Introduction
Depression is associated with about a 2-fold increased risk of Alzheimer's disease (AD), the most common form of dementia (Ownby, Crocco, Acevedo, John, & Loewenstein, Reference Ownby, Crocco, Acevedo, John and Loewenstein2006). Although depression often precedes dementia, there have been debates on whether depression is a modifiable risk factor for dementia or a preclinical/prodromal disease for it. Studies have shown that amyloid β (Aβ) deposition, the initial hallmark pathology of AD, begins decades before the symptom onset of AD (Jack et al., Reference Jack, Knopman, Jagust, Petersen, Weiner, Aisen and Trojanowski2013; Villemagne et al., Reference Villemagne, Burnham, Bourgeat, Brown, Ellis, Salvado and Masters2013). A large longitudinal cohort study followed the trajectory of depression before dementia over 28 years and revealed that the difference in depressive symptoms between those with and without dementia became apparent 11 years before the diagnosis of dementia, while depressive symptoms in the early phase of the study did not increase the risk of dementia, suggesting depression as a prodromal/preclinical feature of dementia rather than a modifiable risk factor for it (Singh-Manoux et al., Reference Singh-Manoux, Dugravot, Fournier, Abell, Ebmeier, Kivimäki and Sabia2017). Another longitudinal study showed that even depression was associated with an increased risk of dementia, this was accounted for by people developing dementia within 5 years of depression, and the use of antidepressants did not decrease the risk of dementia (Almeida, Hankey, Yeap, Golledge, & Flicker, Reference Almeida, Hankey, Yeap, Golledge and Flicker2017). Whereas, a study of people with mild cognitive impairment (MCI) and a history of depression showed that the use of antidepressants was associated with a delayed progression to AD (Bartels, Wagner, Wolfsgruber, Ehrenreich, & Schneider, Reference Bartels, Wagner, Wolfsgruber, Ehrenreich and Schneider2018).
Although the causal relationship between depression and dementia remains questionable, considering that at least 20% of people with dementia experience depression, clinicians should be concerned about the development of subsequent dementia, particularly in people with late-onset depression (Rubin, Reference Rubin2018). The manifestations of symptoms in people with depression are highly heterogeneous. Notably, previous studies have reported that among the depressive symptoms, appetite/weight loss was uniquely associated with an increased risk of neurocognitive deficit and dementia (Potter, McQuoid, & Steffens, Reference Potter, McQuoid and Steffens2015; Saha, Hatch, Hayden, Steffens, & Potter, Reference Saha, Hatch, Hayden, Steffens and Potter2016). However, among the general population, obesity is one of the risk factors for dementia; a meta-analysis study reported obesity, which is defined as a body mass index (BMI) ⩾30 kg/m2, to be associated with late-life dementia (Albanese et al., Reference Albanese, Launer, Egger, Prince, Giannakopoulos, Wolters and Egan2017). Moreover, weight loss of ⩾2 kg in people with BMI ⩾25 was associated with improvement in attention and memory (Veronese et al., Reference Veronese, Facchini, Stubbs, Luchini, Solmi, Manzato and Fontana2017). This suggests that weight changes in patients with depression may have a different effect on the development of dementia compared to the general population.
Meanwhile, the results of previous studies have suggested that the relationship between weight and dementia could be more complex. While obesity in middle age increases the risk of dementia in older age (Gustafson, Rothenberg, Blennow, Steen, & Skoog, Reference Gustafson, Rothenberg, Blennow, Steen and Skoog2003; Whitmer, Gunderson, Barrett-Connor, Quesenberry, & Yaffe, Reference Whitmer, Gunderson, Barrett-Connor, Quesenberry and Yaffe2005), obesity or overweight in late-life showed a protective effect for dementia (Atti et al., Reference Atti, Palmer, Volpato, Winblad, De Ronchi and Fratiglioni2008; Natale, Zhang, Hanes, & Clouston, Reference Natale, Zhang, Hanes and Clouston2023). In addition, there seems a sex difference in the effects of weight on the risk of dementia; the effect of obesity in middle age was more apparent in women (Goble, Reference Goble2005).
While multiple interaction effects are suggested and the underlying mechanisms are far from being understood, weight change, particularly in middle and old age, seems to be a predictor for future dementia. In addition, considering the increased risk of dementia in patients with depression and the characteristics of depression in that clinical types are distinguishable according to age at onset or cluster of symptoms, the weight change in them could give important information in predicting future dementia. However, data that analyzed the association between weight changes in people with depression and the subsequent development of dementia with consideration of other risk factors for dementia is still lacking. This study aimed to investigate the association between weight change before and after a diagnosis of depression with the subsequent risk of dementia using South Korean national cohort data.
Methods
Data source
We used the National Health Insurance Sharing Service (NHISS) database provided by the National Health Insurance Service (NHIS) of South Korea (Lee, Lee, Park, Shin, & Kim, Reference Lee, Lee, Park, Shin and Kim2017). NHIS is a government organization in charge of operating a mandatory universal health insurance service covering the South Korean population of over 50 million. The NHISS database involves medical service claims data including admissions, outpatient clinic visits, emergency room visits, pharmaceutical service uses, and health screening programs. The general health screening program of South Korea is one of the world's largest health screening programs and aims at the prevention and early intervention of common chronic diseases (Shin, Cho, Park, & Cho, Reference Shin, Cho, Park and Cho2022).
The data of NHISS are anonymized and the Institutional Review Board of Samsung Medical Center exempted this study from review because this study involved retrospective analyses of de-identified data (No. SMC 2022-05-118).
Identification of eligible subjects
Between 2010 and 2016, 4 805 617 were newly diagnosed with depression. The diagnosis of depression was defined as the F32 or F33 code of the International Statistical Classification of Disease and Related Health problems 10th revision (ICD-10). We extracted 1 308 730 eligible subjects ⩾40 years of age who received health screening programs within 2 years before the diagnosis of depression (first health screening) and within 2 years after diagnosis (second health screening), respectively, excluding those who had a previous diagnostic history of dementia between 2002 and the day of second health screening. We followed up on the incidence of dementia from the day of the second health screening program until December 2019 (Fig. 1).
Figure 1. Flow chart showing the identification of subjects.
Weight changes
Through the data from two health screening programs, we identified the subjects' weight changes from the first health screening (conducted within 2 years before the diagnosis of depression) to the second health screening (conducted within 2 years after the diagnosis of depression). And, depending on the proportion of weight change, we classified subjects into five groups: <-10%, −10 to −5%, −5 to 5%, 5 to 10%, and ⩾10%.
Outcomes
The main outcome was a new diagnosis of dementia based on the ICD-10 diagnostic code including AD (F00 or G30), VD (F01), and other dementia (F02, F03, or G31) and prescription of one or more dementia medications such as donepezil, galantamine, rivastigmine, and memantine. In South Korea, to file expense claims for the prescription of these medications, physicians need to document evidence of cognitive dysfunction according to National Health Insurance Reimbursement criteria: a Mini Mental State Examination (MMSE) (Baek, Kim, Park, & Kim, Reference Baek, Kim, Park and Kim2016) ⩽26 and either a Clinical Dementia Rating (CDR) (Berg, Reference Berg1984) ⩾1 or a Global Deterioration Scale (GDS) (Reisberg, Ferris, de Leon, & Crook, Reference Reisberg, Ferris, de Leon and Crook1982) ⩾3 (Yoo et al., Reference Yoo, Shin, Han, Kim, Lee, Jeong and Kim2020). When a subject had more than one code of dementia, the primary diagnostic code was used. If both AD and VD codes were present in the secondary diagnostic codes, the subject was classified based on the primary diagnostic code at the subsequent hospital visit, and if both codes remained as secondary diagnostic codes at that time, the subject was classified as the other dementias group.
Covariates
We identified subjects' baseline characteristics through the data obtained on the second health screening day. Low income was defined as less than the bottom 20% of the total premium of insurance payment, which is determined by income level. Subjects' body mass index (BMI), waist circumference, smoking status, alcohol drinking, and physical activity were also identified. Those who performed vigorous-intensity physical activity ⩾3 days or moderate-intensity physical activity ⩾5 days per week were considered to engage in regular physical activity (Pescatello, Riebe, & Thompson, Reference Pescatello, Riebe and Thompson2014). In addition, age, sex, previous medical history of diabetes (ICD-10 code of E11, E12, E13, or E14 and one or more prescriptions of antidiabetic medication or fasting glucose level ⩾126 mg/dl), hypertension (one or more prescriptions of antihypertensive agents under ICD-10 code of I10–13 or I15 or a systolic/diastolic blood pressure ⩾140/90 mmHg), dyslipidemia (one or more prescriptions of anti-hyperlipidemic agents under ICD-10 code of E78 or total cholesterol ⩾240 mg/dl), hearing loss (a history of disability determination) and traumatic brain injury (ICD-10 code of S06) were identified through medical records.
Statistical analysis
Baseline demographic and clinical characteristics were presented as mean ± standard deviation (s.d.) for continuous variables and numbers with percentages for categorical variables. An analysis of variance and a χ2 test was used to compare the variables according to the weight change group. The incidence rates of all-cause dementia, AD, and VD were calculated and presented per 1000 person-years.
Cox proportional hazard regression analyses were conducted to investigate the association between weight change before and after the diagnosis of depression with the occurrence of outcomes. Hazard ratios and corresponding 95% confidence intervals were presented. Model 1 was non-adjusted, Model 2 was adjusted for age and sex, Model 3 was adjusted for low income, hearing loss, traumatic brain injury, hypertension, diabetes mellitus, dyslipidemia, current alcohol drinking, current smoking, and regular physical activity in addition to Model 2, and Model 4 was adjusted for baseline BMI in addition to Model 3. In addition, subgroup analyses were conducted according to sex and age group, and the interaction effects were examined.
Results
Baseline characteristics
Table 1 shows the baseline characteristics of subjects by weight change group. The mean age of those with weight change of <-10%, −10 to −5%, −5 to 5%, 5 to 10%, and ⩾10% were 62.6 years (s.d., 12.2 years), 61.1 years (s.d., 11.2 years), 59.2 years (s.d., 10.6 years), 58.4 years (s.d., 10.9 years), and 58.8 years (s.d., 11.7 years). The proportion of females was highest in those with a weight change of <-10% (65.6%) and lowest in those with a weight change of −5 to 5% (58.8%). The BMI was lowest in those with a weight change of <-10% and increased as the weight change increased.
Table 1. Baseline characteristics of subjects by weight change
BMI, body mass index.
Incidence of dementia
The average follow-up period was 5.2 years (s.d., 2.0 years) (online Supplementary Table S1). 69 373 subjects were newly diagnosed with dementia with an incidence rate of 10.2 per 1000 person-years. Among them, 56 351 were diagnosed with AD with an incidence rate of 8.3 per 1000 person-years, and 6877 were diagnosed with VD with an incidence rate of 1.0 per 1000 person-years (online Supplementary Fig. S1).
Risks of dementia according to weight changes
Table 2 shows the HRs and corresponding 95% CIs of weight change groups on the risk of newly diagnosed dementia. Regarding all outcomes, compared to those with a minimal weight change (−5 to 5%), all groups with weight gain or loss showed increased risks of dementia, in all analysis models with a dose–response relationship, showing a U-shaped association.
Table 2. Hazard ratios and 95% confidence intervals of weight change on the occurrence of dementia
Model 1, non-adjusted; Model 2, adjusted for age and sex; Model 3, adjusted for age, sex, low income, hearing loss, traumatic brain injury, hypertension, diabetes mellitus, dyslipidemia, current alcohol drinking, current smoking, and regular physical activity; Model 4, adjusted for age, sex, low income, hearing loss, traumatic brain injury, hypertension, diabetes mellitus, dyslipidemia, current alcohol drinking, current smoking, regular physical activity, and baseline body mass index.
In the fully adjusted model, compared to those with weight change of −5 to 5%, those with weight loss (adjusted hazard ratio [aHR], 1.49 [95% CI 1.45–1.54] for <−10%; aHR, 1.22 [95% CI 1.20–1.25] for −10 to −5%) and those with weight gain (aHR, 1.17 [95% CI 1.14–1.20] for 5 to 10%; aHR, 1.47 [95% CI 1.42–1.53] for ⩾10%) showed increased risks of all dementia. Likewise, in terms of AD, compared to those with weight change of −5 to 5%, those with weight loss (aHR, 1.47 [95% CI 1.42–1.52] for <−10%; aHR, 1.21 [95% CI 1.18–1.24] for −10 to −5%) and those with weight gain (aHR, 1.15 [95% CI 1.11–1.18] for 5 to 10%; aHR, 1.44 [95% CI 1.39–1.51] for ⩾10%) showed increased risks of AD. And in terms of VD, compared to those with weight change of −5 to 5%, those with weight loss (aHR, 1.66 [95% CI 1.50–1.83] for <−10%; aHR, 1.32 [95% CI 1.24–1.42] for −10 to −5%) and those with weight gain (aHR, 1.24 [95% CI 1.14–1.34] for 5 to 10%; aHR, 1.77 [95% CI 1.58–1.98] for ⩾10%) showed increased risks of VD (Fig. 2 and online Supplementary Fig. S2).
Figure 2. Forest plots showing the hazard ratios and corresponding 95% confidence intervals of weight changes on the risk of the occurrence of dementia. Adjusted for age, sex, low income, hearing loss, traumatic brain injury, hypertension, diabetes mellitus, dyslipidemia, current alcohol drinking, current smoking, regular physical activity, and baseline body mass index.
Subgroup analyses
Table 3 shows the subgroup analyses according to sex and age groups. There were significant interaction effects between sex and weight change on the occurrence of all dementias (p-value for interaction = 0.013) and AD (p-value for interaction = 0.017). The increased risks of all dementia and AD by weight loss or gain were prominent in male subjects. There also were significant interaction effects between age groups and weight change on all dementias (p-value for interaction <0.001), AD (p-value for interaction <0.001), and VD (p-value for interaction = 0.003). The increased risks of all dementias, AD, and VD by weight loss or gain were greater when subjects were younger.
Table 3. Adjusted hazard ratios and 95% confidence intervals by weight change in subgroups according to sex and age group
IR, incidence rate; aHR, adjusted hazard ratio; CI, confidence interval.
* Per 1000 person-years.
† Adjusted for age, sex, low income, hearing loss, traumatic brain injury, hypertension, diabetes mellitus, dyslipidemia, current alcohol drinking, current smoking, regular physical activity, and baseline body mass index.
Online Supplementary Tables S2–S4 show the results according to age at onset of depression and the presence of recurrent depression. Overall, the pattern of outcome in each subgroup was similar to the main outcome. However, the risk of dementia according to weight change was more prominent in those who developed depression before the age of 60 compared to those with late onset (⩾60 years) (p-value for interaction <0.001).
Sensitivity analyses
Online Supplementary Tables S5 and S6 show the results among those with ages <55 years and ⩾55 years based on the index date, respectively. The occurrence of dementia was lower in those with age <55 years (4%) compared to those over 55 years old (26.6%). However, the pattern of dementia risks according to weight change was similar to the main outcome.
Discussion
In this population-based cohort study, we found a U-shaped association between weight change before and after the diagnosis of depression and the subsequent occurrence of dementia. Compared to those whose weight change was within 5%, those who gained or lost more than 5% had increased risks of all-cause dementia, AD, and VD, with a dose-response relationship. In addition, the lower the age at the onset of depression, the stronger the association between weight change and the risk of dementia.
To our knowledge, this study is the first to examine the association between weight changes after the diagnosis of depression and the risk of dementia, and it is difficult to directly compare the results with those of previous studies using the same methodology. However, a study that looked at the risk of dementia according to weight change in type 2 diabetes patients from the same data source showed that weight loss or weight gain of more than 10% increased the risk of dementia (34–38%), while the weight changes within 5–10% did not increase the risk of dementia (Nam et al., Reference Nam, Park, Han, Kim, Koh, Kim and Kwon2019). In comparison, in our study, the risk of dementia increased even with a smaller weight change, and the range of increases in risks was also larger.
While the underlying explanations for the association between weight change and the increased risk of dementia in patients with depression remain poorly understood, the severity of depression might affect the results. The severity of depression is known to be related to the subsequent risk of dementia (Chen et al., Reference Chen, Hu, Wei, Qin, McCracken and Copeland2008), and since a change in appetite or weight is a symptom of depression, more severe depression could cause larger changes in weight. Often, many patients experience decreased appetite and weight loss, but some patients, especially those with atypical features, experience weight/appetite gain as a reversed neurovegetative symptom (American Psychiatric Association, 2013). In the present study, both weight loss and gain were in a dose–response relationship with dementia, which may be mediated by the severity of depression. In addition, factors that can be induced by symptoms of depression and are risk factors for dementia, such as alcohol consumption, smoking, social isolation, and physical inactivity (Livingston et al., Reference Livingston, Huntley, Sommerlad, Ames, Ballard, Banerjee and Mukadam2020), might affect the results.
Regarding weight gain, obesity due to weight gain might increase the risk of dementia. Previous studies have consistently shown the result that obesity increases the risk of dementia (Albanese et al., Reference Albanese, Launer, Egger, Prince, Giannakopoulos, Wolters and Egan2017; Kivimäki et al., Reference Kivimäki, Luukkonen, Batty, Ferrie, Pentti, Nyberg and Jokela2018). Obesity contributes to incident cardiovascular risk factors (Powell-Wiley et al., Reference Powell-Wiley, Poirier, Burke, Després, Gordon-Larsen, Lavie and St-Onge2021), which can be a possible mechanism that increases the risk of VD. However, in the present study, the risk of dementia in Model 3 (in which baseline BMI was not adjusted) was similar to that in Model 4 (in which baseline BMI was adjusted), bolstering evidence that the weight change per se could be a risk factor for dementia independent of BMI. Regarding weight loss, our results are consistent with the results of a study conducted among elderly patients with depression, where the appetite and weight loss symptoms were associated with an increased risk of AD (HR of 1.69) and non-AD (HR of 2.10) (Saha et al., Reference Saha, Hatch, Hayden, Steffens and Potter2016). Dementia has been known as an important cause of weight loss (Shatenstein, Kergoat, & Nadon, Reference Shatenstein, Kergoat and Nadon2001), and previous studies suggest that weight loss may have appeared as a preclinical symptom of dementia through analysis of trajectory before the diagnosis of MCI or dementia (Guo et al., Reference Guo, Wang, Dove, Chen, Yuan, Bennett and Xu2022; Stewart et al., Reference Stewart, Masaki, Xue, Peila, Petrovitch, White and Launer2005). The volume changes in the mesial temporal cortex (Grundman, Corey-Bloom, Jernigan, Archibald, & Thal, Reference Grundman, Corey-Bloom, Jernigan, Archibald and Thal1996) and Apolipoprotein E (APOE) ε4 allele (Jefferson et al., Reference Jefferson, Burlinson, Burns, Mann, Pickering-Brown, Owen and Horan2001; Vanhanen et al., Reference Vanhanen, Kivipelto, Koivisto, Kuusisto, Mykkänen, Helkala and Laakso2001) have been suggested as possible mechanisms of the weight loss in dementia. Considering the maximum duration of the follow-up period of dementia from the index date in our study was less than 10 years, weight loss could be a preclinical symptom of dementia since the initial pathologic change of AD, the accumulation of Aβ is generally known to precede symptoms by 10–15 years (Drew, Reference Drew2018).
Several methodological issues limit interpretations of our findings. First, the diagnosis of depression was based only on diagnostic codes but not structured evaluation. Second, only those who had participated in consecutive health screening programs before and after the initial diagnosis of depression were included. Therefore, those included in the analyses were more likely to engage in health-promoting behavior or might have already experienced some physical or mental discomfort or symptoms than those excluded from the analyses due to a lack of health screening data. Third, the temporal relationship between the diagnosis of depression and weight change is not determined. Previous studies have suggested the bidirectional relationships between weight and depression (Emery, Finkel, Gatz, & Dahl Aslan, Reference Emery, Finkel, Gatz and Dahl Aslan2020) and between weight and cognitive change (Karlsson, Zhan, Gatz, Reynolds, & Dahl Aslan, Reference Karlsson, Zhan, Gatz, Reynolds and Dahl Aslan2021). Therefore, weight could have changed first, leading to subsequent depression, and then to increased risk of dementia, or depression could have occurred first, leading to subsequent weight change. Also, the results of our study do not provide information about weight changes over a longer period after diagnosis of depression. Fourth, although we excluded those who developed dementia before the diagnosis of depression, the pathology of dementia might have started earlier, which could have resulted in a reversed causal inference. Fifth, the treatment factors for depression were not included in the analysis. In particular, it is known that antidepressants have different profiles for weight gain depending on the drug (M et al., Reference M, Correll, Bobes, Cetkovich-Bakmas, Cohen, Asai and Leucht2011), but an analysis considering these characteristics has not been performed. Sixth, the analyses did not include the severity of depression, which, as mentioned above, is likely to mediate the association between weight change and dementia. Finally, the follow-up time was relatively short, which might be insufficient to observe the occurrence of dementia, especially at younger ages.
Despite these limitations, our study has strengths in that dementia was followed longitudinally in a large sample with depression, and its relevance was analyzed by comprehensively correcting for potential risk factors for dementia. In addition, we defined the occurrence of dementia through both the diagnostic codes and the prescription of dementia drugs. For the prescription of these drugs, evaluation through MMSE and CDR (or GDS) must be performed and the criteria must be met. Therefore, this definition could have reduced the overdiagnosis or misdiagnosis of dementia and reflected the occurrence of the outcome more accurately. Moreover, the association was consistent across all age groups. In particular, the risk of dementia according to weight change increased prominently at the young age group. While previous studies have focused on the association between late-onset depression and dementia (Almeida et al., Reference Almeida, Hankey, Yeap, Golledge and Flicker2017; Vilalta-Franch et al., Reference Vilalta-Franch, López-Pousa, Llinàs-Reglà, Calvó-Perxas, Merino-Aguado and Garre-Olmo2013), our study showed that weight change increased the risk of dementia in all age groups ⩾40 years. These suggest that weight change, which appears as a symptom of depression, can be a strong predictor of dementia.
In conclusion, both weight loss and gain after the diagnosis of depression were associated with an increased risk of all-cause dementia, AD, and VD with a dose-response relationship. Our findings suggest that weight change as a symptom of depression could be a predictor for the future development of dementia.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S0033291723003124
Data availability statement
Publicly available datasets were analyzed in this study. This data can be found here: https://nhiss.nhis.or.kr/.
Acknowledgements
This research was supported by the Bio & Medical Technology Development Program of the National Research Foundation (NRF) funded by the Ministry of Science & ICT (No. 2021M3A9E 4 080 784) and by the Korea Health Technology R&D Project through the Korea Health Industry Development Institute (KHIDI), sponsored by the Ministry of Health & Welfare, Republic of Korea (No. HR21C0885).
Author contributions
Hyewon Kim: conceptualization, methodology, writing – original draft, writing – review and editing. Jin Hyung Jung: formal analysis, writing – review & editing. Kyungdo Han: conceptualization, methodology, formal analysis, writing – review and editing. Hong Jin Jeon: conceptualization, project administration, methodology, writing – review and editing,
Competing interests
No disclosures to report.
