Identification of high-risk symptom cluster burden group among midlife peri-menopausal and post-menopausal women with metabolic syndrome using latent class growth analysis

Abstract

Background:

Midlife peri-menopausal and post-menopausal women with metabolic syndrome experience multiple co-occurring symptoms or symptom clusters, which often result in significant symptom cluster burden. While they are a high-risk symptom burden group, there are no studies that have focused on identifying symptom cluster trajectories in midlife peri-menopausal and post-menopausal women with metabolic syndrome.

Objectives:

The objectives were to identify meaningful subgroups of midlife peri-menopausal and post-menopausal women with metabolic syndrome based on their distinct symptom cluster burden trajectories, and to describe the demographic, social, and clinical characteristics of different symptom cluster burden subgroups.

Design:

This is a secondary data analysis using the longitudinal data from Study of Women’s Health Across the Nation.

Methods:

Multi-trajectory analysis using latent class growth analysis was conducted to join the different developmental trajectories of symptom clusters to identify meaningful subgroups and high-risk subgroup for greater symptom cluster burden over time. Then, descriptive statistics were used to explain the demographic characteristics of each symptom cluster trajectory subgroup, and bivariate analysis to examine the association between each symptom cluster trajectory subgroup and demographic characteristics.

Results:

A total of four classes were identified: Class 1 (low symptom cluster burden), Classes 2 and 3 (moderate symptom cluster burden), and Class 4 (high symptom cluster burden). Social support was a significant predictor of high symptom cluster burden subgroup and highlights the need to provide routine assessment.

Conclusion:

An understanding and appreciation for the different symptom cluster trajectory subgroups and their dynamic nature will assist clinicians to offer targeted and routine symptom cluster assessment and management in clinical settings.

Keywords

menopause midlife symptom women

Introduction

Metabolic syndrome is a constellation of metabolic risk factors, which increases an individual’s likelihood of developing chronic disease conditions, such as diabetes, cancer, and cardiovascular disease.¹ These metabolic risk factors include central obesity, hypertension, insulin resistance, and dyslipidaemia, among which at least three of the metabolic risk factors need to co-occur for a clinical diagnosis of metabolic syndrome to be made.² In the United States, an estimated 40–50 million adults are clinically diagnosed with metabolic syndrome and significantly affected by its health consequences.^2,3 The overall prevalence of metabolic syndrome is higher in women than all other genders.⁴ Several factors unique to women contribute to the high prevalence, which include pregnancy-related weight gain, hormonal contraceptive use, polycystic ovary syndrome, and transition from peri-menopause to post-menopause.^5,6 When stratified by age group, midlife women are the highest risk population to develop metabolic syndrome.⁷ Aging and menopause during midlife leads to adverse changes in their hormonal and lipid profiles, such as decreased level of oestrogen and increased level of progesterone.⁷ Such adverse changes cause accumulation of fat in the intra-abdominal and gluteo-femoral area, and reduce lean body mass muscle, which characterize the metabolic risk factors (e.g. central obesity, dyslipidaemia) for metabolic syndrome.⁸ These midlife peri-menopausal and post-menopausal women with metabolic syndrome experience multiple co-occurring symptoms or symptom clusters that often result in significant symptom burden because they experience symptoms associated with both menopause and metabolic syndrome concurrently.⁹

Midlife peri-menopausal and post-menopausal women experience distinct symptom trajectories over time.^10–12 For example, previous study identified four heterogeneous patterns of vasomotor symptom trajectories among midlife women that include onset early with decline after menopause, onset near the final menstrual period with later decline, onset early with persistently high frequency, and persistently low frequency.¹¹ Another longitudinal study found four distinct trajectories for sleep-related symptoms based on their prevalence in midlife women, such as low prevalence, moderate prevalence, increasing prevalence, and high prevalence trajectory group.¹⁰ In addition, a recent study examined the patterns of symptom clustering in midlife women and the probability of midlife women experiencing symptom clusters.¹³ While there is an established body of literature that supports the different symptom trajectory experiences among midlife peri-menopausal and post-menopausal women, no studies have focused on identifying symptom trajectories in midlife peri-menopausal and post-menopausal women with metabolic syndrome despite being a high-risk symptom burden group. Understanding their symptom trajectories over time will allow for future development of targeted interventions that will change the developmental trajectory of their symptom experience into poor patient outcomes. In addition, we need to further examine whether there are any subgroups who share similar symptom trajectories among midlife peri-menopausal and post-menopausal women since the diagnosis of metabolic syndrome and their associated characteristics. This new knowledge will allow the clinicians to understand the different levels of risk for symptom burden among subgroups of midlife peri-menopausal and post-menopausal women with metabolic syndrome, and prioritize the need for patient care in clinical settings.

Given that women do not experience these symptoms in isolation but in clusters of symptoms, one way of studying the co-occurring symptoms of midlife peri-menopausal and post-menopausal women with metabolic syndrome would be to study the symptom trajectory for each symptom. Previous study has found that midlife peri-menopausal and post-menopausal women with metabolic syndrome experienced three symptom clusters that include the psychological/somatic/sexual cluster, sleep/urinary cluster, and vasomotor/genital cluster.⁹ This study examined the number and types of symptom clusters among midlife women in peri-menopause and post-menopause together.⁹ A similar cross-sectional study of midlife women has identified symptom clusters among midlife women and found that having moderate to high severity of physical and psychological symptoms concurrently is associated with the risk of developing metabolic syndrome. Yet, this study does not examine how symptom clusters change over time since their initial diagnosis of metabolic syndrome for midlife women who already have a clinical diagnosis of metabolic syndrome.¹⁴ While there are conflicting results on how their symptom experience may differ by menopausal stage (e.g. peri-menopause versus post-menopause), there was a subsequent study that examined and compared the symptom network structure between midlife women in peri-menopause and post-menopause.⁹ With the use of network comparison test, the symptom networks were similar between the two groups in terms of global strength, network structure, and specific centrality measures.⁹ Thus, it is appropriate to use the identified symptom clusters from previous study to study symptom trajectories in midlife peri-menopausal and post-menopausal women with metabolic syndrome. Studying symptom trajectories within the dimension of symptom cluster provides an opportunity for us to examine symptom burden for each cluster using a symptom cluster composite score instead of examining each individual symptom. The identification of the high-risk subgroup for greater symptom cluster burden over time can provide opportunity for clinicians to deliver targeted, timely, and personalized symptom management interventions in clinical settings.^15,16 It may also help researchers to identify underlying mechanism for different symptom cluster experiences in midlife peri-menopausal and post-menopausal women with metabolic syndrome, and distinct characteristics associated with each subgroup.

Multi-trajectory analysis using latent class growth analysis (LCGA) has been used to analyse relationships among the developmental trajectories of multiple outcomes of interest that evolve simultaneously,¹⁷ identifies distinct joint trajectory patterns, and further examines the common correlates of these outcomes—all of which could provide an important basis for future development of universal interventions.^17,18 Given the high correlation among symptoms, the three symptom clusters identified in previous research are assumed to be distinct but related to each other, and experienced simultaneously by midlife peri-menopausal and post-menopausal women with metabolic syndrome.⁹ Yet, there are no studies that have identified joint trajectories of three symptom clusters together. The use of multi-trajectory analysis can join the different developmental trajectories of three symptom clusters to identify meaningful subgroups and high-risk subgroup for greater symptom cluster burden over time.¹⁹

Objective

The aims of this current study are to identify meaningful subgroups of midlife peri-menopausal and post-menopausal women with metabolic syndrome based on their distinct symptom cluster burden trajectories, and to describe the demographic, social, and clinical characteristics of different symptom cluster burden subgroups.

Design

This is a secondary data analysis using longitudinal data from baseline to visit 10 from the Study of Women’s Health Across the Nation (SWAN).

Methods

Description of the data set

The SWAN is an ongoing prospective cohort study that aims to examine the health and wellness of midlife women in the community across multiple domains (e.g. physical, biological, psychological, and social). The study includes midlife women from multiple ethnic groups across the United States. These women were initially assessed at baseline in 1996 and followed up annually until 2015–2016. Midlife women were eligible to participate at baseline study if they were: (1) 42–52 years of age, (2) had an intact uterus and at least one ovary, (3) had at least one menstrual cycle in the past 3 months, (4) did not use reproductive hormones in the past 3 months, and (5) self-identified as one of the following: White, African American, Hispanic, Japanese, and Chinese. A total of 3302 participants met the eligibility criteria and were enrolled at baseline. More details on the SWAN study can be find elsewhere.²⁰

Participants

The eligibility criteria for the current study was (1) self-reported to be in peri-menopause or post-menopause, (2) met diagnostic criteria for metabolic syndrome at any point from baseline, visit 1, visit 3, visit 5, visit 7, and visit 9, based on the National Cholesterol Education Program Adult Treatment III guidelines. The variables required for the clinical diagnosis of metabolic syndrome were collected only during these visits. We excluded women in pre-menopause or not meeting diagnostic criteria for metabolic syndrome at any time point. A total of 557 participants met the eligibility criteria for midlife peri-menopausal and post-menopausal women with metabolic syndrome. It has been recommended that a minimum sample size of 200 is required when the number of latent classes is 2 and a sample size of 1000 when the number of latent classes is close to 5.²¹ However, previous studies using LCGA have used 100 sample for four latent classes.^22,23 Given the small number of study variables (three symptom clusters), we believe that the current sample size is sufficient to conduct statistical analyses.

Measures

Demographic and clinical characteristics

Demographic characteristics were obtained from self-report using the SWAN-designed questionnaire. Demographic characteristics included age, race/ethnicity, level of education, marital status, current employment status, annual household income, degree of difficulty paying for basics, and level of social support. Social support was measured on a 5-point Likert-type scale using four items from the Medical Outcome Study Social Support Survey.²⁴ Clinical characteristics included health perception, number of comorbidities, menopausal status, and body mass index (BMI). Health perception was measured using a single question from the Short Form (SF)-36 as a part of the Medical Outcome Study.²⁴ Menopausal status was obtained from self-reported menstrual bleeding patterns in the past 12 months. A woman is considered to be in peri-menopause when she has had menstrual period in the past 3 months with change in cycle length or regularity within the past 12 months or no menstrual period in the past 3 months with intermittent menstrual bleeding within the past 12 months. A woman is considered to be in post-menopause when she has had no menstrual period in the past 12 months.^25,26 BMI was derived from the height and weight measured during each visit.²⁷ We used the demographic and clinical characteristics at the time of initial diagnosis of metabolic syndrome for all participants.

Metabolic syndrome characteristics

The SWAN study used standardized study protocols to measure height, weight, waist circumference, blood pressure, and fasting blood work—all of which are needed for clinical diagnosis of metabolic syndrome. Height and weight were measured without shoes using the stadiometer. Waist circumference was measured at the umbilicus. Blood pressure was measured twice with a 2-min interval using standard mercury sphygmomanometers and these two blood pressure values were averaged. Fasting blood work was conducted in the morning after fasting for at least 8 h to obtain values for blood glucose, triglyceride, high-density lipoprotein (HDL) cholesterol. A clinical diagnosis of metabolic syndrome was made when an individual has three or more following metabolic risk factors at any time point: waist circumference over 35 inches, blood pressure over 130/85 mmHg, fasting triglyceride level over 150 mg/dl, fasting HDL cholesterol level less than 50 mg/dl, and fasting blood glucose over 100 mg/dl.²

Symptom clusters and cluster composite score

The types of symptom clusters were selected based on our earlier work which reported that midlife peri-menopausal and post-menopausal women with metabolic syndrome experienced the psychological/somatic/sexual cluster, sleep/urinary cluster, and vasomotor/genital cluster based on symptom severity dimension. Psychological/somatic/sexual cluster included anxiety, depression, frequent mood change, forgetfulness, stiffness, or soreness in joints, neck, or shoulder, and sexual disturbance. Sleep/urinary cluster included sleep disturbance and getting up from sleep to urinate. Vasomotor/genital cluster included cold sweat, night sweat, hot flash, and vaginal dryness.⁹ Some symptoms, such as sexual disturbance and getting up from sleep to urinate, were not measured at all visits. A composite symptom cluster score was derived for each symptom cluster that represents the symptom burden, with scores ranging from 0 to 3 (0 = none, 1 = mild, 2 = moderate, and 3 = severe). Missing data were considered to be missing at random (MAR) and multiple imputation was conducted to create values for sexual disturbance and getting up from sleep using the complete values from other related symptoms in the same symptom cluster.²⁸

Ethics approval and consent to participant

The SWAN study database, codebook, and survey questionnaires are currently available for public access from the Inter-university Consortium for Political and Social Research (ICPSR) website.²⁹ All the materials were downloaded to a secure, encrypted server at Duke University. Only the de-identified information was archived and analysed. All data sets, corresponding codebooks, and statistical programs were stored on a secure, encrypted server at Duke University. The current study received Duke University institutional review board declaration of exemption (Pro00106232). Informed consent was not obtained from the current study because this was a secondary data analysis using the SWAN data set and does not contain any direct interaction with human participants by any of the authors.

Statistical analyses

Data analysis was conducted using SAS 9.4 (SAS Ins., Cary, NC, USA). The SWAN data were panel survey data that were collected from baseline to visit 10. To study each participant’s symptom trajectory after their initial diagnosis of metabolic syndrome, we first realigned all participant data for the starting time to be their years since initial diagnosis. As a result, all participants had the same starting value of 0 year since initial diagnosis. After the realignment, some participants had fewer or more time points than others in the analysis data depending on when they were initially diagnosed with metabolic syndrome. In other words, a participant had more time points if she had an initial diagnosis of metabolic syndrome early, for example, at baseline. Contrastingly, a participant had fewer time points if she had an initial diagnosis of metabolic syndrome later on, for example, at visit 7. Then, a multi-trajectory modelling using LCGA was used to explore distinct patterns of symptom cluster burden trajectory in midlife peri-menopausal and post-menopausal women with metabolic syndrome. This approach assigns class membership in trajectory classes across all three symptom clusters while considering the longitudinal change over time.¹⁹ In addition, the multi-trajectory modelling is designed to measure the linkages between the trajectories of three distinct but related outcomes and to obtain the joint estimation of trajectory models.¹⁷ The model was built in two stages. The first stage was an exploratory step to learn the overall trend (linear and quadratic) of the trajectory of each symptom cluster composite score which helped build the final multi-trajectory model in the second stage. Using the information from the first stage, the multi-trajectory modelling was conducted to jointly model the trajectories for three symptom clusters composite scores together. This allowed us to examine three symptom clusters together, instead of each symptom cluster separately.

Stage 1: exploring symptom cluster composite score trajectories

Each symptom cluster composite score was continuous with approximately normal distributions and was modelled individually using the censored norm (CNORM) model in SAS Proc Traj. First, we examined the empirical summary plot to illustrate the trajectories of three symptom clusters: the psychological/somatic/sexual cluster, sleep/urinary cluster, and vasomotor/genital cluster. Linear and quadratic trend models were tested and parameter estimates were obtained for each symptom cluster. Statistical fit indices, such as Akaike information criterion (AIC) and Bayesian information criterion (BIC) were considered to determine the number of symptom cluster trajectory subgroups. The statistical fit of the model was tested for two-, three-, four-, five-class models. The final model was chosen based on the statistical fit indices, clinical interpretability of each symptom cluster trajectory subgroup, and clinical judgment of the authors. Then, the conditional probabilities linking latent class membership across the symptom cluster trajectory subgroups were calculated. Based on the conditional probabilities, there was a high correlation between the vasomotor/genital cluster and psychological/somatic/sexual cluster that led us to conduct Stage 2 and to model multi-trajectories based on the three symptom clusters together.

Stage 2: modelling multi-trajectories of three symptom clusters

From Stage 1, we decided that a quadratic trend model is suitable for each symptom cluster. Thus, parameter estimates for quadratic trend model (e.g. intercept, slope, quadratic) were obtained for each symptom cluster and used as starting values for the multi-trajectory model. The multi-trajectory model of three symptom clusters was tested for two-, three-, four-, and five-class models. All equal conditional group percentages were assumed based on the high correlation among the symptom clusters and thus each group of conditional group percentages added up to 100. Then, each multi-trajectory model was fitted with linear and quadratic trend. Statistical fit indices, such as AIC, BIC, clinical interpretability of each symptom cluster trajectory subgroup, and clinical judgment of the authors were used to select the optimal number of classes.

After the two-stage group-based trajectory modelling, descriptive statistics was used to explain the participant demographic characteristics of each symptom cluster trajectory subgroup at the time of initial diagnosis of metabolic syndrome. Then, bivariate analysis (analysis of variance, Kruskal–Wallis test) was conducted to examine the association between each symptom cluster trajectory subgroup and demographic characteristics.

Results

Characteristics of study sample

Our total study sample included 557 participants with a mean age of 45.76 years. The highest percentage of race/ethnicity was White (49.25%), followed by African American (29.89%), Hispanic (10.90%), Japanese (5.83%), and Chinese (4.14%). More than a quarter of the participants received high school education (28.16%) and some college education (30.32%). The majority were currently working, with annual household income US$20,000–US$49,999 (27.65%) and US$50,000–US$99,999 (37.70%),and no difficulty paying for basics (56.62%). The mean time since initial diagnosis of metabolic syndrome was 3.30 years and the mean BMI was 28.06 kg/m² in the overweight range. Table 1 further details the participant characteristics.

Table 1.

Characteristics of study sample.

	Total(n = 557)
Age, mean (SD)	45.79 (2.66)
Race/ethnicity
White	262 (49.25)
African American	159 (29.89)
Chinese	22 (4.14)
Japanese	31 (5.83)
Hispanic	58 (10.90)
Education
Less than high school	48 (8.66)
High school graduate	156 (28.16)
Some college	168 (30.32)
College graduate	89 (16.06)
Post-graduate	93 (16.79)
Marital status
Single, never married	63 (11.39)
Married	358 (64.74)
Separated	113 (20.43)
Widowed	19 (3.44)
Currently working	434 (78.34)
Annual household income
Less than US$19,999	87 (15.62)
US$20,000–US$49,999	154 (27.65)
US$50,000–US$99,999	210 (37.70)
US$100,000 or more	87 (15.62)
Refused	19 (3.41)
Difficulty paying for basics
Not hard at all	312 (56.62)
Somewhat hard	170 (30.85)
Very hard	69 (12.52)
Social support
None of the time	12 (2.15)
A little of the time	25 (4.49)
Some of the time	61 (10.95)
Most of the time	190 (34.11)
All of the time	269 (48.29)
Health perception
Poor	17 (3.06)
Fair	91 (16.40)
Good	185 (33.33)
Very good	174 (31.35)
Excellent	88 (15.86)
Comorbidity
0	181 (32.50)
1–2	311 (55.83)
>2	65 (11.67)
Menopausal status
Peri-menopause	242 (44.08)
Post-menopause	307 (55.92)
Time since initial diagnosis of metabolic syndrome, mean (SD)	3.30 (3.83)
Body mass index in kg/m², mean (SD)	28.06 (6.82)

Empirical summary plot

The empirical summary plot presents the overall trend of trajectory for each symptom cluster across all the participants (Figure 1). Based on the empirical summary plot, the severity for psychological/somatic/sexual cluster drops temporarily until year 1, increases until year 7, and then remains constant over time. In contrast, the severity for both sleep/urinary cluster and vasomotor/genital cluster fluctuates over time.

Figure 1.

Empirical summary plot for symptom clusters.

Multi-trajectory model results

A four-class model with quadratic trend for multi-trajectory model of three symptom clusters was selected based on the statistical fit indices (AIC =-10897.16, BIC = -11055.25), clinical interpretability, and clinical judgment of the authors (Table 2). While the five-class model had the lowest AIC and BIC, the identified trajectory subgroups were not clinically interpretable or distinct from each other. The Voung–Lo–Mendell–Rubin (VLMR) likelihood ratio test and Lo–Mendell–Rubin adjusted LRT (LMR-LRT), both indices that test whether the current model (n) is better than the previous model (n–1), was non-significant for all models but was close to reaching statistical significance for the four-class model. Thus, it shows that four-class model is highly likely to be better than the three-class model, and thus, we chose the four-class as the final model.

Table 2.

Multi-trajectory model fit by number of classes.

Number of classes	AIC	BIC	Log Bayes factor	Log-likelihood	BLRT	VLMR	LMR-LRT	Entropy	% per class
2	−11,717.21	−11,800.22	N/A	−11,695.21	N/A	N/A	N/A	0.890	66.0%, 34.0%
3	−11,164.90	−11,285.35	1029.74	−11,132.90	< .001	0.2209	0.2233	0.834	38.1%, 44.9%, 17.0%
4	−10,897.16	−11,055.25	460.2	−10,855.16	< .001	0.0661	0.0668	0.857	32.5%, 25.5%, 25.6%, 16.4%
5	−10,690.64	−10,886.38	337.74	−10,638.64	< .001	0.7070	0.7079	0.862	32.7%, 21.6%, 16.4%, 14.4%, 14.8%

AIC: Akaike information criterion; BIC: Bayesian information criterion; VLMR: Voung–Lo–Mendell–Rubin likelihood ratio test; LMR-LRT: Lo–Mendell–Rubin-adjusted likelihood ratio test.

For the psychological/somatic/sexual cluster (Figure 2), Class 1 was the consistently low symptom burden subgroup (32.5%) with an initial decrease in symptom severity and no significant changes over time. Class 2 was the moderately high symptom burden subgroup (25.5%) and Class 3 in the moderately low symptom burden subgroup (25.6%). When comparing Classes 2 and 3, they showed a similar pattern, but Class 2 remained consistently higher. Class 4 was the consistently severe symptom burden subgroup that was significantly higher than all other classes for most of the time. For the sleep/urinary cluster (Figure 2), Class 1 was the consistently low symptom burden subgroup (32.5%) given its overall trend. In contrast to the psychological/somatic/sexual cluster, Class 2 was the moderately low symptom burden subgroup with increasing trend and Class 3 was the moderately high symptom burden subgroup with consistent trend. Class 4 was the severe symptom burden subgroup with fluctuating trend over time. For the vasomotor/genital cluster (Figure 2), Classes 1 and 2 were the low symptom burden subgroups that were interactive with each other. Herein, interactive relationship indicates that the symptom clusters show a similar trend over time and are related to each other. Class 3 was the moderate symptom burden subgroup with an increasing trend and Class 4 was the severe symptom burden subgroup with an initial sudden drop in its severity and a significant quadratic trend (β = –0.004, p = 0.017, Table 3). Similarly, Classes 3 and 4 had an interactive relationship with each other.

Figure 2.

Multi-trajectory models: (a) psychological/somatic/sexual cluster; (b) sleep/urinary cluster; (c) vasomotor/genital cluster. Herein, X-axis (T) indicates years since initial diagnosis of metabolic syndrome and Y-axis (outcome) indicates symptom composite score.

Table 3.

Latent class growth analysis multi-trajectory model results.

Types of symptom cluster	Trajectory subgroup	Parameter	β	SE	t-value	p-value
Group membership	Class 1		35.257	2.523	12.892	< 0.001*
	Class 2		25.499	2.515	10.140	< 0.001*
	Class 3		25.623	2.452	10.449	< 0.001*
	Class 4		16.351	1.727	9.469	< 0.001*
Psychological/somatic/sexual cluster, sleep/urinary cluster, vasomotor/genital cluster	Class 1	Intercept	0.647	0.108	5.981	< 0.001*
		Linear	–0.008	0.022	-0.379	0.704
		Quadratic	0.001	0.001	1.254	0.210
	Class 2	Intercept	0.706	0.163	4.334	< 0.001*
		Linear	0.087	0.032	2.755	0.006*
		Quadratic	–0.002	0.001	-1.570	0.116
	Class 3	Intercept	1.295	0.102	12.690	< 0.001*
		Linear	0.044	0.022	1.998	0.046*
		Quadratic	–0.002	0.001	–1.731	0.084
	Class 4	Intercept	1.244	0.182	6.835	< 0.001*
		Linear	0.106	0.036	2.960	0.003*
		Quadratic	–0.004	0.002	–3.249	0.017*

p < 0.05.

Individual characteristics of multi-trajectory subgroups

When comparing the four classes, Class 4 had the highest mean age of 46.09 years and Class 2 had the lowest mean age of 45.73 years. There was the highest percentage of African American (34.11%) and Japanese (6.98%) in Class 2, Chinese (5.93%) and Hispanic (12.59%) in Class 3, and White (52.69%) in Class 4. In relation to the level of education, Class 1 had the highest percentage of participants with high school graduate degree or less (42.31%) and Class 4 with college graduate degree or beyond (39.14%). Majority of the participants (> 70%) were currently working across all classes with no difficulty or somewhat difficulty paying for basics. In addition, Class 1 had the highest percentage of participants in peri-menopause (51.96%), and Class 3 (61.54%) and Class 4 (61.29%) in post-menopause. However, all of these differences in their demographic characteristics were not statistically significant. A statistically significant difference was examined in the level of social support (p = 0.02) in which Class 4 had the highest percentage of receiving social support none of the time (6.45%) to a little of the time (7.53%). In contrast, Class 1 had the highest combined percentage of receiving social support most of the time (33.15%) to all of the time (51.09%). Table 4 further details the participant characteristics for the four classes.

Table 4.

Characteristics for each multi-trajectory group.

	Comparisons among latent classes, N (%)
	Class 1 (n = 184)	Class 2 (n = 135)	Class 3 (n = 145)	Class 4 (n = 93)	p-value
Age, mean (SD)	45.82 (2.64)	45.73 (2.75)	45.61 (2.53)	46.09 (2.76)	0.58
Race/ethnicity					0.89
White	89 (50.86)	60 (46.51)	64 (47.41)	49 (52.69)
African American	48 (27.73)	44 (34.11)	38 (28.15)	29 (31.18)
Chinese	8 (4.57)	4 (3.10)	8 (5.93)	2 (2.15)
Japanese	11 (6.29)	9 (6.98)	8 (5.93)	3 (3.23)
Hispanic	19 (10.86)	12 (9.30)	17 (12.59)	10 (10.75)
Education					0.10
Less than high school	14 (7.69)	13 (9.63)	14 (9.66)	7 (7.61)
High school graduate	63 (34.62)	28 (20.74)	43 (29.66)	22 (23.91)
Some college	44 (24.18)	49 (36.30)	48 (33.10)	27 (29.35)
College graduate	36 (19.78)	21 (15.56)	14 (9.66)	18 (19.57)
Post-graduate	25 (13.74)	24 (17.78)	26 (17.93)	18 (19.57)
Marital status					0.92
Single, never married	16 (8.74)	18 (13.33)	19 (13.19)	10 (10.99)
Married	123 (67.21)	81 (60.00)	93 (64.58)	61 (67.03)
Separated	37 (20.22)	31 (22.96)	27 (18.75)	18 (19.78)
Widowed	7 (3.83)	5 (3.70)	5 (3.47)	2 (2.20)
Currently working	144 (79.12)	105 (77.78)	115 (79.31)	70 (76.09)	0.93
Annual household income					0.43
Less than US$19,999	31 (16.85)	18 (13.33)	22 (15.17)	16 (17.20)
US$20,000–US$49,999	49 (26.63)	38 (28.15)	44 (30.34)	23 (24.73)
US$50,000–US US$99,999	71 (38.59)	54 (40.00)	52 (35.86)	33 (35.48)
US$100,000 or more	26 (14.13)	21 (15.56)	23 (15.86)	17 (18.28)
Refused	7 (3.80)	4 (2.96)	4 (2.76)	4 (4.30)
Difficulty paying for basics					0.45
Not hard at all	107 (58.79)	68 (51.13)	77 (53.85)	60 (64.52)
Somewhat hard	53 (29.12)	45 (33.83)	46 (32.17)	26 (27.96)
Very hard	22 (12.09)	20 (15.04)	20 (13.99)	7 (7.53)
Social support					0.02*
None of the time	2 (1.09)	3 (2.22)	1 (0.69)	6 (6.45)
A little of the time	9 (4.89)	5 (3.70)	4 (2.76)	7 (7.53)
Some of the time	18 (9.78)	17 (12.59)	20 (13.79)	6 (6.45)
Most of the time	61 (33.15)	48 (35.56)	59 (40.69)	22 (23.66)
All of the time	94 (51.09)	62 (45.93)	61 (42.07)	52 (55.91)
Health perception					0.37
Poor	7 (3.83)	3 (2.22)	7 (4.86)	0 (0.00)
Fair	31 (16.94)	22 (16.30)	27 (18.75)	11 (11.83)
Good	57 (31.15)	48 (35.56)	52 (36.11)	28 (30.11)
Very good	61 (33.33)	38 (28.15)	40 (27.78)	35 (37.63)
Excellent	27 (14.75)	24 (17.78)	18 (12.50)	19 (20.43)
Comorbidity					0.83
0	65 (35.33)	42 (31.11)	44 (30.34)	30 (32.26)
1–2	101 (54.89)	73 (54.07)	85 (58.62)	52 (55.91)
>2	18 (9.78)	20 (14.81)	16 (11.03)	11 (11.83)
Menopausal status					0.12
Peri-menopause	93 (51.96)	58 (43.28)	55 (38.46)	36 (38.71)
Post-menopause	86 (48.04)	76 (56.72)	88 (61.54)	57 (61.29)
Body mass index in kg/m², mean (SD)	28.48 (6.54)	27.57 (7.49)	27.90 (6.59)	28.20 (6.76)	0.69

Note. *p < 0.05.

High symptom cluster burden group

Class 4 remained high across all three symptom clusters. The mean age was 46.09 years with majority participants being White (52.69%) or African American (31.18%). There was a similar percentage of participants with high school degree (23.91%), some college degree (29.35%), college graduate degree (19.57%), and post-graduate degree (19.57%). More than half were married (67.03%) with the other half separated (19.78%), single/never married (10.99%), or widowed (2.20%). A majority was currently working (76.09%) with annual household income US$50,000–US$99,999 (35.48%). While half of the participants (55.91%) reported having social support all of the time, there was the highest percentage of participants receiving social support none of the time (6.45%) to a little of the time (7.53%). These participants had good (30.11%) to excellent (20.43%) perception of their health with one to two health comorbidities (55.91%). There were more participants in post-menopause (61.29%) than peri-menopause (61.29%). The mean time since initial diagnosis of metabolic syndrome was the longest (3.72 years) and the mean BMI was 28.20 kg/m² in the overweight range.

Discussion

Previous studies have found different trajectory subgroups based on an outcome of interest (e.g. health behaviours) in various populations and the associated characteristics of each trajectory subgroup.³⁰ Similar to their approach, our study used LCGA but with multi-trajectory modelling approach to identify homogeneous subpopulations based on symptom cluster trajectories within the heterogeneous population of midlife peri-menopausal and post-menopausal women with metabolic syndrome, which is critical in understanding the similarities and differences in their individual characteristics of homogeneous subpopulations based on their similar symptom trajectories over time. A total of four classes were identified (Classes 1–4), which showed distinct and interactive symptom cluster trajectories over time. Our study findings provide a critical knowledge basis for clinicians to identify high-risk symptom cluster burden group and to effectively manage their symptom clusters through targeted symptom assessment and management in clinical settings.

Overall, there was a temporary sharp decrease or increase in symptom cluster severity after initial diagnosis of metabolic syndrome for the identified classes across three symptom clusters. The diagnosis of metabolic syndrome may be an additional psychological burden for some people and can negatively influence a healthy lifestyle and development of negative symptoms related to lifestyle, such as feeling of loneliness and cognitive decline.^31,32 This may further explain the sharp increase in symptom cluster severity shortly after the initial diagnosis of metabolic syndrome among midlife peri-menopausal and post-menopausal women. In contrast, there is an established literature that lifestyle modifications are effective in resolving or reducing the severity of metabolic syndrome.^33–35 Such lifestyle modifications have been used in various patient population, which has shown to improve symptoms and quality of life among patients with irritable bowel syndrome.³⁶ It may be possible that some midlife women adapted lifestyle modifications after their initial diagnosis of metabolic syndrome that led to sharp decrease in symptom cluster severity.

Our study identified four classes based on the psychological/somatic/sexual cluster, sleep/urinary cluster, and vasomotor/genital cluster. The severity trend for Class 1 (low symptom burden) and Class 4 (high symptom burden) remained relatively stable across all three symptom clusters. However, the severity trend for Classes 2 and 3 (both moderate symptom burden) changed for the psychological/somatic/sexual cluster and sleep/urinary cluster. More specifically, Class 2 was the moderate-high symptom burden group and Class 3 was the moderate-low symptom burden group in the psychological/somatic/sexual cluster. In contrast, in the sleep/urinary cluster, Class 2 becomes the moderate-low symptom burden group and Class 3 becomes the moderate-highs symptom burden group. To date, studies have identified the same severity trend across all symptoms and symptom clusters and do not consider the dynamic, changing nature of symptom clusters and their severity over time.^37,38 For example, two different symptom subgroups of a consistently lessening symptom group and moderately worsening symptom group were found to be identical across all different symptoms among formerly abused women.³⁷ However, symptoms are not always associated with other symptoms, which means that, in other words, a severely depressed patient may not have severe urinary incontinence problem but rather none to mild urinary incontinence problem.³⁹ This is further supported by our study findings in which severity trend for moderate-low and moderate-high symptom burden groups changed based on the psychological/somatic/sexual cluster and sleep/urinary cluster. With this new knowledge, clinicians should understand the dynamic nature of symptom clusters and their severity trend and carefully assess for any changes in symptom cluster severity over time. In addition, future research needs to focus on developing targeted interventions tailored to each class of midlife peri-menopausal and post-menopausal women with metabolic syndrome, which will help prevent the developmental trajectory changing into greater symptom burden.

In addition, an interactive relationship among the different symptom cluster subgroups was identified in the vasomotor/genital cluster. Herein, an interactive relationship indicates a similar trend of symptom clusters over time, which is evident in Classes 1 and 2, as well as in Classes 3 and 4. Our study finding adds to the current body of literature because previous studies have only identified symptom cluster subgroups that are distinctly different with each other.⁴⁰ There is a strong evidence that the symptoms within a symptom cluster may share interactive mechanisms, such as common physiological and endocrine changes that occur together.⁴¹ For example, a study of older women with vasomotor symptoms reported to experience vaginal dryness, both symptoms that consist the vasomotor/genital cluster.⁴² However, none of the studies have reported the potential interactive relationship among the symptom cluster subgroups within a symptom cluster. This highlights the need for future research to understand the potential interactions among symptom cluster subgroups and their underlying mechanisms, which can propose common pathways that may underlie this relationship. Furthermore, the use of symptom management interventions that target both the interactive clusters should be considered, such as cognitive behavioural therapy that has shown to be effective in improving symptom clusters among cancer patients undergoing chemotherapy.⁴³

The demographic, social, and clinical characteristics for each class were identified and tested for within-class differences. Among them, social support was significantly different among the four classes. Class 4 (high symptom burden) had the highest percentage of participants who had social support none of the time to a little of the time. This aligns with previous study findings in which lower social support was associated with greater symptom burden.^44,45 Furthermore, social support predicted patient outcomes, such as health-related quality of life and well-being that emphasizes the importance of social support.⁴⁶ As midlife women experience challenges in their social, psychological, and biological domains, including menopause, they often suffer from significant symptom burden and stress.⁴⁷ Therefore, it is important to understand the role of social support in reducing their symptom burden and improving health outcomes, to offer social resources, or to involve friends or family members, if necessary, to clinical care.

While not statistically significant, the potential impact of menopausal status at the initial diagnosis of metabolic syndrome should be considered. Class 1 (low symptom burden) had the highest percentage of midlife women in peri-menopause and Class 4 (high symptom burden) had the highest percentage in post-menopause. Our study findings support the current body of literature where post-menopausal women experienced more prevalent and severe symptoms than those in peri-menopause.⁴⁸ In consequence, among all menopausal stages, the health-related quality of life is most affected during the post-menopause stage.⁴⁹ This may be due to different hormonal profile (oestrogen and follicle-stimulating hormone) and rate of adipose tissue metabolism between peri-menopause and post-menopause among midlife women with metabolic syndrome.^50,51 With this in mind, future research should be conducted with a larger sample size where statistical significance may be reached in regards to their menopausal stage.

Our study has several limitations to consider. First, our sample consists of midlife peri-menopausal and post-menopausal women with metabolic syndrome of five racial/ethnic groups. When LCGA is applied to a larger sample for replication, the results may be different even with acceptable model fit indices because LCGA is specific to the study population which is one of the limitations of LCGA.³⁰ Second, there was a small sample size of 557 who met the study eligibility criteria. With the four-class model, there was an average of 139 participants in each class, thereby limiting its generalizability. Third, data were collected retrospectively via self-report that is prone to recall bias. Therefore, the results should be carefully interpreted with caution.

Conclusion

The current study identified four symptom cluster trajectory subgroups in midlife peri-menopausal and post-menopausal women with metabolic syndrome. In addition, we found a dynamic and interactive nature of symptom cluster trajectory subgroups. Social support was a significant predictor of symptom cluster trajectory subgroup that needs to be provided routinely. Clinicians should understand the different symptom cluster trajectory subgroups and their dynamic nature, and offer targeted and routine symptom cluster assessment and management in clinical settings.

Supplemental Material

sj-docx-1-whe-10.1177_17455057231160955 – Supplemental material for Identification of high-risk symptom cluster burden group among midlife peri-menopausal and post-menopausal women with metabolic syndrome using latent class growth analysis

Supplemental material, sj-docx-1-whe-10.1177_17455057231160955 for Identification of high-risk symptom cluster burden group among midlife peri-menopausal and post-menopausal women with metabolic syndrome using latent class growth analysis by Se Hee Min, Sharron L Docherty, Eun-Ok Im, Xiao Hu, Daniel Hatch and Qing Yang in Women’s Health

Footnotes

Acknowledgements

Not applicable.

Declarations

ORCID iD

Se Hee Min

Supplemental material

Supplemental material for this article is available online.

References

Rochlani

Pothineni

Kovelamudi

, et al. Metabolic syndrome: pathophysiology, management, and modulation by natural compounds. Ther Adv Cardiovasc Dis 2017; 11(8): 215–225.

Han

Lean

MEJ

. Metabolic syndrome. Medicine (Baltimore) 2015; 43(2): 80–87.

Oladejo

AO.

Overview of the metabolic syndrome; an emerging pandemic of public health significance. Ann Ib Postgrad Med 2011; 9(2): 78–82.

Aguilar

Bhuket

Torres

, et al. Prevalence of the metabolic syndrome in the United States, 2003-2012. JAMA 2015; 313(19): 1973–1974.

Bentley-Lewis

Koruda

Seely

EW.

The metabolic syndrome in women. Nat Clin Pract Endocrinol Metab 2007; 3(10): 696–704.

Marc

Genetic succeptibility to metabolic syndrome. EJIFCC 2007; 18(1): 7–14.

Harlow

Derby

CA.

Women’s midlife health: why the midlife matters. Womens Midlife Health 2015; 1: 5, https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6214213/ (accessed April 6, 2021).

Carr

MC.

The emergence of the metabolic syndrome with menopause. J Clin Endocrinol Metab 2003; 88(6): 2404–2411.

Min

Yang

Docherty

, et al. Symptom clusters and key symptoms among midlife peri-menopausal and post-menopausal women with and without metabolic syndrome. Nurs Res 2022; 71: E28–E38, https://journals.lww.com/nursingresearchonline/Abstract/9000/Symptom_Clusters_and_Key_Symptoms_Among_Midlife.99601.aspx (accessed March 21, 2022).

10.

Kravitz

Janssen

Bromberger

, et al. Sleep trajectories before and after the final menstrual period in the study of women’s health across the nation (SWAN). Curr Sleep Med Rep 2017; 3(3): 235–250.

11.

Tepper

Brooks

Randolph

Jr , et al. Characterizing the trajectories of vasomotor symptoms across the menopausal transition. Menopause 2016; 23(10): 1067–1074.

12.

Woods

Mitchell

ES.

Symptoms during the perimenopause: prevalence, severity, trajectory, and significance in women’s lives. Am J Med 2005; 118(12, Suppl. 2): 14–24.

13.

Harlow

Karvonen-Gutierrez

Elliott

, et al. It is not just menopause: symptom clustering in the study of women’s health across the nation. Womens Midlife Health 2017; 3(1): 2.

14.

Reeves

Elliott

Brooks

, et al. Symptom clusters predict risk of metabolic-syndrome and diabetes in midlife: the Study of Women’s Health Across the Nation. Ann Epidemiol 2021; 58: 48–55.

15.

Miaskowski

Future directions in symptom cluster research. Semin Oncol Nurs 2016; 32(4): 405–415.

16.

Miaskowski

Barsevick

Berger

, et al. Advancing symptom science through symptom cluster research: expert panel proceedings and recommendations. J Natl Cancer Inst 2017; 109(4): djw253.

17.

Jones

Nagin

DS.

Advances in group-based trajectory modeling and an SAS procedure for estimating them. Sociol Methods Res 2007; 35(4): 542–571.

18.

Zhou

Zheng

Liang

, et al. Joint developmental trajectories of bullying and victimization from childhood to adolescence: a parallel-process latent class growth analysis. J Interpers Violence 2022; 37(3–4): NP1759–NP1783.

19.

Fanti

KA.

Using a person-centered methodology to investigate the co-occurrence between psychopathological problems. Ann Depress Anxiety 2014; 1(8): 1–5.

20.

Torréns

Sutton-Tyrrell

Zhao

, et al. Relative androgen excess during the menopausal transition predicts incident metabolic syndrome in mid-life women: SWAN. Menopause N Y N 2009; 16(2): 257–264.

21.

Lampousi

Möller

Liang

, et al. Latent class growth modelling for the evaluation of intervention outcomes: example from a physical activity intervention. J Behav Med 2021; 44(5): 622–629.

22.

Steenkamp

Dickstein

Salters-Pedneault

, et al. Trajectories of PTSD symptoms following sexual assault: is resilience the modal outcome? J Trauma Stress 2012; 25(4): 469–474.

23.

Greene

Gelkopf

Grinapol

, et al. Trajectories of traumatic stress symptoms during conflict: a latent class growth analysis. J Affect Disord 2017; 220: 24–30.

24.

Bromberger

Matthews

Schott

, et al. Depressive symptoms during the menopausal transition. J Affect Disord 2007; 103(1–3): 267–272.

25.

Bromberger

Kravitz

Chang

, et al. Major depression during and after the menopausal transition: study of women’s health across the nation (SWAN). Psychol Med 2011; 41(9): 1879–1888.

26.

Min

Yang

Docherty

, et al. Symptom clusters and key symptoms among midlife perimenopausal and postmenopausal women with and without metabolic syndrome. Nurs Res 2022; 71(4): E28-E38.

27.

Hall

Okun

Sowers

, et al. Sleep is associated with the metabolic syndrome in a multi-ethnic cohort of midlife women: the SWAN sleep study. Sleep 2012; 35(6): 783–790.

28.

Jakobsen

Gluud

Wetterslev

, et al. When and how should multiple imputation be used for handling missing data in randomised clinical trials—a practical guide with flowcharts. BMC Med Res Methodol 2017; 17(1): 162.

29.

ICPSR, https://www.icpsr.umich.edu/web/pages/ (accessed January 17, 2023).

30.

Lee

Yang

Wolever

, et al. Health behavior trajectories in high cardiovascular risk populations: secondary analysis of a clinical trial. J Cardiovasc Nurs 2021; 36(6): E80–E90.

31.

Furihata

Konno

Suzuki

, et al. Unhealthy lifestyle factors and depressive symptoms: a Japanese general adult population survey. J Affect Disord 2018; 234: 156–161.

32.

Miley-Akerstedt

Jelic

Marklund

, et al. Lifestyle factors are important contributors to subjective memory complaints among patients without objective memory impairment or positive neurochemical biomarkers for Alzheimer’s disease. Dement Geriatr Cogn Dis Extra 2018; 8(3): 439–452.

33.

Ciccone

Baldi

, et al. Effectiveness of a lifestyle intervention on metabolic syndrome. A randomized controlled trial. J Gen Intern Med 2007; 22(12): 1695–1703.

34.

Kwaśniewska

Kaleta

Dziankowska-Zaborszczyk

, et al. Healthy behaviours, lifestyle patterns and sociodemographic determinants of the metabolic syndrome. Cent Eur J Public Health 2009; 17(1): 14–19.

35.

Yamaoka

Tango

Effects of lifestyle modification on metabolic syndrome: a systematic review and meta-analysis. BMC Med 2012; 10(1): 138.

36.

Kang

Choi

Lee

, et al. The effects of lifestyle modification on symptoms and quality of life in patients with irritable bowel syndrome: a prospective observational study. Gut Liver 2011; 5(4): 472–477.

37.

Cao

Yang

Steinberg

, et al. Symptom trajectory among formerly abused women: an exploratory study. Issues Ment Health Nurs 2021; 42(11): 989–997.

38.

Guérin

Goldfield

Prud’homme

Trajectories of mood and stress and relationships with protective factors during the transition to menopause: results using latent class growth modeling in a Canadian cohort. Arch Womens Ment Health 2017; 20(6): 733–745.

39.

van der Vaart

Roovers

de Leeuw

, et al. Association between urogenital symptoms and depression in community-dwelling women aged 20 to 70 years. Urology 2007; 69(4): 691–696.

40.

Hockenberry

Hooke

Rodgers

, et al. Symptom trajectories in children receiving treatment for leukemia: a latent class growth analysis with multitrajectory modeling. J Pain Symptom Manage 2017; 54(1): 1–8.

41.

Parker

Kimble

Dunbar

, et al. Symptom interactions as mechanisms underlying symptom pairs and clusters. J Nurs Scholarsh 2005; 37(3): 209–215.

42.

Zeleke

Bell

Billah

, et al. Vasomotor and sexual symptoms in older Australian women: a cross-sectional study. Fertil Steril 2016; 105(1): 149–155.

43.

Zhang

Wang

Zhang

, et al. A cognitive-behavioral intervention for the symptom clusters of Chinese patients with gastrointestinal tract cancer undergoing chemotherapy: a pilot study. Cancer Nurs 2019; 42(6): E24–E31.

44.

Atkins

Rubenstein

Sota

, et al. Impact of social support on cognitive symptom burden in HIV/AIDS. AIDS Care 2010; 22: 793–802, https://www.tandfonline.com/doi/full/10.1080/09540120903482994 (accessed March 2, 2022).

45.

Santos Salas

Watanabe

Tarumi

, et al. Social disparities and symptom burden in populations with advanced cancer: specialist palliative care providers’ perspectives. Support Care Cancer 2019; 27(12): 4733–4744.

46.

Alshraifeen

Al-Rawashdeh

Alnuaimi

, et al. Social support predicted quality of life in people receiving haemodialysis treatment: a cross-sectional survey. Nurs Open 2020; 7(5): 1517–1525.

47.

Thomas

Mitchell

Woods

NF.

The challenges of midlife women: themes from the Seattle midlife Women’s health study. Womens Midlife Health 2018; 4: 8.

48.

Ruan

Cui

, et al. Prevalence of climacteric symptoms comparing perimenopausal and postmenopausal Chinese women. J Psychosom Obstet Gynaecol 2017; 38(3): 161–169.

49.

Ceylan

Özerdoğan

Menopausal symptoms and quality of life in Turkish women in the climacteric period. Climacteric 2014; 17(6): 705–712.

50.

Ferrara

Lynch

Nicklas

, et al. Differences in adipose tissue metabolism between postmenopausal and perimenopausal women. J Clin Endocrinol Metab 2002; 87(9): 4166–4170.

51.

Randolph

Jr Sowers

Bondarenko

, et al. Change in estradiol and follicle-stimulating hormone across the early menopausal transition: effects of ethnicity and age. J Clin Endocrinol Metab 2004; 89(4): 1555–1561.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB