PERsonalised Knowledge to reduce the risk of Stroke (PERKS-International): a randomised controlled trial testing the efficacy of an mHealth application to reduce risk factors for the primary prevention of stroke

In a randomized controlled trial involving 862 adults with at least two stroke risk factors, the Stroke Riskometer mobile app failed to significantly improve overall cardiovascular health scores compared to usual care at six months, despite showing a borderline increase in physical activity.

The Gist

The Big Idea: Can a Phone App Save Your Brain?

Imagine your body is a high-performance car. You know that if you ignore the oil, the tires, and the fuel, the engine might eventually blow a gasket. In the human body, that "engine blowout" is a stroke.

Scientists have long known that most strokes are preventable if you manage your "car maintenance" (things like blood pressure, diet, exercise, and smoking). But knowing what to do is hard, and actually doing it is even harder.

So, researchers asked a simple question: If we give people a fancy, personalized "dashboard" on their smartphones to track their health, will they actually drive their cars better?

This study, called PERKS-International, tested exactly that using an app called the Stroke Riskometer™.

The Experiment: The "GPS" vs. The "Paper Map"

The researchers gathered 862 volunteers from Australia and New Zealand. These were everyday people (ages 35–75) who already had a few "warning lights" on their dashboard (like high blood pressure or being overweight), but who hadn't had a stroke yet.

They were split into two groups:

1. The "GPS" Group (Intervention): These people got access to the Stroke Riskometer app. It was like having a personal health coach in your pocket. It calculated their specific risk of having a stroke, gave them personalized tips, and let them track their progress over time.
2. The "Paper Map" Group (Usual Care): These people got a single email with a summary of their health numbers and a link to some general health articles. It was helpful, but it was static—like getting a paper map once and then having to figure out the rest of the journey on your own.

The Goal: To see if the "GPS" group improved their health scores (specifically the "Life's Simple 7" score, which measures 7 key health factors) more than the "Paper Map" group after 6 months.

The Results: The GPS Didn't Change the Route

Here is the twist: The app didn't work as hoped.

After six months, the researchers checked the "dashboard" of both groups.

• The GPS Group did not show a significantly better improvement in their overall health scores compared to the Paper Map Group.
• Both groups actually improved a little bit on their own (perhaps just by paying attention to the study), but the app didn't give the "GPS" group a magic boost.

The One Small Win:
There was one tiny exception. The "GPS" group did a little bit more physical activity (walking, running, etc.) than the other group. It was like the app successfully convinced them to take the stairs instead of the elevator, but it didn't convince them to eat better, stop smoking, or lower their blood pressure.

Why Didn't the App Work? (The "Why" Behind the "What")

The researchers offered a few creative explanations for why the app didn't save the day:

• The "Ghost Rider" Problem: Even though people were told to download the app, about 20% of them never actually opened it. It's like giving someone a Ferrari key, but they leave it in the glovebox and keep driving their old sedan.
• The "Real World" Test: The study was designed to be "real world." The researchers didn't call the participants every day to nag them. They just sent a few emails. In the real world, if you don't have a coach standing over your shoulder, you might forget to use the app.
• The "Teachable Moment" Gap: The study focused on healthy-ish people. Research suggests that apps work best when people have just had a scare (like a heart attack). That event acts as a "wake-up call." For people who feel fine, an app is just another notification on a phone full of notifications.
• The App Might Be "Old Tech": The app used standard tools. It didn't have the latest "gamification" (like earning points or badges) or AI chatbots that make health feel like a fun game rather than a chore.

The Takeaway

The Bottom Line:
Simply handing a person a health app and saying, "Here, track your stroke risk," isn't enough to change their life. In this study, the app was a nice tool, but it wasn't a magic wand.

What This Means for the Future:
To really help people prevent strokes, apps might need to be more engaging (like a video game), or they need to be paired with real human support (like a doctor or a coach calling you).

Think of it this way: An app is like a recipe book. It can tell you exactly how to bake a cake (or how to lower your blood pressure), but if you don't actually go into the kitchen, mix the ingredients, and turn on the oven, you're still not going to get a cake. The PERKS study showed that while the recipe book was good, most people didn't actually start cooking.

Technical Summary

1. Problem Statement

Stroke incidence rates have plateaued, with some evidence of increasing rates in younger demographics. Approximately 90% of strokes are preventable through the management of modifiable risk factors (e.g., hypertension, obesity, physical inactivity, smoking). Despite the high prevalence of these risk factors in populations like Australia and New Zealand, effective strategies for early identification and management in primary prevention settings remain a research priority.

While mobile health (mHealth) applications offer a scalable solution due to high smartphone penetration, existing evidence is limited by:

• A preponderance of small pilot studies.
• A focus on secondary prevention (patients with existing cardiovascular disease) rather than primary prevention.
• A lack of large-scale, adequately powered RCTs with objectively measured risk factors.

The study aimed to evaluate the efficacy of the Stroke Riskometer™ app in improving the Life's Simple 7® (LS7®) cardiovascular risk score among adults with at least two stroke risk factors but no history of cardiovascular disease.

2. Methodology

Study Design:

• Type: Phase III, prospective, pragmatic, open-label, single-blinded (outcome assessor) 2-arm RCT.
• Setting: Australia (Hobart, Melbourne, Adelaide) and New Zealand (Auckland, Hamilton).
• Timeline: Recruitment from August 2021 to January 2024; 6-month follow-up completed by July 2024.
• Participants: $N=862$ adults aged 35–75 years with $\ge$2 "poor" LS7® risk factors, smartphone ownership, and no prior stroke, myocardial infarction, or cognitive impairment.
• Randomisation: 1:1 allocation to Intervention or Usual Care, minimised by site.

Intervention vs. Control:

• Intervention Group: Received access to the Stroke Riskometer™ app. The app provided personalised absolute/relative stroke risk estimates, evidence-based feedback, self-monitoring tools, and push notifications. Participants received technical support only; no clinical advice was given via the app.
• Usual Care Group: Received a single email containing their baseline risk factor results and links to generic, evidence-based stroke prevention information. They were not informed about the app to prevent contamination.

Outcomes:

• Primary Outcome: Mean between-group difference in the change of the total LS7® score (0–14 scale) from baseline to 6 months. The score comprises seven domains: blood pressure, cholesterol, glucose, BMI, smoking, physical activity, and diet.
• Secondary Outcomes: Changes in individual LS7® items.
• Data Collection: Objective measurements (BP, non-fasting cholesterol/glucose via Cardiochek®, height, weight) were taken face-to-face by blinded research assistants. Lifestyle data were collected via questionnaires.

Statistical Analysis:

• Primary Analysis: Intention-to-Treat (ITT) using Analysis of Covariance (ANCOVA) and linear mixed models.
• Missing Data: Handled via Multiple Imputation using Chained Equations (MICE) with 20 imputations.
• Sensitivity Analysis: Per-protocol analysis excluding participants who did not download/use the app, those with unblinded assessors, or those with incomplete data.

3. Key Contributions

• Scale and Rigor: This is one of the few fully powered, Phase III mHealth RCTs focused on primary prevention with objectively measured biomedical risk factors, addressing a significant gap in the literature.
• Real-World Design: The study employed a "pragmatic" design with minimal staff-patient interaction to simulate real-world app usage, avoiding the "Hawthorne effect" common in highly controlled trials.
• Comprehensive Evaluation: It evaluated a specific, validated algorithm (Stroke Riskometer™) against a control group receiving standard risk feedback, providing a clear benchmark for the efficacy of standalone app-based interventions.

4. Results

Demographics:

• Mean age: 58 ± 11 years; 63% female; 74% Caucasian.
• Baseline characteristics were generally balanced, though the usual care group had a slightly higher proportion of participants in the "high" LS7® category.

Primary Outcome (Total LS7® Score):

• ITT Analysis: No statistically significant difference between groups.
  • Mean difference in change: 0.03 (95% CI: -0.19 to 0.25; $p=0.79$).
  • Both groups showed similar, small improvements from baseline (Intervention: +0.18; Usual Care: +0.15).
• Per-Protocol Analysis: Results remained non-significant, though the effect size was slightly larger (Mean difference: 0.20; $p=0.11$).
• Subgroup Analysis: No significant effect modification by age, sex, country, socioeconomic status, or baseline LS7® score.

Secondary Outcomes (Individual LS7® Items):

• Physical Activity: The intervention group showed a borderline significant increase in metabolic equivalent of task (MET) minutes/week compared to usual care.
  • Mean difference in change: 313.42 MET-mins/week (95% CI: -2.80 to 629.65; $p=0.05$).
  • In per-protocol analysis, this reached statistical significance ($p=0.04$).
• Other Items: No significant differences were observed for BMI, dietary quality, blood glucose, cholesterol, systolic blood pressure, or smoking status.

Adherence:

• 64% of the intervention group self-reported downloading the app within one week.
• However, 20% of the intervention group were determined (via survey or lack of data) to have never downloaded or used the app.
• Only 32% of the intervention group consented to share app usage data, limiting fidelity assessment.

5. Significance and Conclusion

Conclusion:
In a real-world setting, providing access to the Stroke Riskometer™ app did not result in a significant improvement in the overall cardiovascular risk profile (LS7® score) compared to receiving standard risk information via email after 6 months. The only notable finding was a marginal increase in physical activity in the intervention group.

Implications:

• Limitations of Standalone Apps: The study suggests that simply providing risk information and a tracking tool via a mobile app is insufficient to drive significant behavioral change in a primary prevention population without additional support mechanisms.
• Engagement Challenges: The lack of effect may be attributed to low engagement (20% non-adherence) and the "real-world" design which lacked intensive follow-up.
• Future Directions: The authors suggest that future mHealth interventions for stroke prevention may require:
  • Enhanced engagement features (e.g., gamification, AI chatbots, wearable integration).
  • Integration with face-to-face or telehealth support (hybrid models).
  • Updates to the app to provide more specific, actionable feedback on individual risk factors.

This study provides critical evidence that while mHealth is a promising tool, its efficacy in primary stroke prevention is limited when deployed as a standalone, low-touch intervention.