Multiple pathways to metabolic syndrome identified by hierarchical clustering of Japanese longitudinal health checkup data

By applying hierarchical clustering to longitudinal health checkup data from Japanese men, this study identified six distinct pre-onset pathways to metabolic syndrome characterized by varying sequences of abdominal obesity, hypertension, dyslipidemia, and hyperglycemia, thereby highlighting the heterogeneous nature of MetS development and supporting the need for targeted early intervention strategies.

The Gist

Imagine your body is a busy city. For this city to run smoothly, it needs a few key things: a manageable population size (weight), steady traffic flow (blood pressure), clean fuel lines (cholesterol), and a reliable power grid (blood sugar).

Metabolic Syndrome (MetS) is what happens when several of these systems start failing at the same time, putting the whole city at risk of a major blackout (heart disease or diabetes).

For a long time, doctors thought everyone's city failed in the exact same way: "First you get fat, then your blood pressure goes up, then your sugar gets high." But this new study suggests that's not true. There isn't just one road to MetS; there are six different highways.

Here is the story of how the researchers found these six paths, explained simply.

The Detective Work: Watching the City Before the Crash

The researchers acted like detectives looking at security footage from a Japanese health clinic. They didn't just look at people after they got sick; they looked at the three years leading up to the diagnosis.

They tracked 296 men who eventually developed MetS and compared them to 296 healthy men who were similar in age and starting health. They watched four specific "city sensors":

1. Waistline (Abdominal Obesity)
2. Blood Pressure (Traffic jams)
3. Cholesterol (Dirty fuel)
4. Blood Sugar (Power grid issues)

The Six Different "Disaster Movies"

Using a computer program that groups similar patterns together (like sorting movies by genre), they found that these men didn't all get sick the same way. They fell into six distinct clusters, or "storylines":

• Storyline 1 (The "Belly & Pressure" Start): About 1/3 of the men started with a growing belly and high blood pressure. The dirty fuel (cholesterol) was the last thing to go wrong.
• Storyline 2 (The "Belly & Fuel" Start): These men started with a growing belly and dirty fuel. The traffic jams (blood pressure) came later.
• Storyline 3 (The "Belly Only" Start): These men started with just a growing belly. Later, both the traffic and the power grid failed.
• Storyline 4 (The "Traffic & Fuel" Start): A different group started with high blood pressure and dirty fuel. Surprisingly, their bellies didn't grow until the very end!
• Storyline 5 (The "Traffic" Start): These men started with high blood pressure alone, and then the belly, fuel, and power grid all failed in quick succession.
• Storyline 6 (The "Everything Except Belly" Start): These men had high blood pressure, dirty fuel, and power grid issues before their bellies even started to grow.

The Healthy Controls: The men who didn't get sick? Their city sensors stayed relatively stable. They didn't have these dramatic shifts.

The Big Surprise: The "Belly" is the Final Straw

The most important discovery is that everyone in the "sick" group had a rapidly growing waistline right before they got diagnosed, even if they started with a normal waist.

Think of it like a balloon. Some people started with a slightly inflated balloon (obesity) and added air to the other parts. Others started with a flat balloon but had to pump it up very fast at the end to reach the breaking point.

The Takeaway: Even if you aren't overweight right now, if your waistline suddenly starts expanding quickly, it's a huge red flag. It's the "canary in the coal mine" that warns you a MetS disaster is coming, regardless of which of the six paths you are on.

Why This Matters: No More "One-Size-Fits-All"

In the past, doctors might have treated everyone the same way. But this study says: "Stop guessing. Look at the pattern."

• If you are on Storyline 1, your doctor should watch your cholesterol closely because it's the next thing to fail.
• If you are on Storyline 6, your doctor should watch your blood sugar and blood pressure immediately, even if your waist looks fine, because your body is already struggling internally.

The "Smoking" Twist

The study also found something interesting about lifestyle. People who quit smoking often gained weight (a common side effect). The study confirmed that quitting smoking was linked to a faster-growing waistline. However, the researchers emphasize: Don't start smoking again! The heart benefits of quitting are worth the weight gain, but you just need to be extra careful about your diet and exercise to keep your waistline in check.

The Bottom Line

Metabolic syndrome is complex, like a city with many different ways to fail. This study gives us a map of those six different failure paths. By knowing which path you are on, doctors can give you a personalized warning system to stop the disaster before it happens, rather than just treating the symptoms after the city has already burned down.

In short: Watch your waistline, know your specific risk pattern, and get help before the whole system crashes.

Technical Summary

1. Problem Statement

Metabolic Syndrome (MetS) is a cluster of interrelated abnormalities (abdominal obesity, hypertension, dyslipidemia, and hyperglycemia) that significantly increase the risk of cardiovascular disease and type 2 diabetes. While the components of MetS are well-defined, the temporal progression and heterogeneous pathways leading to its onset remain underexplored. Previous studies using Markov models have simulated disease transitions but often lack the granularity to distinguish multiple distinct developmental trajectories. There is a critical need to identify pre-onset patterns to enable targeted early interventions, particularly given that individuals may follow different sequences of symptom emergence before meeting the full diagnostic criteria.

2. Methodology

Study Design and Population:

• Type: Retrospective longitudinal cohort study.
• Data Source: Japanese health checkup data from the Hokuriku Health Service Association (Toyama Prefecture) collected between April 2012 and March 2021.
• Participants:
  • Cases: 296 male participants diagnosed with MetS (Year 0).
  • Controls: 296 male participants without MetS, matched 1:1 to cases based on clinical profiles from the initial year of a four-year window.
  • Matching Criteria: Euclidean distance matching based on a five-dimensional vector (Waist Circumference [WC] value + 3-level categorical classifications of WC, Hypertension [HT], Dyslipidemia [DLP], and Hyperglycemia [HG]).
  • Exclusion: Non-fasting blood samples, missing key variables, or data gaps in the 3 years prior to onset.
• Timeframe: Analysis focused on the three years preceding MetS onset (Year -3 to Year -1) and the year of onset (Year 0).

Variable Encoding and Clustering:

• Input Variables: Four core MetS components (WC, HT, DLP, HG).
• Encoding: Each component was categorized into three levels (0: Normal, 1: Intermediate, 2: MetS criteria) based on specific thresholds.
• Feature Vector: For each participant, the 3-year longitudinal status of the 4 components was encoded into a 12-digit ternary code (4 components × 3 years).
• Algorithm: Hierarchical Clustering using Euclidean distance and Ward's method.
  • Optimization: The optimal number of clusters was determined using silhouette scores (selecting 6 clusters over 4, despite a local maximum at 4, to capture population heterogeneity).
  • Validation: K-means clustering was tested but yielded inconsistent results (Adjusted Rand Index ~0.60), leading the authors to rely on the deterministic nature of hierarchical clustering.

Statistical Analysis:

• Comparisons: Unpaired t-tests or Wilcoxon rank-sum tests for continuous variables; Chi-square or Fisher's exact tests for categorical variables.
• Trend Analysis: Annual changes in Waist Circumference (WC) and Body Mass Index (BMI) were estimated using simple linear regression.
• Predictors: Multiple linear regression was used to identify factors (e.g., smoking cessation, lifestyle efforts) associated with WC changes.
• Biomarkers: Triglyceride-glucose index (TyG) combined with WC (TyG-WC) and BMI (TyG-BMI) were calculated to assess insulin resistance and cardiometabolic risk.

3. Key Contributions

1. Identification of Six Distinct Pathways: The study successfully identified six unique temporal trajectories leading to MetS onset, challenging the notion of a single linear progression.
2. Heterogeneity of Onset: Demonstrated that while abdominal obesity (OB) is a required criterion for diagnosis, it does not always precede other components. In some clusters, HT and DLP appeared years before significant OB.
3. Control Group Stability: Confirmed that non-MetS controls with similar baseline profiles did not exhibit the same dynamic shifts in component combinations, validating that the observed trajectories are specific to MetS progression.
4. Sex-Specific Insights: Preliminary analysis of female participants suggested different patterns (e.g., earlier emergence of Hyperglycemia and more gradual progression), highlighting the need for sex-specific models.

4. Key Results

The Six Identified Clusters (Male Participants):

• Cluster 1 (32.8%): Early OB + HT $\rightarrow$ Late DLP. (Most common pattern).
• Cluster 2 (12.2%): Early OB + DLP $\rightarrow$ Late HT.
• Cluster 3 (16.2%): Early OB $\rightarrow$ Late HT + HG.
• Cluster 4 (14.9%): Early HT + DLP $\rightarrow$ Late OB. (Non-obese onset of other components).
• Cluster 5 (15.5%): Early HT $\rightarrow$ Late OB + DLP + HG.
• Cluster 6 (8.4%): Early HT + DLP + HG $\rightarrow$ Late OB.

Clinical Dynamics:

• Temporal Shifts: Significant changes in component combinations occurred primarily between Year -1 and Year 0 (the year of onset).
• Waist Circumference (WC): WC increased significantly in all MetS clusters compared to controls. Notably, Clusters 4, 5, and 6 exhibited rapid WC increases in the final year, despite starting with non-obese profiles.
• TyG Indices: TyG-WC and TyG-BMI were significantly elevated in MetS groups across all clusters.
  • Cluster 2 showed the highest TyG-WC (highest cardiometabolic risk).
  • Cluster 5 showed the lowest TyG-WC among MetS clusters.
• Lifestyle Factors: No significant differences in diet, physical activity, or sleep were found between MetS and control groups within clusters. However, smoking cessation was significantly associated with WC gain in the MetS group.
• Age: Contrary to some literature, the MetS group was slightly younger than the matched controls, potentially due to the matching methodology or specific Japanese population trends regarding obesity-based criteria.

5. Significance and Implications

• Personalized Intervention: The findings suggest a "one-size-fits-all" approach to MetS prevention is insufficient. Interventions should be tailored to the specific cluster:
  • Clusters 1-3: Focus on managing existing obesity and hypertension to prevent the emergence of the third component.
  • Clusters 4-6: Crucial for early detection in non-obese individuals; monitoring for rapid WC gain and metabolic abnormalities (HT, DLP, HG) is vital even before obesity is clinically apparent.
• Early Warning Systems: The study highlights that the year immediately preceding diagnosis (Year -1 to 0) is a critical window where rapid deterioration occurs. Shorter follow-up intervals are recommended for individuals showing rapid WC increases.
• Methodological Framework: The use of hierarchical clustering on longitudinal health checkup data provides a flexible, data-driven framework for characterizing disease trajectories that can be applied to other complex chronic conditions.
• Limitations: The study is limited to a Japanese male population (limiting generalizability to other ethnicities/sexes), lacks genetic/socioeconomic data, and has a relatively short follow-up (4 years) for long-term outcome prediction.

Conclusion:
This study provides robust evidence that MetS develops via multiple, distinct pathways. By identifying these trajectories, healthcare providers can move from reactive diagnosis to proactive, cluster-specific monitoring and intervention, potentially preventing the onset of MetS and its associated cardiovascular risks.