Factors associated with patient preferences for statin therapy in primary prevention

Abstract

Introduction

The smallest worthwhile difference (SWD) – the minimum absolute benefit required to justify a treatment against its burdens – varies by personal preference. We investigated factors associated with SWD variation for statin therapy in primary prevention of atherosclerotic cardiovascular disease (ASCVD) to inform shared decision-making strategies.

Methods

This pre-planned secondary analysis used survey data from the US (n = 602) and Japan (n = 665). Participants determined their SWD for statins across three 10-year ASCVD risk scenarios (2%, 10%, 20%) using the Benefit-Harm Trade-off Method. Subgroup analyses compared SWD by statin use history, self-perceived baseline risk, and hypercholesterolemia diagnosis. Associations between participant characteristics and SWD were examined using proportional odds regression adjusted for age and sex. A post hoc analysis using a directed acyclic graph explored the causal role of prior cholesterol education.

Results

The median SWD exceeded the benefits typically provided by statins across all scenarios. Predictors of lower SWDs (greater willingness to initiate statins) were mostly consistent across countries. Higher self-perceived ASCVD risk was associated with lower SWD: in the moderate-risk scenario, median SWDs decreased from 7.5–10.0% to 2.5–5.0% as self-perceived risk increased. Current statin use, hypercholesterolemia history, and cardiovascular comorbidities were each associated with lower SWDs. After adjusting for confounders, prior cholesterol education was associated with 38% (US) and 47% (Japan) lower odds of selecting a higher SWD (OR 0.62 [95%CI 0.42–0.90] and 0.53 [0.31–0.90], respectively), indicating lower treatment efficacy expectations.

Conclusions

Risk perception and understanding of treatment benefits may predict statin acceptance. Educational interventions strengthening treatment-specific knowledge may represent a promising strategy to align patient preferences with clinical evidence.

1. Introduction

Statins are well-established for the primary prevention of atherosclerotic cardiovascular disease (ASCVD). Evidence indicates that statins reduce ASCVD risk by approximately 25% (i.e., relative risk [RR] ≈ 0.75) per 1mmol/L reduction in low density lipoprotein cholesterol (LDL-C), with broadly similar proportional effects across baseline risk groups [[1], [2], [3]]. Accordingly, the approximate absolute risk reduction varies by baseline risk: 1.3% reduction for individuals with a 5% baseline ASCVD risk, and roughly 5% reduction for those with a 20% baseline risk [1]. Given this evidence and statins’ generally favorable safety profile [4], guidelines worldwide recommend statins for individuals at moderate to high ASCVD risk.

These recommendations primarily reflect expert interpretation of clinical evidence, with limited incorporation of patient perspectives. To better understand patient preferences, we previously applied the smallest worthwhile difference (SWD) to statin therapy for primary prevention in general population samples in the US and Japan [5]. The SWD represents the minimum absolute benefit an individual requires to consider a treatment worthwhile after weighing its potential burdens, including harms, costs, and inconveniences [6]. Across baseline risk scenarios, median SWDs exceeded the absolute benefits typically achieved with statins, and fewer than one-third indicated willingness to initiate statins at risk thresholds commonly recommended in guidelines. These findings suggest a potential gap between specialist recommendations and patient expectations, which may impair adherence and challenge truly informed shared decision making (SDM).

Understanding the factors that influence patients’ benefit-burden trade-off judgments, and lead to variation in SWD, could help clinicians tailor communication strategies during SDM. Self-perceived ASCVD risk is one plausible factor. Patients’ subjective appraisal of their own risk may be as important as objectively estimated risk based on clinical characteristics for shaping treatment preferences, as it was shown to be associated with greater willingness to initiate treatments [7,8]. A prior diagnosis of hypercholesterolemia may exert similar influence, as such individuals may perceive statin therapy as more personally relevant. In addition, disease-specific knowledge is indicated to affect risk comprehension and treatment preferences in preventive contexts and represents a potentially modifiable factor [9,10]. Notably, in our primary study, Japanese respondents reported substantially lower prior cholesterol education and lower comprehension of statin-related information compared with US respondents [5], raising the possibility that differences in prior education may contribute to variation in SWD.

In the present study, we conducted pre-specified subgroup analyses based on a priori hypotheses to examine whether SWD distributions differed by (i) age and statin use history, (ii) self-perceived ASCVD risk, or (iii) prior diagnosis of hypercholesterolemia. We also examined the association of additional baseline characteristics through multivariable regression. In addition, we performed a post hoc causal analysis to explore the potential causal contribution of prior cholesterol education to SWD, and a post hoc analysis to explore the relationship between age and SWD. Because education is potentially modifiable and intervenable, clarifying its role may inform strategies to support more preference-concordant shared decision-making.

2. Methods

2.1. Study design and ethical approval

This is a secondary analysis of a cross-sectional online survey. The primary study, including survey development and primary analysis, has been reported elsewhere [5]. The present analysis was specified in part a priori in the original protocol and in part conducted post hoc for exploratory purposes.

The study was approved by the Ethics Committee of the Kyoto University Graduate School and Faculty of Medicine (R4587). All participants provided electronic informed consent prior to participation. The study protocol is publicly available on the Open Science Framework (OSF): https://osf.io/f8yud.

2.2. Parent study overview

Between November 2024 and March 2025, we recruited general population samples from two research crowdsourcing services (Prolific for the US and Macromill for Japanese participants) [[11], [12], [13]] using stratified sampling to approximate national census demographic distributions. Eligible participants were adults aged 18 years or older.

The questionnaire was designed using the Benefit-Harm Trade-off Method (BHTM) [14]. Participants first reviewed standardized information on ASCVD and on the benefits (prevention of ASCVD) and burdens (adverse effects, costs, and inconvenience) of statin therapy, then completed two comprehension questions with correct answers displayed afterward. As clinical data such as serum LDL-C and blood pressure were likely to be inaccessible in an online survey setting, we did not provide individualized risk estimation and instead presented three hypothetical scenarios with an assumed baseline 10-year ASCVD risk of 2%, 10%, and 20%, corresponding to low-, moderate-, and high-risk groups, respectively. These risk groups were defined to be interpreted consistently under both the American and Japanese guidelines [4,15]. All participants responded to all three scenarios, presented in random order. Each scenario opened with a visual explanation of the untreated baseline risk in both percentage and frequency formats, after which the SWD was obtained through an iterative back-and-forth sequence (Methods Supplement of the original report [5]). Taking the moderate-risk scenario as an example, participants were first asked whether they would take statins if they could reduce their risk from 10% to 0%. If they answered “No”, indicating that the burdens outweighed even the maximum possible benefit, the scenario ended and the participant was classified as “unwilling”. If they answered “Yes”, they were then asked about a smaller reduction (e.g., from 10% to 5%), and the process continued with progressively smaller reductions until the smallest reduction at which the participant would still opt for statin therapy was identified. This value was taken as the SWD for that scenario, yielding three SWD outcomes per participant.

2.3. Secondary analyses

The primary publication focused on estimating SWD distributions in adults aged 40–75 who have never used statins (i.e., statin-naïve participants) [5]. The current study extends that work by examining potential sources of variation in SWD. We conducted four analyses:

• 1.Prespecified subgroup analyses by age and statin use history, self-perceived ASCVD risk, or prior diagnosis of hypercholesterolemia.
• 2.Prespecified multivariable regression analyses of various baseline variables, adjusted for age and sex (high-risk scenario only).
• 3.A post hoc descriptive analysis examining the relationship between age and SWD.
• 4.A post hoc exploratory analysis examining the potential role of prior cholesterol education (high-risk scenario only).

As with the primary analysis, all the analyses were performed separately in the US and Japanese samples.

1. Subgroup analyses.

We descriptively compared SWD distributions across three subgroup classifications defined in the original protocol:

(i) Age and statin use history:

Based on age and statin use history, participants were categorized into five groups (Fig. 1):

• a)Young (age<40 years, regardless of statin use history).
• b)Older (age>75 years, regardless of statin use history).
• c)Statin-naïve (age 40–75 years, no prior statin use).
• d)Former statin users (age 40–75 years, prior statin use but currently discontinued).
• e)Current statin users (age 40–75 years, ongoing statin use).

Fig. 1.

Participant groups. Abbreviations: ASCVD: atherosclerotic cardiovascular disease.

The age range of 40–75 years represents the population for whom ASCVD primary prevention risk assessment is recommended by current guidelines.

(ii) Self-perceived 10-year ASCVD risk:

In the questionnaire, participants were asked to assess their own 10-year ASCVD risk group after reviewing the necessary medical knowledge on statins (“Considering your current health status, how likely do you think you are to develop an atherosclerotic cardiovascular disease over the next 10 years? There is no right or wrong answer to this question, so please respond with your honest impression.”). There were three options: low risk (<5%), borderline or intermediate risk (5–20%), and high risk (≥20%).

(iii) Prior diagnosis of hypercholesterolemia:

Participants reported whether they had ever been diagnosed with hypercholesterolemia or hyperlipidemia (“Have you ever been diagnosed with high cholesterol or high lipid levels [hypercholesterolemia/hyperlipidemia] during a health-checkup or by a medical professional? – Yes/Probably yes/No/Probably no”).

Consistent with the primary analysis, subgroup analyses were restricted to participants who correctly answered the comprehension test on statin benefits to ensure a basic understanding of statin therapy. Given the exploratory nature of subgroup analyses, results were primarily presented descriptively with visualization. Hypothesis testing was limited to assessing whether any overall difference existed across subgroup categories, rather than all pairwise comparisons. To account for the ordinal nature of the SWD outcome and the within-person correlation arising from each participant contributing three SWD measurements across risk scenarios, we used a mixed-effects cumulative link model with participant as a random effect and scenario as a fixed effect. An omnibus test for the subgroup variable was obtained, yielding a single p-value for whether any difference existed among the groups.

2. Regression analyses.

As specified in the protocol, we used a proportional odds logistic regression to explore associations between various background variables and SWD in the high-risk (20%) scenario. SWD was modeled as an ordinal variable from the smallest measurable SWD to the “unwilling” category, with higher values representing greater reluctance to initiate statins. Rather than fitting a single multivariable model, we fitted separate models for each predictor with age and sex as covariates. This approach was chosen because regression coefficients from a fully adjusted model can only be interpreted conditional on all other variables in the model, but since covariate sets are likely to differ across studies, such coefficients would have limited interpretability and could not serve as a meaningful reference for future research [16]. As comprehension check items were among the factors examined as potential correlates of SWD, this analysis included all participants.

Specifically, for each factor $X$, we fitted the following model: $\text{logit}\left(P[SWD>j]\right)={\alpha }_j+{\beta }_{x}X+{\beta }_{1}Age+{\beta }_{2}Sex$, where $j$ denotes each threshold level of SWD (i.e., 2.5, 5, …, 17.5, 20, unwilling). The coefficient ${\beta }_x$ reflects the association between $X$ and SWD after accounting for age and sex and can be exponentiated to yield an odds ratio (OR). Importantly, ${\beta }_x$ remains constant across thresholds $j$ – this is the proportional odds assumption – meaning a single OR summarizes the association between $X$ and SWD across all thresholds, while the intercept ${\alpha }_j$ varies across $j$ to account for the different cumulative probabilities at each threshold. We did not formally test the proportional odds assumption as standard tests are prone to low statistical power and inflated type I error rates when applied to numerous predictors and models. Furthermore, given our sample size, more flexible modeling alternatives were at risk of overfitting. Consequently, these results should be interpreted as a parsimonious approximation of the data trends. An OR < 1 indicates that, compared to the reference group ($X=0$), participants in the group $X=1$ have lower odds of being above any higher SWD category – that is, they show greater willingness to initiate statin therapy. Conversely, an OR > 1 indicates lower willingness to treat relative to the reference group. For each factor $X$, we estimated the OR and its 95% confidence interval (CI) to quantify the direction and magnitude of association with SWD, alongside the Nagelkerke R² and c-statistic to approximate the degree to which each variable explains variation in and discriminates between levels of SWD. Both statistics were corrected for optimism using 1000 bootstrap resamples.

3. Relationship between age and SWD.

Because age was included as a covariate in the regression analyses described above, its relationship with SWD was not examined directly. Therefore, a post hoc descriptive analysis was conducted to explore the relationship between age and SWD. Unlike the other characteristics examined above, age is a continuous variable, and its relationship with SWD may be non-linear; modeling such associations would have complicated interpretation. Consequently, age was grouped into bands for descriptive purposes, and no formal statistical testing or modeling was performed. To visualize the relationship between age and SWD, SWD was treated as a continuous variable; thus, only participants willing to accept statins under the high baseline risk scenario were included. The distribution of SWD across age bands was summarized using violin plots.

4. Association between cholesterol education and SWD.

Because the primary study identified differences in prior cholesterol education and comprehension checks between US and Japanese respondents [5], and because the education is a modifiable variable, we conducted a post hoc exploratory analysis to examine the potential causal relationship between prior cholesterol education and SWD. This analysis was hypothesis-generating and should not be interpreted as establishing causality.

Prior education was identified by the question, “Have you ever received any medical advice or instructions about high cholesterol or high lipid levels (hypercholesterolemia/hyperlipidemia) from a medical professional?” We hypothesized that prior education might reduce SWD by enhancing appreciation of statin benefits. A directed acyclic graph (DAG) was constructed to identify potential confounders (Figure S1 in the Appendix). We then fitted a proportional odds logistic regression adjusting for those confounders. All analyses were conducted in R [17].

3. Results

A total of 602 US and 665 Japanese participants completed the survey. To ensure a basic understanding of statin therapy, subgroup analyses were restricted to participants who correctly answered the statin benefits comprehension check, comprising 593 (98.5%) US and 527 (79.2%) Japanese participants. All other analyses were conducted on the full samples.

1. Subgroup analyses.

Table 1 summarizes the median SWD and proportion of participants willing to accept statins based on current evidence across prespecified subgroups, with a visual representation provided in Appendix Figure S3. Though specific values varied slightly, the overall trends were remarkably consistent in two countries.

Table 1.

Subgroup analysis results.

		US (N = 593)								Japan (N = 527)
			Low-risk scenario (ASCVD risk = 2%)		Moderate-risk scenario (ASCVD risk = 10%)		High-risk scenario (ASCVD risk = 20%)				Low-risk scenario (ASCVD risk = 2%)		Moderate-risk scenario (ASCVD risk = 10%)		High-risk scenario (ASCVD risk = 20%)
		N	Median SWD	% willing to start statins at its current effect sizea	Median SWD	% willing to start statins at its current effect sizea	Median SWD	% willing to start statins at its current effect sizea	P valueb	N	Median SWD	% willing to start statins at its current effect sizea	Median SWD	% willing to start statins at its current effect sizea	Median SWD	% willing to start statins at its current effect sizea	P valueb
(i) Age and statin use history	Young	183	Unwilling	15.8%	7.5	22.4%	7.5	35.0%	<0.0001	119	Unwilling	21.8%	7.5	26.1%	12.5	30.3%	<0.0001
	Older	1	-	-	-	-	-	-		54	Unwilling	27.8%	7.5	38.9%	5	53.7%
	Statin-naïve	254	Unwilling	13.8%	7.5	23.6%	7.5	34.2%		297	Unwilling	19.2%	7.5	23.9%	12.5	30.6%
	Former user	37	Unwilling	10.8%	7.5	35.1%	7.5	48.6%		8	1	25%	7.5	37.5%	2	75%
	Current user	118	1.5	32.2%	2.5	62.7%	2.5	76.3%		49	0.5	53.1%	1	61.2%	2	71.4%
(ii) Self-perceived risk	Low risk	275	Unwilling	13.1%	7.5	20.0%	10	33.8%	<0.0001	218	Unwilling	18.8%	10	23.4%	15	30.7%	<0.0001
	Moderate risk	245	Unwilling	20.1%	5	38.1%	5	50.4%		227	Unwilling	25.1%	5	31.7%	10	39.6%
	High risk	73	2	28.8%	2.5	54.8%	5	58.9%		82	2	34.1%	5	40.2%	7.5	48.8%
(iii) Prior diagnosis of hypercholesterolemia	No hypercholesterolemia	308	Unwilling	13.3%	7.5	19.5%	10	32.5%	<0.0001	368	Unwilling	19.8%	10	24.5%	12.5	30.7%	<0.0001
	Probable hypercholesterolemia	285	Unwilling	22.9%	5	45.1%	5	55.9%		159	1.5	33.3%	5	41.5%	5	52.8%

^a“Current effect size” is based on an RR of 0.75 per 1mmol/L reduction in LDL-C for primary prevention.

^bP-value is derived from an omnibus test for the subgroup variable of a mixed-effects cumulative link model with participant as a random effect and scenario as a fixed effect.

Abbreviations: ASCVD: atherosclerotic cardiovascular disease. SWD: smallest worthwhile difference.

Regarding subgroups defined by age and statin use history, detailed numbers and proportions of participants for each subgroup in the two countries are presented in Figure S2 in the Appendix. Among US participants aged 40–75 (n = 409), 254 (62.1%) had never used statins and 118 (28.9%) were current users. In Japan, 354 were aged between 40 and 75, with 297 (83.9%) statin-naïve, a notably higher proportion than in the US sample. Current statin users demonstrated less reluctance to initiate statin therapy compared to statin-naïve participants (Table 1, Figure S3 (i)). For instance, in the US, the median SWDs at moderate- and high-risk scenarios were both 7.5 among statin-naïve participants but dropped to 2.5 among current users, with willingness to accept statins at their current efficacy from 23.6% and 34.2% to 62.7% and 76.3%, respectively. A similar pattern was observed in Japan, where median SWD fell from 7.5 and 12.5 to 1 and 2 across the same scenarios, and willingness increased from 23.9% and 30.6% to 61.2% and 71.4%.

Participants with higher self-perceived ASCVD risks showed greater willingness to accept statins (Table 1, Figure S3 (ii)), as did those with a prior hypercholesterolemia diagnosis (Table 1, Figure S3 (iii)). Taking the moderate-risk scenario as an example, the median SWD reduced from 7.5 to 2.5 in the US and from 10 to 5 in Japan as self-perceived risk increased from low to high, with willingness rising from 20% to 54.8% and from 23.4% to 40.2%, respectively.

2. Regression analysis.

Table 2 presents the estimated ORs indicating the association between background variables and SWD, adjusted for age and sex. In both the US and Japan, presence of comorbidities, ASCVD history, hypercholesterolemia, prior education on blood cholesterol, current statin use, and moderate to high self-perceived risk were associated with a lower SWD, indicating greater willingness to initiate statin therapy. Beyond these consistent findings, some country-specific associations were also observed: lack of health insurance was associated with a higher SWD in the US sample, whereas a correct understanding of statin benefits was associated with a lower SWD in the Japanese sample. However, the optimism-corrected Nagelkerke R² and c-statistics were consistently low (R² < 0.1 and c-statistic < 0.6 for nearly all models), indicating that age, sex, and any single background variable together had limited ability to discriminate and predict different SWD levels.

Table 2.

Association between background variables and the SWD.

Variables	US				Japan
	OR, 95%CI	P value	Nagelkerke R2	c-statistic	OR, 95%CI	P value	Nagelkerke R2	c-statistic
Race/Ethnicity			0.010	0.54			-	-
White	Ref				-	-
Othera	1.34 (0.997 to 1.80)	0.05			-	-
Education			0.006	0.53			0.013	0.56
Undergraduate	Ref				Ref
High School or less	0.83 (0.60 to 1.14)	0.25			1.27 (0.95 to 1.70)	0.10
Graduate	0.72 (0.48 to 1.08)	0.12			0.98 (0.50 to 1.92)	0.94
Employment			0.017	0.55			0.007	0.55
Working full-time	Ref				Ref
Working part-time	1.10 (0.73 to 1.66)	0.64			0.93 (0.60 to 1.44)	0.74
Unemployed	1.60 (0.93 to 2.74)	0.09			1.21 (0.65 to 2.29)	0.54
Unemployed, not lookingb	0.64 (0.44 to 0.93)	0.02			0.99 (0.70 to 1.42)	0.98
Insurance			0.018	0.55			-	-
Private insurance	Ref				-	-
Public insurance	1.14 (0.84 to 1.53)	0.40			-	-
Uninsured	2.36 (1.40 to 4.02)**	0.001			-	-
Smoking status			0.007	0.54			0.017	0.56
Never a smoker	Ref				Ref
Former smoker	1.04 (0.74 to 1.45)	0.83			1.08 (0.75 to 1.55)	0.69
Currently smoking	0.68 (0.44 to 1.05)	0.08			1.24 (0.84 to 1.83)	0.28
Non-ASCVD comorbidities (HTN, DM, CKD)
No	Ref		0.050	0.59	Ref		0.045	0.59
Yes	0.45 (0.33 to 0.60)**	<0.001			0.47 (0.34 to 0.64)**	<0.001
ASCVD history			0.017	0.55			0.023	0.56
No	Ref				Ref
Yes	0.51 (0.32 to 0.80)*	0.004			0.43 (0.23 to 0.78)*	0.006
Hypercholesterolemia history
No	Ref		0.071	0.60	Ref		0.053	0.60
Yes	0.37 (0.27 to 0.50)**	<0.001			0.44 (0.32 to 0.60)**	<0.001
Received education on blood cholesterol
No	Ref		0.064	0.60	Ref		0.065	0.59
Yes	0.39 (0.29 to 0.53)**	<0.001			0.36 (0.26 to 0.51)**	<0.001
Experience of statin therapy
Never	Ref		0.130	0.62	Ref		0.079	0.58
Past	0.63 (0.35 to 1.12)	0.11			0.41 (0.13 to 1.28)	0.13
Current	0.16 (0.10 to 0.24)**	<0.001			0.20 (0.12 to 0.32)**	<0.001
Self-perceived ASCVD risk
Low	Ref		0.039	0.58	Ref		0.022	0.57
Moderate	0.55 (0.40 to 0.74)**	<0.001			0.68 (0.50 to 0.92)*	0.01
High	0.38 (0.23 to 0.60)**	<0.001			0.61 (0.41 to 0.90)*	0.01
Test on statins’ benefits
Incorrect	Ref		0.010	0.54	Ref		0.027	0.57
Correct	0.25 (0.06 to 0.93)	0.04			0.56 (0.39 to 0.80)**	0.001
Test on statins’ adverse effects
Incorrect	Ref		0.005	0.53	Ref		0.015	0.55
Correct	1.18 (0.78 to 1.78)	0.43			0.83 (0.61 to 1.12)	0.22

^aOther includes Indian (Asian Indian subcontinent); Middle Eastern, Arab, or Persian; Native American or Alaska Native; Native Hawaiian or Pacific Islander; and multiracial.

^bStudent, retired, homemaker, or disabled.

All models were adjusted by age and sex. Nagelkerke R² and c-statistic were optimism-corrected based on 1000 bootstrap resamples. An OR less than 1 indicated a higher likelihood of selecting a smaller SWD category, suggesting greater willingness to accept statins at a smaller effect size. In contrast, an OR larger than 1 indicated a higher likelihood of selecting a larger SWD category, suggesting lower willingness to accept statins. For descriptive purposes, * and ** indicate p≤0.01 and p≤0.001, respectively, highlighting potentially noteworthy associations. No formal significance threshold was applied given the exploratory nature of the analyses.

3. Relationship between age and SWD.

Figure S4 shows the relationship between age and SWD. In both countries, SWD tended to be lower at higher ages, corresponding to greater willingness to accept statins.

4. Association between cholesterol education and SWD.

Finally, after adjusting for confounders identified by the DAG (Figure S1), prior cholesterol education was associated with a lower SWD with OR=0.62 (95%CI, 0.42 to 0.90, p = 0.01) in the US and OR=0.53 (95%CI, 0.31 to 0.90, p = 0.02) in Japan.

4. Discussion

In this secondary analysis of survey data from the US and Japan, we explored potential sources of variation in SWD for statin therapy to better understand factors underlying treatment preferences. Although exploratory in nature, most analyses were prespecified, and we observed consistent patterns across the US and Japan despite divergent ASCVD risk profiles and clinical communication environments, suggesting the potential robustness of our study results. In both countries, participants who perceived themselves to be at higher baseline ASCVD risks demonstrated greater willingness to initiate statins, reflected in lower median SWDs and higher proportions willing to accept statins at their current effect size. A history of hypercholesterolemia, ASCVD, or other cardiovascular comorbidities, alongside current statin use, and older age, also predicted greater treatment willingness. However, despite these observed associations, the optimism-corrected Nagelkerke R² and c-statistics were consistently low, suggesting that no single background variable can adequately predict SWD, indicating that treatment preferences may reflect a complex interplay of factors. Notably, prior education about blood cholesterol was associated with lower SWD after adjusting for potential confounders, with educated participants having 38% (US) and 47% (Japan) lower odds of selecting a higher SWD category, that is, greater willingness to initiate statin therapy, compared to those without prior education. Taken together, these findings suggest that perceived susceptibility and knowledge of the benefit-harm profile may meaningfully influence treatment acceptance. Efforts to strengthen them through education may therefore represent a practical approach that warrants further investigation as a means of promoting rational prescription incorporating informed patient judgements.

The SWD is a preference-based metric that quantifies the minimum absolute benefit required for an individual to deem a therapy worthwhile, explicitly balancing benefits against harms and burdens. Unlike the minimum important change (MIC, also known as minimum important difference or minimum clinically important difference), which captures within-person changes in outcomes [18,19], SWD is conceptually better suited for between-treatment comparisons essential to treatment selection [20]. As healthcare increasingly emphasizes patient-centered care and public involvement, SWD has gained recognition as a structured tool across diverse clinical contexts [[21], [22], [23], [24], [25]].

Our findings showed that participants with ASCVD-related comorbidities and higher self-perceived ASCVD risk exhibited lower SWD, indicating that perceiving one’s risk as personally relevant and serious is associated with greater willingness to initiate statins. This is consistent with previous studies suggesting that low risk perception, rather than objectively estimated risk, was associated with lower adherence [8,10,26], and that increasing perceived risk was associated with higher willingness to consider preventive therapy [7]. Interestingly, perceived and algorithmically estimated risk are often poorly correlated, and individuals frequently overestimate their ASCVD risk [7]. A similar pattern was observed in our sample: 53.6% of US and 58.6% of Japanese participants rated themselves at moderate or high risk, higher than the 39.8% prevalence of an algorithm-predicted 10-year ASCVD risk ≥7.5% reported among US adults aged ≥40 years, despite our younger sample and the generally lower objective risk in Japan [27]. Differences in risk perception may also help explain discrepancies between specialist recommendations and patient preferences: clinicians, particularly specialists, frequently encounter severe cases, whereas the general population seldom do, potentially resulting in systematically distinct risk intuitions. Along the same lines, the lower SWD observed among older participants in both countries may reflect a similar mechanism, as cardiovascular disease tends to become more personally relevant with age.

In addition to risk perception, our results showed that a correct understanding of statin benefits was associated with a lower SWD (regression analyses: OR in Japan 0.56 (95%CI 0.39 to 0.80), US 0.25 (0.06 to 0.93)), suggesting that treatment-specific knowledge may contribute to treatment acceptance. This association aligns with broader evidence linking health literacy and disease knowledge to preventive health behaviors and medication uptake [10,28]. Notably, although side effects are commonly perceived as a major barrier to treatment, our findings, which are consistent with a previous report, suggest that understanding of statin benefits might be more strongly associated with preferences and behaviors [29]. One possible explanation is that rare adverse effects are often overrepresented in public information environments, potentially inflating perceived harm and influencing treatment decisions. Indeed, even statin product labels appear to overstate harm: a recent individual-participant-data meta-analysis of large double-blind randomized trials found no evidence that statins cause most of the listed adverse effects [30]. In our questionnaire, however, adverse events were presented with evidence-based frequencies, which were objectively low. Thus, concerns about harms may exert less influence on treatment thresholds, potentially explaining why adverse-effect considerations were not strongly associated with SWD variation. In addition, accurate risk perception and health literacy are interrelated constructs: risk appraisal depends on comprehension of probabilities and outcomes, making it difficult to fully disentangle their contributions. Indeed, the greater willingness observed among current statin users may reflect both mechanisms, with lived experience reinforcing personal risk awareness, while simultaneously deepening understanding of statin benefit-harm trade-offs.

Our post hoc analysis found that prior cholesterol education was associated with lower SWD after adjusting for confounders identified through a DAG framework. Although causal inference cannot be firmly established, the persistence of this association after adjustment supports the hypothesis that education may influence treatment thresholds, potentially through simultaneous effects on risk perception and literacy. Education delivered by healthcare professionals may be particularly important, given that statin-related information in online environments is often exaggerated or unreliable [31], and medical risk communication is inherently complex [32]. Educational strategies could be implemented through either community-based programs [33] or individual health checkup consultations [34], both of which may improve understanding of ASCVD risk and statin benefit-harm profiles and thereby influence SWD. This may be especially pertinent in Japan, where rates of prior cholesterol education and comprehension in our sample were notably lower than in the US. As ASCVD remains the leading cause of death in the US, extensive risk factor control programs have been implemented [35], which may partially explain the higher cholesterol education and comprehension observed among US participants. It should also be noted that the US participants demonstrated higher educational attainment, which may have further contributed to this difference. In Japan, the overall population risk is relatively low and preventive education infrastructures may be comparatively less intensive. Strengthening cholesterol-related health literacy may represent a particularly actionable strategy to enhance engagement with preventive therapy.

Our study has limitations of note. As with the primary study, sample representativeness and response quality in online surveys cannot be taken for granted. We incorporated comprehension check questions and provided correct answers to aid participant understanding; however, bias cannot be excluded. Second, this secondary analysis was exploratory in nature and intended to generate hypotheses rather than draw definitive conclusions. Nevertheless, the three subgroup analyses were pre-planned, and to avoid misinterpretation, we refrained from using multivariable regression models containing all available variables to search for “predictors”, as coefficients may vary greatly depending on model specifications. Regarding the relationship between cholesterol education and SWD, although we employed a DAG approach to identify confounders, this analysis was not pre-specified, and there is always room for unmeasured confounding. Nonetheless, these findings may inform future confirmatory studies aimed at developing prediction models for SWD or examining the causal effects of interventional factors on SWD. Third, when estimating the proportion of participants willing to accept statins at their actual efficacy, we used the RR per 1 mmol/L LDL-C reduction. This approach might be imprecise at high baseline ASCVD risks, as individuals with high risk often receive intensified statin therapy which could achieve >1 mmol/L LDL-C reductions, potentially resulting in greater effect sizes.

5. Conclusion

Individual subjective risk perception and understanding of the benefit-harm profile are associated with treatment preferences and acceptance. Educational interventions aimed at strengthening treatment-specific health literacy, whether embedded within shared decision-making or delivered through broader educational initiatives, may represent a practical and potentially modifiable pathway to rational and informed healthcare decision-making.

Central Illustration.

Declaration of Generative AI and AI-assisted Technologies in the Manuscript Preparation Process

During the preparation of the central illustration, generative AI technology (ChatGPT) was used to assist in its development. After using this tool, the authors carefully reviewed and edited the resulting content as appropriate and take full responsibility for the content of the published article.

CRediT authorship contribution statement

Yan Luo: Writing – original draft, Visualization, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Hiroshi Kawakami: Writing – review & editing, Validation, Resources, Investigation. Satoshi Funada: Writing – review & editing, Validation, Investigation. Shoma Ozawa: Writing – review & editing, Validation, Project administration, Investigation. Ethan Sahker: Writing – review & editing, Writing – original draft, Validation, Resources, Investigation. Kenji Omae: Writing – review & editing, Validation, Methodology, Investigation. Kazumichi Yamamoto: Writing – review & editing, Validation, Resources. Osamu Yamaguchi: Writing – review & editing, Supervision, Resources. Aran Tajika: Writing – review & editing, Validation, Resources, Investigation, Funding acquisition. Toshi A. Furukawa: Writing – review & editing, Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Conceptualization.

Declaration of competing interest

YL reports funding from the Japan Society for the Promotion of Science (JSPS Grant-in-Aid for Scientific Research, KAKENHI Grant Number JP22K21112 and JP23K16355) outside the submitted work. YL is employed in the Department of Next-Generation Organ Transplantation, The University of Tokyo Hospital, which is an endowed department supported by the N28 General Incorporated Association, a non-profit organization in Japan. KO reports consultant fees and lecture fees from Astellas, Kyowa Hakko Kirin; and lecture fees from Ono, Kyorin, Pfizer, Nippon Shinyaku, and Kissei, outside the submitted work. OY reports receiving honoraria for lectures outside the submitted work from AstraZeneca, Otsuka, Daiichi Sankyo, Novartis Japan, Bayer Yakuhin, Nippon Boehringer Ingelheim, Ono, Takeda, Medtronic Japan, Alnylam Japan, Pfizer, Kowa, Tsumura, Eisai, Bristol-Myers Squibb, Sumitomo Dainippon, Mochida, Nipro, Toa Eiyo, Kyowa Kirin, MSD, Viatris, Novo Nordisk, Sanwa Kagaku Kenkyusho, Johnson & Johnson, and Janssen; and grants outside the submitted work from Otsuka, Chugai, Taiho, Kyowa Kirin, Mochida, Nippon Boehringer Ingelheim, Kyorin, Nippon Kayaku, Sumitomo, Boston Scientific Japan, Teijin, Mitsubishi Tanabe; and a contract outside the submitted work from Canon Medical Systems. AT reports personal fees from Eisai and Shionogi outside the submitted work. TAF reports personal fees from Daiichi Sankyo, DT Axis, Micron, Shionogi, SONY and UpToDate, and a grant from DT Axis, outside the submitted work; in addition, TAF has a patent 7448125 and a pending patent 2022–082495 and has licensed intellectual properties for Kokoro-app to DT Axis. All the other authors report no competing interests to declare.

Data availability

Relevant data may be available from the corresponding author upon reasonable request and at their discretion.