with the cessation as dependent variable. The first model was calculated with “BMI” as the predictor variable (χ2 (3) = 149.5, p < 0.001, Goodness of fit Pearson χ2 (6589) = 6596.2, p = 0.45). The second model was calculated with “BMI”, “QoL environment”, “family discouragement for diet”, “taking medication or supplement”, and “money spent on PA activities” as predictor variables (χ2 (15) = 367.0, p < 0.001, Goodness of fit Pearson χ2 (6642) = 6679.3, p = 0.37). Comparison of the models suggested significant improvement when predictors were added (χ2 = 367.0 – 149.5 = 217.5, df = 15 – 3 = 12, χ2 (12) = 217.5 p < 0.001).
Sex and degree of education as moderating variables for cessation
Sex
Multinomial logistic regression with “sex” as moderating variable was calculated with cessation as dependent and “age”, “degree of education”, “primary healthcare utilization”, “social support”, “moods”, and “QoL” as predictor variables. The overall the model was significant (χ2 (105) = 387.4, p < 0.001) with good data fit (Pearson χ2 (6552) = 6711.0, p = 0.083). Of the predictor variables, significant interaction was observed for “sex” * “taking medication or supplements” (χ2 (6) = 24.7, p < 0.001), “sex” * “QoL Environment” (χ2 (6) = 58.0, p < 0.001), and “sex” * “family discouragement for diet” (χ2 (6) = 19.4, p = 0.004).
Being woman was associated with lower perceived family support for diet changes within all groups (very early, early, and late cessation groups). For women lower likelihood of taking medication or supplements was observed in very early cessation group (group1) and for men in early cessation group (group 2). Further, only for men in very early cessation group (group 1) lower perceived environmental QoL was observed and only for women in early cessation group (group 2). Results and parameter estimates for variables associated with different group memberships are summarized in Table
4.
Degree of education
Multinomial logistic regression was calculated with “degree of education” as moderating variable, “cessation” as a dependent variable, and “age”, “healthcare utilization”, “social support”, “moods”, and “QoL” as predictor variables. The overall model was significant (χ2 (267) = 590.7, p < 0.001), but without good data fit (Pearson χ2 (6390) = 6592.5, p = 0.038). Of the predictor variables, significant interaction was observed for different interactive combinations: “degree of education” * “taking medication or supplements” (χ2 (15) = 33.3, p = 0.004), “degree of education” * “QoL environment” (χ2 (15) = 65.8, p < 0.001), and “degree of education” * “family discouragement for diet” (χ2 (15) = 38.2, p = 0.001).
For all groups (very early, early, and late cessation groups), a university degree was associated with lower perceived family support for diet changes. Further, university degree was associated with lower likelihood of taking medication or supplements among those in very early cessation group (group 1). Lower perceived environmental QoL was associated with university degree, but only for early cessation group (group 2). Results and parameter estimates for variables associated with different group memberships are summarized in Table
4.
Predictors of chronic stress
Multinomial logistic regression with “chronic stress” as an outcome variable and “age”, “degree of education”, “healthcare utilization”, “social support”, and “QoL” as predictor variables indicated that overall model significance (χ2 (66) = 1729.6, p < 0.001) but without good data fit (Pearson χ2 (6591) = 44,469.8, p < 0.001). Of the predictor variables “QoL physical health” (χ2 (3) = 21.3, p < 0.001), “QoL phycological health” (χ2 (3) = 54.6, p < 0.001), “QoL environment” (χ2 (3) = 85.2, p < 0.001), “mood states” (χ2 (3) = 447.3, p < 0.001), and “sex” (χ2 (3) = 11.8, p = 0.008) were significant predictors.
Medium–low, medium–high, and high stress groups were associated with lower QoL for both psychological health and environment, as well as higher mood disturbances. Furthermore, medium–high and high stress were both associated with woman sex, with high stress being also associated with lower reported physical health QoL. Results and parameter estimates for variables associated with different group memberships are summarized in Table
5.
Table 5
Parameter estimates for significant predictor variables for chronic stress compared to low chronic stress
VARIABLES ASSOCIATED WITH STRESS |
Low-medium chronic stress |
QoL Psychological health | -.17 | .05 | 13.22 | 1 | p < .001 | .85 |
QoL Environment | -.22 | .04 | 25.75 | 1 | p < .001 | .80 |
Mood disturbance (POMS) | .04 | .01 | 64.74 | 1 | p < .001 | 1.04 |
High-medium chronic stress |
QoL Psychological health | -.33 | .05 | 38.24 | 1 | p < .001 | .72 |
QoL Environment | -.41 | .05 | 68.07 | 1 | p < .001 | .67 |
Mood disturbance (POMS) | .07 | .01 | 148.11 | 1 | p < .001 | 1.07 |
Sex—Women | .50 | .16 | 9.17 | 1 | P = .002 | 1.64 |
High chronic stress |
QoL Physical health | -.19 | .05 | 12.36 | 1 | p < .001 | .83 |
QoL Psychological health | -.43 | .06 | 46.52 | 1 | p < .001 | .65 |
QoL Environment | -.47 | .06 | 64.81 | 1 | p < .001 | .63 |
Mood disturbance (POMS) | .10 | .01 | 269.08 | 1 | p < .001 | 1.10 |
Sex—Women | .66 | .21 | 9.49 | 1 | p = .002 | 1.93 |
SEX AS MODERATOR OF STRESS |
Low-medium chronic stress |
QoL Psychological health – Women | -.20 | .06 | 12.85 | 1 | p < .001 | .82 |
QoL Environment – |
Women | -.18 | .05 | 11.51 | 1 | p = .001 | 84 |
Men | -.33 | .08 | 18.66 | 1 | p < .001 | .72 |
Between levels no significant difference | | | | | | |
Mood disturbances – |
Women | .04 | .01 | 31.78 | 1 | p < .001 | 1.04 |
Men | .05 | .01 | 32.19 | 1 | p < .001 | 1.05 |
Between levels no significant difference | | | | | | |
High-medium chronic stress |
QoL Psychological health – |
Women | -.33 | .06 | 27.14 | 1 | p < .001 | .72 |
Men | -.32 | .10 | 10.82 | 1 | p = .001 | .73 |
Between levels no significant difference | | | | | | |
QoL Environment – |
Women | -.38 | .06 | 41.24 | 1 | p < .001 | .69 |
Men | -.46 | .09 | 25.53 | 1 | p < .001 | .63 |
Between levels no significant difference | | | | | | |
Mood disturbance – |
Women | .06 | .01 | 75.38 | 1 | p < .001 | 1.06 |
Men | .08 | .01 | 72.97 | 1 | p < .001 | 1.08 |
Between levels no significant difference | | | | | | |
High chronic stress |
QoL Physical health - |
Women | -.24 | .06 | 14.20 | 1 | p < .001 | .78 |
QoL Psychological Health – |
Women | -.40 | .07 | 28.57 | 1 | p < .001 | .67 |
Men | -.56 | .12 | 20.32 | 1 | p < .001 | .57 |
Between levels no significant difference | | | | | | |
QoL Environment – |
Women | -.40 | .07 | 33.45 | 1 | p < .001 | .67 |
Men | -.67 | .11 | 35.13 | 1 | p < .001 | .51 |
Between levels no significant difference | | | | | | |
Mood disturbance – |
Women | .09 | .01 | 156.68 | 1 | p < .001 | 1.10 |
Men | .11 | .01 | 105.36 | 1 | p < .001 | 1.11 |
Between levels no significant difference | | | | | | |
DEGREE OF EDUCATION AS MODERATOR OF STRESS |
Low-medium chronic stress |
QoL Psychological health – University education | .25 | .08 | 9.35 | 1 | p = .002 | .78 |
QoL Environment – |
Secondary vocational education | -.70 | .12 | 10.32 | 1 | p = .001 | .69 |
Higher vocational education | -.45 | .12 | 14.73 | 1 | p < .001 | .64 |
Between levels no significant differences | | | | | | |
Mood disturbance - |
Up to secondary education | .05 | .01 | 12.56 | 1 | p < .001 | 1.05 |
Secondary vocational education | .03 | .01 | 7.18 | 1 | p = .007 | 1.03 |
Higher vocational education | .04 | .01 | 11.75 | 1 | p = .001 | 1.05 |
University education | .05 | .01 | 27.36 | 1 | p < .001 | 1.05 |
Between levels of up to secondary and secondary vocational z-score = 2.43 significant | | | | | | |
High-medium chronic stress |
QoL Psychological health - |
Secondary vocational education | -.362 | .13 | 8.02 | 1 | p = .005 | .70 |
Higher vocational education | -.408 | .14 | 8.44 | 1 | p = .004 | .67 |
University education | -.398 | .09 | 19.45 | 1 | p < .001 | .67 |
Between levels no significant differences | | | | | | |
QoL Environment – |
Secondary vocational education | -.49 | .12 | 15.47 | 1 | p < .001 | .61 |
Higher vocational education | -.55 | .14 | 15.33 | 1 | p < .001 | .58 |
University education | -.35 | .08 | 18.74 | 1 | p < .001 | .71 |
Other | -.69 | .17 | 15.55 | 1 | p < .001 | .50 |
Between levels no significant differences | | | | | | |
Mood disturbance – |
Up to secondary education | .07 | .01 | 24.53 | 1 | p < .001 | 1.07 |
Secondary vocational education | .05 | .01 | 18.37 | 1 | p < .001 | 1.05 |
Higher vocational education | .07 | .01 | 25.59 | 1 | p < .001 | 1.07 |
University education | .08 | .01 | 62.58 | 1 | p < .001 | 1.08 |
Other | .06 | .01 | 18.07 | 1 | p < .001 | 1.07 |
Between levels no significant differences | | | | | | |
High chronic stress |
QoL Psychological health - |
Secondary vocational education | -.46 | .15 | 9.80 | 1 | p = .002 | .63 |
Higher vocational education | -.58 | .19 | 8.93 | 1 | p = .003 | .56 |
University education | -.49 | .11 | 20.39 | 1 | p < .000 | .61 |
Between levels no significant differences | | | | | | |
QoL Environment – |
Up to secondary education | -.36 | .14 | 63.6 | 1 | p = .012 | .70 |
Secondary vocational education | -.53 | .15 | 13.04 | 1 | p < .001 | .59 |
Higher vocational education | -.83 | .19 | 19.49 | 1 | p < .001 | .43 |
University education | -.37 | .09 | 14.93 | 1 | p < .001 | .69 |
Other | -.65 | .21 | 9.83 | 1 | p = .002 | .52 |
Between levels no significant differences | | | | | | |
Mood disturbance – |
Up to secondary education | .10 | .01 | 46.03 | 1 | p < .001 | 1.11 |
Secondary vocational education | .07 | .01 | 26.55 | 1 | p < .001 | 1.07 |
Higher vocational education | .10 | .02 | 39.55 | 1 | p < .001 | 1.11 |
University education | .12 | .01 | 116.33 | 1 | p < .001 | 1.13 |
Other | .10 | .02 | 33.15 | 1 | p < .001 | 1.10 |
Between levels of up to secondary and secondary vocational z-score = 2.42 significant | | | | | | |
Chronic stress—Independence of the significant predictors from BMI
Two logistic regression models were calculated with chronic stress as the dependent variable. The first model was calculated with “BMI” as the predictor variable (χ2 (3) = 61.18, p < 0.001, Goodness of fit Pearson χ2 (6582) = 6578.85.0, p = 0.51). The second model was calculated with “BMI”, “QoL psychological health”, “QoL physical health”, “QoL environment”, “mood disturbances”, and “sex” as predictor variables (χ2 (18) = 1638.42, p < 0.001, Goodness of fit Pearson χ2 (6639) = 17,229.36, p < 0.001). Comparison of the models suggested significant improvement between the models when predictors were added (χ2 = 1638.42 – 61.18 = 1557.24, df = 18 – 3 = 15, χ2 (15) = 1557.24 p < 0.001).
Sex and degree of education as moderating variables for chronic stress
Sex
Multinomial logistic regression with sex as moderating variable was calculated with “age”, “degree of education”, “primary healthcare utilization”, “social support”, and “QoL”. The overall the model was significant (χ2 (105) = 1756.7, p < 0.001) but without good data fit (Pearson χ2 (6552) = 35,745.11 p < 0.001). Of the predictor variables, significant interaction with sex was observed for “QoL physical health” (χ2 (6) = 28.1, p < 0.001), “QoL psychological health” (χ2 (6) = 58.1, p < 0.001), “QoL environment” (χ2 (6) = 92.5, p < 0.001), and “mood disturbances” (χ2 (6) = 453.0, p < 0.001).
Being man or woman was found to moderate the associations for low-medium and high stress. For low-medium stress group lower psychological health QoL and for high stress group lower physical health QoL were associated with women but not with men. Results and parameter estimates for variables associated with different group memberships are summarized in Table
5.
Degree of education
Multinomial logistic regression with “degree of education” as moderating variable was calculated. The overall model was significant (χ2 (267) = 1962.9, p < 0.001) but Goodness-of-Fit test did not indicate good data fit (Pearson χ2 (6390) = 12,240.0 p < 0.001). Of the predictor variables, significant interaction with “degree of education” was observed for “QoL physical health” (χ2 (15) = 33.0, p = 0.005), “QoL psychological health” (χ2 (15) = 67.2, p < 0.001), “QoL environment” (χ2 (15) = 100.2, p < 0.001), and “mood states” (χ2 (15) = 440.1, p < 0.001).
Also, degree of education moderated the association for low-medium and high stress groups. For the low-medium stress group university degree was associated with lower psychological health QoL. For both low-medium and high stress groups, those with up to a secondary degree of education reported fewer mood disturbances than those with higher degree of education. Results and parameter estimates for variables associated with different group memberships are summarized in Table
5.
Discussion
Achieving weight-loss and weight-loss maintenance, key components of T2D-prevention, can be very challenging even when supportive behavioral interventions are offered [
17,
18]. In the present study, variables and pathways, i.e. interactions between intervention inputs, individuals, and context variables [
19,
21,
22], associated with premature intervention cessation and chronic stress at the start of an intervention were examined [
16,
36]. Our results supported the notion that successful intervention completion is a complex and dynamic process, relying on interactions between intervention inputs and personal factors [
19,
21,
22]. Findings indicated that pathways between QoL, social support, primary care utilization, mood and chronic stress as well as cessation were moderated by both sex and degree of education, a prominent and significant dimension of SES.
As expected based on the previous research, overall lower QoL was associated with both intervention cessation and chronic stress [
14,
15]. Although only lower environmental QoL was associated with cessation [
16,
28,
29], higher chronic stress was more broadly associated with lower QoL [
8,
9]. It is not clear why only environmental QoL, which according to WHO [
64] encompasses aspects such as safety, access to medical services, availability of resources, and opportunities for skills acquisition was associated with cessation. Furthermore, the result is difficult to interpret as lower environmental QoL was associated with very early intervention cessation especially with men and with early cessation especially for women and those with a university degree, which, in itself, was associated with higher QoL [
47].
In accordance with previous research [
29,
71], lack of family support was associated with earlier intervention cessation especially for women in this study. Furthermore, SES, represented here as degree of education [
5,
56], was found to moderate between intervention cessation and social support, especially lack of family support for diet changes. Lack of family support was associated particularly with university degree. While in former studies being a man who had reached only a lower degree of education had been associated with less favorable intervention outcomes [
18,
38,
45], in this study women with higher degrees of education were at risk of poorer intervention outcomes. Although degree of education and sex can moderate relationship between intervention outcomes [
38,
41,
50], in the present study we observed that higher degrees of education bear overall a risk for non-completing the PREVIEW intervention. Combined with the observation that in PREVIEW intervention single parents were least likely to achieve weight-loss, our results highlighted the lack of resources such as time as a factor for less favorable outcomes especially for university educated women with family responsibilities.
As mood disturbances are closely associated with stress [
52], association between the variables was expected. Despite previous research indicating sex as a potential moderating variable [
42], in this study sex was not found to moderate the relationship between mood disturbances and chronic stress, thereby adding to the inconclusive body of literature examining mood disturbances in association with intervention cessation [
22]. Nonetheless, it could be postulated that the lack of association may be attributable to participant selection [
58], given that those with major mental health difficulties were excluded.
While it was hypothesized that higher primary care utilization prior to intervention enrolment [
36] would be associated with cessation and chronic stress, only higher non-usage of medication or supplements was associated with very early intervention cessation, especially for women and those with university education. This result emphasized further the complexity of pathways leading to unsuccessful intervention completion [
27,
72]. As hesitancy about lifestyle changes and their necessity may hinder participation [
24‐
26], participants without comorbidities requiring medication may perceive themselves at lower risk of adverse consequences from prediabetes, thus leading to a higher risk of cessation.
Elevated stress is considered to lower the likelihood of successful weight-loss and weight-loss maintenance in lifestyle change interventions [
18,
36]. Participants living with children and those economically not active due to, e.g. caring responsibilities, were least likely to report low chronic stress at the start of the intervention. Overall, as expected, higher chronic stress was associated with lower psychological QoL and higher mood disturbances, indicating that factors such as low self-esteem, negative feelings, and negative body image may have amplified chronic stress [
18]. Although men from lower socio-economic background have been reported to be particularly vulnerable to experience stress [
31], in here, especially women reported medium–high and high chronic stress.
From the results, it was notable that high chronic stress was associated with significantly lower physical health QoL especially for women. Physical health QoL encompasses concepts such as energy and fatigue, sleep and rest, and mobility. As stress has been associated with physical inactivity [
36] and the risk of weight gain [
32,
55], the current results suggested that lower physical health QoL at the start of the intervention may predispose especially women to higher chronic stress and thus to suboptimal weight-loss outcomes. Further, although lower SES has been associated with increased stress and consequently to worse weight-loss outcomes [
44‐
46], in the present study there was only the results indicated only limited influence of SES to chronic stress.
Targeted strategies are required to improve T2D-prevention especially in primary care settings [
37]. Success of behavior change interventions in T2D-prevention is based on complex interactions between participants and intervention [
26,
72]. Both intervention cessation and chronic stress at the start of the intervention are important determinants of successful weight-loss and weight-loss maintenance [
26,
28,
45]. Most differences were found between intervention completers and those who discontinued the intervention very early or early, and participants reporting high or low chronic stress. Lower environment al QoL and lack of family support for diet changes emerged as important predictors for cessation with women and those with higher SES especially affected. In turn, high chronic stress was predicted by higher mood disturbances and lower QoL for psychological and physical health, with, yet again, women more affected. Finally, the analyses indicated that the identified predictor variables were independent of participant BMI [
25], further highlighting the complexity of pathways that healthcare professionals need to consider in planning and delivering T2D-prevention interventions. For public health promotion, the results indicate that intervention developers and practitioners engaged in T2D-prevention need to consider how flexible intervention elements could be incorporated into the design and delivery to ensure better fit of varied participant`s needs.
There are numerous strengths associated with the study, particularly the large sample size. Nonetheless, the study is not without limitations. Specifically, participants were divided in the groups retrospectively, and it is recognized that different group divisions could have influenced the results. While logistic regression as an analysis method places few limitations on the data, stress and mood states were, as expected, correlated (
r = .07). Furthermore, associations between the outcome variables of chronic stress and cessation were not examined and additional work at this area would be needed. Also, degree of education was used as a measurement of SES [
5,
49,
56], and it can argued that other measurements e.g. incorporating income in to measurement of SES might have been more appropriate [
6], although not unproblematic in international research. In addition, no adjustments were made regarding different access to university education between countries. Number of predictor variables included in the analyses was also restricted, limiting ability to test different pathways. Finally, the interpretation of the results should be done carefully, as due to large number of participants, even small differences could produce statistically significant associations.
Acknowledgements
PREVIEW Study concept and design: Edith Feskens, Wageningen University, Netherlands.
PREMIT behaviour modification intervention: Daniela Kahlert and Annelie Unyi-Reicherz (University of Stuttgart, Germany).
The following contributors listed below assisted in conduct of the trial during recruitment, intervention and/or data collection:
University of Copenhagen, Denmark: Ulla Skovbæch Pedersen, Marianne Juhl Hansen, Bettina Belmann Mirasola, Maria Roed Andersen, Anne Wengler, Lene Stevner, Jane Jørgensen, Sofie Skov Frost, Eivind Bjørås, Grith Møller, Lone Vestergaard Nielsen.
University of Helsinki, Finland: Saara Grönholm (née Kettunen), Karoliina Himanen, Heini Hyvärinen, Martta Jalavisto (née Nieminen), Heidi Jokinen, Laura Kainu (née Korpipää), Pauliina Kokkonen, Liisi Korhonen, Tiia Kunnas, Elina Malkamäki, Pihla Mäkinen, Tuulia Onali (née Ingman), Tiina Pellinen, Kirsi Pietiläinen, Heli Pikkarainen, Sanna Ritola, Heikki Tikkanen, Sonja Toijonen, Jaana Valkeapää .
University of Nottingham, United Kingdom: Liz Simpson, Moira Taylor, Shelley Archer, Natalie Bailey-Flitter, Nicky Gilbert, Laura Helm, Sally Maitland, Melanie Marshall, Theresa Mellor, Grace Miller, Seodhna Murphy, Vicky Newman, Amy Postles, Jakki Pritchard, Maria Papageorgiou, Cheryl Percival, Clare Randall, Sue Smith, Sarah Skirrow.
University of Navarra, Spain: Blanca Martinez de Morentin Aldabe, María Hernández Ruiz de Eguilaz, Salomé Pérez Diez, Rodrigo San-Cristobal, Maria dels Angels Batlle, Laura Moreno-Galarraga, Alejandro Fernández-Montero, Marian Nuin, Javier Baquedano, Maria Eugenia Ursúa, Francisco Javier Martinez Jarauta, Pilar Buil, Lourdes Dorronsoro, Juana María Vizcay, Teodoro Durá-Travé, and all general practitioners and nurses from the Navarra Health Services who collaborated in the recruitment of the participants.
Medical University of Sofia, Bulgaria: Nadka Boyadjieva, Pavlina Gateva-Andreeva, Georgi Bogdanov, Galina Dobrevska.
University of Auckland, New Zealand: Amy Liu, Lindsay Plank, Anne-Thea McGill, Madhavi Bollineni, Clarence Vivar, Kelly Storey, Nicholas Gant, Jonathon Woodhead,
University of Sydney, Australia: Kylie Simpson, Michele Whittle, Kirstine Bell, Shannon Brodie, Jessica Burk.
We want to acknowledge all the additional people who have worked and are currently working for PREVIEW including trainees, post- and undergraduate students. Finally, a respectful thank you to all the study participants that participated in PREVIEW.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.