Long-term Impact of Behavioral Weight Loss Intervention on Cognitive Function

Abstract

Background.

It is unknown whether intentional weight loss provides long-term benefits for cognitive function.

Methods.

An ancillary study to a randomized controlled clinical trial was conducted in overweight and obese individuals (N = 978), aged 45–76 years at enrollment, with type 2 diabetes. An intensive behavioral intervention designed to promote and maintain weight loss through caloric restriction and increased physical activity was compared with diabetes support and education. Standardized assessments of cognitive function were collected an average of 8.1 years after trial enrollment.

Results.

Participants assigned to intensive lifestyle intervention lost a mean (SE) 11.1% (0.4%) and 7.2% (0.5%) of weight at Years 1 and 8, respectively, compared with 1.0% (0.2%) and 3.3% (0.5%) in the control group (p < .001). Covariate-adjusted mean composite cognitive function test scores were similar for the two groups (p = .69), and no significant differences were found for any individual cognitive test. There was some evidence of a differential effect (nominal interaction p = .008) for a prespecified comparison: Intensive lifestyle intervention was associated with a relative mean benefit for composite cognitive function of 0.276 (95% confidence interval: 0.033, 0.520) SDs among individuals with body mass index less than 30kg/m² at baseline compared with a relative mean deficit of 0.086 (−0.021, 0.194) SDs among individuals with body mass more than or equal to 30kg/m².

Conclusions.

Eight years of intensive lifestyle intervention did not alter cognitive function in obese adults with type 2 diabetes; however, there was evidence for benefit among overweight but not obese individuals. Changes in cognition were not assessed in this cross-sectional study.

There is considerable evidence that obesity in midlife increases the risk of cognitive decline and dementia (1–4). Later in life, relationships are less consistent, and some studies have found overweight, obesity, and central adiposity to protect against cognitive impairment and dementia in older individuals (5,6) and that lower body mass index (BMI) may be associated with poorer cognition (7) and greater cognitive decline (8).

Little is known about the long-term effects of weight loss on cognitive function. One meta-analysis of short-term (<2 years) interventions reports a modest protective effect among obese, but not overweight, individuals (9). However, midlife all-cause (ie, intentional or unintentional) weight changes in either direction, compared to weight stability, may be associated with increased risk of dementia (10). In later life, weight loss is linked to poorer cognitive function (11,12). This may be complicated in that unintentional weight loss is a marker for increased risk of dementia (8,13) and selection for obesity-related mortality may have occurred.

We describe findings from the first randomized controlled clinical trial of an intensive lifestyle intervention (ILI) that has sufficient size, duration, and adherence to examine whether lifestyle-induced intentional weight loss has long-term effects on cognitive function among overweight and obese individuals with type 2 diabetes, a condition that independently increases risk for cognitive deficits (14).

Methods

The Action for Health in Diabetes (Look AHEAD) recruited individuals with type 2 diabetes who were 45–76 years of age and had BMI more than or equal to 25kg/m² (≥27kg/m² if taking insulin), HbA1c less than 11% (97 mmol/mol), systolic blood pressure less than 160 mmHg, diastolic blood pressure less than 100 mmHg, and triglycerides less than 600mg/dL (15). They underwent a maximal graded exercise test and completed 2 weeks of self-monitoring prior to enrollment.

The Look AHEAD Movement and Memory Study (M&M) enrolled Look AHEAD participants at 4 (Baton Rouge, Denver, Memphis, and Pittsburgh) of its 16 centers to participate in an ancillary study to assess cognitive function at follow-up years 8 or 9. Only those who were currently active (ie, had not been lost to follow-up or refused further involvement) and who provided separate informed consent were eligible. The protocol and consent forms were approved by local Institutional Review Boards.

Interventions

When enrolled in Look AHEAD, individuals had been randomly assigned by center, with equal probability, to ILI or a Diabetes Support and Education (DSE) control condition. The ILI included diet modification and physical activity and was designed to induce at least a 10% weight loss at year 1 and to maintain this weight loss through subsequent years (16). ILI participants were assigned a calorie goal (1,200–1,800 based on initial weight), with less than 30% of total calories from fat (<10% from saturated fat) and a minimum of 15% of total calories from protein. The physical activity goal was more than or equal to 175 minutes per week through activities similar in intensity to brisk walking.

ILI participants were seen weekly for the first 6 months and 3 times per month for the next 6 months, with a combination of group and individual contacts. During years 2–4, participants were seen individually at least once a month, contacted another time each month by phone or e-mail, and offered a variety of centrally approved group classes. Subsequently, monthly meetings continued to be offered and annual campaigns were used to promote adherence.

DSE participants were invited to three group educational sessions each year (17). These used standardized curricula with focus on diet, physical activity, or social support. Information on behavioral strategies was not presented and participants were not weighed.

Cognitive Function

The following validated tests of cognitive functions, which have been demonstrated to be related to obesity, were used: Trail Making Test-Parts A and B (TMT-A, TMT-B) (18), Modified Stroop Color and Word Test (SCWT) (19,20), Digit Symbol-Coding Test (DSC) (21), Rey Auditory Verbal Learning Test (RAVLT) (22), and the Modified Mini-Mental State Exam (3MS) (23). These were administered at annual follow-up Visits 8 or 9 by centrally trained, masked examiners. Participants who were unable to visit clinics were administered the Telephone Interview for Cognitive Status (24), however because this measure of global cognitive function is not interchangeable with the 3MS clinic-based test (25), its data are not included.

Risk Factors for Cognitive Deficits

At Look AHEAD baseline, a number of risk factors for cognitive deficits were measured by certified staff and, self-reported characteristics and conditions were obtained using standardized questionnaires. The Short Form-36 Health Survey (SF-36) assessed health-related quality of life (26). Scores more than or equal to 11 on the Beck Depression Inventory (27) marked symptoms of mild-to-moderate depression. Blood pressure was measured using a Dinamap Monitor Pro100 automated device. A maximal graded exercise test was administered (28). Fitness was measured as the estimated metabolic equivalents: 4 metabolic equivalents approximates walking on flat ground at just under 4 miles per hour. Fasting blood specimens were analyzed centrally using standardized laboratory procedures. Weight was measured annually using a digital scale by certified, masked staff.

Statistical Analysis

Each cognitive test subscale was converted to z scores by dividing its difference from the mean by the cohort’s standard deviation and ordered, so that positive scores reflected better performance. A 5% winsorization was first applied to limit the effect of extremes: scores less than 5th percentile or more than 95th percentile were replaced with the 5th and 95th percentiles of the distribution. Global cognition was based on the total 3MS score. Verbal fluency was the based on the 3MS four-legged animals item. Immediate and delayed verbal memory scores were calculated from the RAVLT and averaged. Attention was based on TMT-A times. Executive function—set shifting was based on the TMT-B times and executive function—response inhibition was based on the SCWT interference score: Composite executive function was formed by averaging scores from these two tests. Processing speed was based on the DSC test score. Composite cognitive function was formed by averaging standardized scores for each tests and re-normalizing it to have unit standard deviation.

Analyses of covariance, with varying levels of covariate adjustment, were used to compare mean values between intervention groups. Tests of interaction were used to contrast prespecified subgroups based on gender and age and BMI at enrollment. Inverse probability weighting was used to assess the sensitivity of findings with respect to attrition that occurred between the initial randomization and time of cognitive data collection used in analyses (29).

Results

The four clinics had originally enrolled 1,331 participants into Look AHEAD. When M&M enrollment began, 22 had previously withdrawn, 65 had died, and 12 had follow-up curtailed other reasons, leaving 1,232 potential recruits. Of these, 1,092 (89%) consented to enroll in M&M, of which 1,081 (99%) were seen either at the clinic or assessed by telephone. We analyzed data from the 978 participants assessed at the clinic, 73% of the original Look AHEAD enrollees and 90% of those assessed by M&M. We examined whether rates of non-enrollment between intervention groups varied according to the Table 1 baseline risk factors for cognitive impairment: No interaction tests reached statistical significance (ie, all p > .05). We also assessed whether the balance afforded by the original randomization extended to this 73% subset (Table 1). The distribution of all risk factors was balanced between intervention groups (Table 1), except that DSE participants were slightly less likely to have elevated Beck Depression Inventory scores.

Table 1.

Characteristics at the Time of Enrollment Into the Look AHEAD (Action for Health in Diabetes) Trial of Participants From the Look AHEAD Movement and Memory Study Included in This Report: Frequency (%) or Mean (SD)

	Diabetes Support and Education (N = 479)	Intensive Lifestyle Intervention (N = 499)	p Value
Age
45–59	262 (54.7)	281 (56.3)	.61
60–76	217 (45.3)	218 (43.7)
Gender
Female	272 (56.8)	281 (56.3)	.88
Male	207 (43.2)	218 (43.7)
Race/ethnicity
African-American	96 (20.0)	100 (20.0)	.95
Non-Hispanic White	343 (71.6)	360 (72.1)
Other/multiple	40 (8.4)	39 (7.8)
Education (missing = 21)
HS or less	71 (15.3)	71 (14.4)	.19
Post–HS	205 (44.2)	194 (39.4)
College graduate	188 (40.5)	228 (46.2)
Body mass index (kg/m2)
25–29	69 (14.4)	95 (19.0)	.15
30–39	310 (64.7)	304 (60.9)
≥40	100 (20.9)	100 (20.0)
Weight (kg)	101.2 (18.4)	101.9 (19.0)	.53
Waist girth (cm) (missing = 1)	113.4 (14.4)	114.0 (15.2)	.55
Glycated hemoglobin (HbA1c) (%)
<7.0	231 (48.2)	235 (47.1)	.89
7.0–8.9	203 (42.4)	219 (43.9)
9.0–10.9	45 (9.4)	45 (9.0)
Insulin use (missing = 25)
No	406 (87.5)	419 (85.7)	.41
Yes	58 (12.5)	70 (14.3)
Diabetes duration (missing = 10)
<5 y	226 (47.8)	239 (48.3)	.88
≥5 y	247 (52.2)	256 (51.7)
Hypertension
No	88 (18.4)	83 (16.6)	.47
Yes	391 (81.6)	416 (83.4)
Prior cardiovascular disease
No	415 (86.6)	423 (84.8)	.40
Yes	64 (13.4)	76 (15.2)
Beck depression index (missing = 1)			.03
<11	440 (92.0)	439 (88.0)
≥11	38 (8.0)	60 (12.0)
Alcohol intake
None	323 (67.4)	328 (65.7)	.80
<1/d	130 (27.1)	145 (29.1)
≥1/d	26 (5.4)	26 (5.2)
Cigarette smoking
Never	258 (53.9)	244 (48.9)	.28
Former	200 (41.8)	233 (46.7)
Current	21 (4.4)	22 (4.4)
Fitness, METS	7.53 (2.08)	7.42 (1.98)	.39
Short Form-36 Health Survey (missing = 4)
General health	48.0 (8.2)	47.6 (8.8)	.41
Mental	54.4 (7.1)	54.4 (7.7)	.99
Pain	50.9 (8.8)	50.3 (8.7)	.20
Physical	47.8 (8.2)	47.3 (8.0)	.39
Social function	52.4 (7.0)	52.3 (7.5)	.68
Vitality	53.0 (8.2)	52.8 (9.1)	.67
Visit year
8	466 (97.3)	484 (97.0)	.78
9	13 (2.7)	15 (3.0)

Note: HS = higher secondary; METS = metabolic equivalents.

The ILI had produced substantial long-term changes in weight compared with DSE. The ILI group lost a mean (SE) of 11.1% (0.4%) of their weight at Year 1 and maintained a 7.2% (0.4%) mean loss through Year 8. At these times, 46.6% and 35.5%, respectively, of ILI participants had weight losses more than or equal to 10%. In contrast, the DSE group lost a mean of 0.9% (0.41%) at Year 1 and 3.2% (0.4%) at Year 8, and 3.6% and 21.6%, respectively, had more than or equal to 10% weight losses at these times.

Cognitive functions (Table 2) were assessed an average (range) of 8.1 (7.8–9.3) years after Look AHEAD baseline. Test scores were combined and standardized into six functions (global cognition, verbal fluency, verbal memory, attention, executive function, and processing speed) and a composite measure. Each had strong inverse associations with age (Table 3), with relative mean deficits of 0.17–0.66 SDs for each additional decade. All 95% confidence intervals for mean differences in standardized cognitive function scores between intervention groups (ILI minus DSE) included 0, both with limited and extensive covariate adjustment, indicating no significant difference between the groups. Supporting analyses using inverse probability weighting and propensity scores and to control for attrition each yielded comparable results (data not reported).

Table 2.

Raw Mean (SD) Winsorized Scores From Individual Tests for Participants Grouped by Intervention Assignment

Cognitive Test Score	Diabetes Support and Education	Intensive Lifestyle Intervention	p Value
Trail Making Test
Psychomotor speed: Part A (s)	36.28 (12.37)	35.87 (12.42)	.60
Tracking speed: Part B (s)
Without adjustment for psychomotor speed	102.6 (57.66)	100.8 (57.14)	.62
With adjustment for psychomotor speed (Part A s – Part B s)	62.85 (46.34)	62.87 (46.14)	.99
Stroop Color-Word Test
Interference (Part C-W time + errors)	38.47 (18.68)	39.17 (17.71)	.55
Digit Symbol-Coding
Processing speed (# correct)	41.14 (10.84)	41.61 (10.38)	.49
Rey Auditory Verbal Learning Test
Total learned (sum)	41.66 (9.04)	42.21 (8.64)	.33
Short delay recall	7.73 (2.96)	7.73 (3.08)	.99
Short-term retention (%)	72.73 (20.46)	71.78 (21.71)	.49
Long delay recall	7.30 (3.22)	7.51 (3.28)	.33
Long-term retention (%)	67.79 (23.34)	69.17 (23.61)	.36
Modified Mini-Mental State Examination
Total score	92.19 (6.66)	92.25 (5.94)	.88
Verbal fluency	9.92 (1.41)	9.23 (1.41)	.90

Table 3.

Mean Differences Between Diabetes Support and Education (DSE) and Intensive Lifestyle Intervention (ILI) Participants Expressed in Standard Deviation Units, With Varying Levels of Covariate Adjustment

Cognitive Function	Association With Current Age*	Mean (SE) Difference in Standard Deviation Units: ILI − DSE (p value)
	SD/y(SE) (pvalue)	Adjustment for Clinic, Age, Education, Race/Ethnicity, Gender, and Baseline BMI	Full Covariate Adjustment†
Global cognition	−0.041 (0.004) (<.001)	−0.027 (0.056) (.43)	−0.007 (0.057) (.90)
Verbal fluency	−0.017 (0.005) (<.001)	−0.011 (0.061) (.86)	0.003 (0.062) (.97)
Verbal memory
Short delay	−0.031 (0.005) (<.001)	−0.032 (0.061) (.60)	−0.035 (0.062) (.72)
Long delay	−0.036 (0.004) (<.001)	0.056 (0.058) (.34)	0.074 (0.059) (.21)
Verbal memory Composite	−0.042 (0.004) (<.001)	0.021 (0.058) (.72)	0.032 (0.058) (.58)
Attention	−0.056 (0.004) (<.001)	0.006 (0.056) (.91)	0.027 (0.057) (.64)
Executive function
Tracking	−0.049 (0.004) (<.001)	−0.016 (0.056) (.78)	0.009 (0.057) (.88)
Set shifting	−0.053 (0.004) (<.001)	−0.100 (0.059) (.08)	−0.081 (0.059) (.17)
Executive function composite	−0.059 (0.004) (<.001)	−0.054 (0.054) (.32)	−0.029 (0.055) (.60)
Processing speed	−0.064 (0.004) (<.001)	−0.000 (0.053) (.99)	0.032 (0.054) (.55)
Composite	−0.066 (0.004) (<.001)	−0.022 (0.050) (.67)	0.009 (0.051) (.86)

Notes: Missing categorical covariates are coded as a separate category. BMI = body mass index.

*Linear regression with adjustment for clinic, age, gender, education, race/ethnicity, and baseline BMI.

^†Clinic and, as listed in Table 1, age, gender, race/ethnicity, education, BMI, weight, waist girth, HbA1c, insulin use, diabetes duration, hypertension, prior cardiovascular disease, depressive symptoms, alcohol intake, cigarette smoking, fitness, and SF-36.

There was little evidence that an intervention effect varied by age or gender. The mean (95% confidence interval) intervention effect (ILI minus DSE) for the composite score was 0.043 (−0.098, 0.183) SD units for participants who enrolled in Look AHEAD when aged 46–59 years and −0.089 (−0.246, 0.067) SD units for those aged 60–76 years (interaction p = .22). The mean intervention effect for the composite score among women was −0.015 (−0.149, 0.119) SD units; among men, it was –0.026 (−0.179,0.127) SD units (interaction p = .92).

There was some evidence that the intervention differentially affected overweight compared with obese participants. Among those who were overweight but not obese at enrollment (ie, 25kg/m² ≥ BMI < 30kg/m²), ILI was associated with a mean benefit for composite cognitive function of 0.276 (0.033, 0.520) SDs, compared with a deficit of 0.086 (−0.021, 0.194) SDs among those who were obese (nominal p value for interaction: p = .008). Figure 1 presents point estimates and 95% confidence intervals for intervention effects for overweight and obese participants. Points estimates were positive (favoring ILI) for all individual functions among overweight participants, however only for processing speed did the confidence interval exclude 0.

Figure 1.

Means and 95% confidence intervals for differences (ILI minus DSE) in cognitive function scores for participants grouped at baseline as overweight (BMI < 30kg/m²) or obese (BMI ≥ 30kg/m²), with adjustment for clinic, age, education, gender, and race/ethnicity. The squares represent the mean values and the horizontal bars are the 95% confidence intervals.

We examined whether average percent changes in weight from baseline over years 1–4 (when intervention was most intensive) were related to cognitive function scores using linear regression with covariate adjustment for clinic, age, gender, race/ethnicity, education, and baseline BMI. Tests of interaction were used to assess whether slopes between cognitive function and weight loss varied between intervention groups. None reached statistical significance (all p > .15)—see online Supplementary Figure.

Discussion

Lack of an Overall Intervention Effect on Cognitive Function

The Look AHEAD intervention produced sustained relative differences in weight loss among individuals included in our analysis, resulting in a 4% mean difference at 8 years between intervention groups, which exceeded the 8-year 3% difference seen for the full Look AHEAD trial (30). It also increased physical activity and fitness relative to DSE over time (28) and produced relative benefits in many risk factors for cognitive decline: markers of diabetes control, blood pressure, lipids, depression, physical fitness, and sleep apnea (27,30–32). In addition, ILI increased the odds of diabetes remission (33). These effects varied in intensity and some were not sustained throughout 8 years, however each may be hypothesized to slow cognitive aging. Despite these potential benefits, no overall relative differences were seen in cognitive function at 8 years.

Several explanations may be offered. Others have reported that cognitive deficits associated with diabetes appear to occur early, perhaps as insulin resistance develops prior to the development of frank diabetes (34,35). Diabetes also induces decrements in brain volumes (36). It may be that these consequences of diabetes are insensitive to lifestyle changes. However, obesity is linked to cognitive function independently of diabetes, so that weight change may be expected to affect cognitive function through separate mechanisms (14). Perhaps greater lifestyle changes are necessary to improve cognition and any benefits of the achieved weight losses on cognitive function were too small to be detected. While the lack of a baseline cognitive assessment reduced the available power, post hoc power calculations indicate there was more than 80% power to detect mean differences in the composite cognitive measure of 0.143 SD units. The slope of the regression linking composite cognitive function with age was 0.066 SD/y: Thus, the study was powered to detect a mean difference roughly equivalent to 0.143/0.066 = 2.2 years of cognitive aging in the cohort. Across the average of 8.1 years since randomization, this translates to a 27% reduction in the rate of cognitive aging. Smaller relative differences may be important but have less overall clinical or practical utility. It may be that differential enrollment culled individuals with cognitive deficits, limiting the ability to identify differences. While our analysis cohort was 73% of the original randomized participants, it was balanced between intervention groups and with respect to many risk factors. A sensitivity analysis based on inverse probability weighting found no evidence that this affected the results. Furthermore, we saw no differences between intervention groups in the Telephone Interview for Cognitive Status collected by telephone interviews (data not shown), however the sample size for these provided limited statistical power. It is possible that any intervention effects occurred earlier and had waned by eight years, perhaps due to weight regain or cross-over effects, or that these may emerge later.

Consistency Across Subgroups

The M&M protocol specified subgroup comparisons based on baseline age, gender, and weight. Relationships that obesity has with cognitive function and brain structure may vary with age and gender (3,12,37), however, we saw little evidence that the intervention differentially affected cognition depending on these factors. There was some evidence that the intervention may have improved cognitive function among participants who were overweight (BMI < 30kg/m²) but not obese (BMI ≥ 30kg/m²). Among overweight participants, the largest estimated benefit was for processing speed, however trends for relative benefit were present for all functions with estimated mean effect sizes ranging from 0.14 to 0.40 SDs. If present, these may have important clinical implications in that even smaller relative differences in cognitive function have been observed to translate to significant differences in the risk of cognitive impairment and dementia (38). The baseline BMI of obese participants ranged from 30 to 58kg/m², with an average of 37kg/m². Thus, even in attaining the goal of a 10% weight loss, most of these individuals remained obese and none reached normal weight. Remaining overweight/obese from midlife into older age is associated with poorer cognitive function compared with individuals who were normal weight some time during this interval (39). The relationship with risk for cognitive impairment is nonlinear, with a marked increase at BMI less than or equal to 30kg/m² (40).

Our finding is contrary to a meta-analysis of smaller intentional weight loss studies (9), which found modest evidence of benefit for memory and executive function for obese, but not overweight, individuals, for studies lasting less than 2 years. It is possible that these differences may have been present earlier in Look AHEAD but had disappeared by 8 years.

Relationships With Weight Changes by Intervention Group

It is likely that both arms of the Look AHEAD trial included both individuals who intentionally and unintentionally lost weight, and all were encouraged to lose weight. If there was a greater prevalence of intentional weight loss among ILI participants, as might be expected, this did not appear to alter relationships between weight loss and cognitive function: These were reasonably similar in both groups.

Limitations

As eligible volunteers for a clinical trial of lifestyle intervention, the Look AHEAD cohort may not generalize to other populations. The lack of a baseline cognitive assessment prevented us from examining changes in cognition and whether any intervention effects differed by baseline level of cognitive function.

Summary

Overall, random assignment to 8 years of an intensive lifestyle weight loss intervention did not lead to relatively better levels of cognitive function in adults with type 2 diabetes. On the other hand, weight loss intervention appeared to pose no harm for cognition when prescribed to middle- and older-aged individuals with diabetes. There was evidence of benefit for individuals with BMI less than 30kg/m², however we cannot rule out that this may be a chance finding. Because cognitive function was assessed only at one time point, we are not able to draw any conclusions about changes in cognitive function or shorter term effects.

Supplementary Material

Supplementary material can be found at: http://biomedgerontology.oxfordjournals.org/

Funding

This work was supported by the National Institute on Aging, National Institutes of Health, Department of Health and Human Services (R-01AG033087-01 and R-01AG033087-04S1). The Action for Health in Diabetes is supported through the following cooperative agreements from the National Institutes of Health: DK57136, DK57149, DK56990, DK57177, DK57171, DK57151, DK57182, DK57131, DK57002, DK57078, DK57154, DK57178, DK57219, DK57008, DK57135, and DK56992. The following federal agencies have contributed support: National Institute of Diabetes and Digestive and Kidney Diseases; National Heart, Lung, and Blood Institute; National Institute of Nursing Research; National Center on Minority Health and Health Disparities; Office of Research on Women’s Health; the Centers for Disease Control and Prevention; and the Department of Veterans Affairs. This research was supported in part by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. The Indian Health Service (I.H.S.) provided personnel, medical oversight, and use of facilities. Additional support was received from the University of Colorado Health Sciences Center General Clinical Research Center (M01RR00051) and Clinical Nutrition Research Unit (P30 DK48520); the University of Tennessee at Memphis General Clinical Research Center (M01RR0021140); the University of Pittsburgh General Clinical Research Center (GCRC) (M01RR000056); the Clinical Translational Research Center (CTRC) funded by the Clinical & Translational Science Award (UL1 RR 024153) and NIH grant (DK 046204); and the Frederic C. Bartter General Clinical Research Center (M01RR01346). The following organizations have committed to make major contributions to Look AHEAD: FedEx Corporation; Health Management Resources; LifeScan, Inc., a Johnson & Johnson Company; OPTIFAST of Nestle HealthCare Nutrition, Inc.; Hoffmann-La Roche Inc.; Abbott Nutrition; and Slim-Fast Brand of Unilever North America.