HCV Mosaic-B

What is the HCV-MOSAIC risk score and how was it developed?

The HCV-MOSAIC risk score is a behavior-based risk assessment tool developed from the prospective Dutch MSM Observational Study of Acute Infection with hepatitis C (MOSAIC). It was created using multivariable logistic regression modeling to identify predictors of HCV acquisition. The score consists of six factors associated with HCV transmission: condomless receptive anal intercourse, sharing sex toys, unprotected fisting, injecting drug use, sharing straws during nasally administered drug use, and having an ulcerative sexually transmitted infection. The score is calculated by summing the beta coefficients of these factors when present in an individual's behavior profile .

What populations has the HCV-MOSAIC risk score been validated for?

The HCV-MOSAIC risk score was originally developed and validated for predicting primary early HCV infection in MSM with HIV. Subsequently, it has been evaluated for predicting HCV reinfection in the same population. Validation studies have been conducted using data from the original MOSAIC cohort in the Netherlands and externally validated using data from the Recently Acquired HCV Infection Trial (REACT) conducted in Australia .

What are the key risk factors included in the HCV-MOSAIC score?

The HCV-MOSAIC risk score includes six key behavioral risk factors, each weighted by their beta coefficient from the original regression model:

These risk factors were identified as significant predictors of HCV transmission and are assessed based on self-reported behaviors in the preceding 6-12 months .

How does the predictive capacity of the HCV-MOSAIC score differ between primary HCV infection and HCV reinfection?

The HCV-MOSAIC score demonstrates different predictive capacities for primary infection versus reinfection. For primary early HCV infection, the score achieved an area under the receiver operating characteristic (AUROC) curve of 0.82, with sensitivity and specificity of 78.0% and 78.6%, respectively, using a cut-off value of ≥2.0. For HCV reinfection, the predictive capacity was slightly lower, with an AUROC of 0.74 (95% CI = 0.63-0.84) in the training dataset. Using the same cut-off value (≥2.0), sensitivity was 70.4% (95% CI = 49.8-86.2%) and specificity was 59.2% (95% CI = 47.3-70.4%) for reinfection prediction. In external validation for reinfection prediction, the AUROC decreased to 0.63 (95% CI = 0.53-0.74), with sensitivity of 44.0% (95% CI = 24.4-65.1%) and specificity of 71.2% (95% CI = 61.8-79.4%) .

What methodological considerations should be accounted for when applying the HCV-MOSAIC score in different populations?

When applying the HCV-MOSAIC score to different populations, researchers should consider several methodological factors:

• Behavioral data collection: The risk factors must be assessed using standardized questionnaires covering the appropriate time frame (6-12 months).

• Cultural and contextual differences: Risk behaviors may vary across populations and settings.

• Missing data handling: In the external validation study, two risk factors (sharing sex toys and unprotected fisting) were not measured, potentially affecting score performance.

• Cut-off value recalibration: Consider whether the established cut-off (≥2.0) is appropriate for the specific population or whether recalibration is needed.

• Group sex dynamics: Research indicates that group sex significantly affects the ROC curve (ΔROC = 1.14, 95% CI: 0.19-2.09), suggesting this factor may modify the predictive capacity of the score .

How do covariates influence the performance of the HCV-MOSAIC risk score?

Parametric ROC regression analysis revealed that certain covariates can significantly influence the performance of the HCV-MOSAIC risk score. Notably, group sex had a significant effect on the ROC curve at any given 1-specificity value (ΔROC = 1.14, 95% CI: 0.19-2.09). In fact, when including only those who engaged in group sex, sensitivity reached 100% (although specificity decreased to 29.4%). Other covariates, such as the number of HCV reinfections (ΔROC = 0.02, 95% CI: -0.87, 0.90) and having any anonymous partner (ΔROC = 0.42, 95% CI: -1.18, 2.01), did not significantly affect the ROC curve. These findings suggest that group sex behaviors may be particularly important to consider when utilizing the HCV-MOSAIC score in research contexts .

What statistical approaches should be used to evaluate the performance of the HCV-MOSAIC score in research settings?

When evaluating the performance of the HCV-MOSAIC score, researchers should employ comprehensive statistical approaches including:

How should researchers handle missing data when applying the HCV-MOSAIC score?

Missing data poses challenges when applying the HCV-MOSAIC score. The research demonstrates two main approaches:

• Complete-case analysis: Only including participants with complete data on all six risk factors.

• Restricted score approach: Modifying the score to include only available risk factors. In the external validation study, where data on sharing sex toys and unprotected fisting were not collected, sensitivity analyses were performed by restricting the HCV-MOSAIC risk score in the training dataset to the same risk factors measured in the validation dataset, which yielded comparable results.

What are the optimal questionnaire designs for capturing the behavioral data needed for the HCV-MOSAIC score?

Researchers should design questionnaires that:

• Include all six behavioral risk factors in the score (condomless receptive anal intercourse, sharing sex toys, unprotected fisting, injecting drug use, sharing straws during nasally administered drug use, and ulcerative STI)

• Specify appropriate timeframes (6 months for sexual behaviors, 12 months for drug use and STIs)

• Use self-administered formats to reduce social desirability bias for sensitive behaviors

• Include additional variables that may affect score performance (e.g., group sex, anonymous partners)

• Maintain consistency in question phrasing and response options

• Consider sociodemographic factors (age, ethnicity, education level) that may influence risk behaviors

The MOSAIC study employed self-administered questionnaires with questions about risk behaviors referring to the preceding 6 or 12 months, which proved effective for data collection in this sensitive area .

How do sociodemographic factors correlate with HCV-MOSAIC risk scores?

These findings suggest that while age may be a factor in some populations, the HCV-MOSAIC risk score's performance is relatively stable across different sociodemographic groups .

What are the implications of different cut-off values when implementing the HCV-MOSAIC score in research studies?

Cut-off value selection has significant implications for the implementation of the HCV-MOSAIC score:

• Original validated cut-off (≥2.0):

  • For primary infection: Sensitivity 78.0%, Specificity 78.6%

  • For reinfection (training dataset): Sensitivity 70.4%, Specificity 59.2%, proportion correctly classified 0.62

  • For reinfection (validation dataset): Sensitivity 44.0%, Specificity 71.2%, proportion correctly classified 0.66

• Alternative cut-off (≥1.2) identified in post-hoc analysis for reinfection:

  • Training dataset: Sensitivity 77.8%, Specificity 57.9%, proportion correctly classified 0.63

  • Validation dataset: Sensitivity 44.0%, Specificity 66.7%, proportion correctly classified 0.63

The proportion of individuals recommended for HCV-RNA testing also varies by cut-off: 48.5% of the study population when using ≥2.0 and 51.5% when using ≥1.2 in the training dataset. This highlights the trade-off between sensitivity, specificity, and resource allocation that researchers must consider when selecting cut-off values .

How do the individual risk factors in the HCV-MOSAIC score compare between cases and controls in predicting HCV reinfection?

Individual risk factors show varying distributions between cases (those with HCV reinfection) and controls across both the training and validation datasets:

*Statistically significant difference (p<0.05)
RAI: receptive anal intercourse; NAD: nasally administered drug; STI: sexually transmitted infection

In the training dataset, four risk factors (condomless RAI, sharing sex toys, unprotected fisting, and injecting drug use) were significantly more common among cases than controls. In the validation dataset, only condomless RAI showed a significant difference. These differences in risk factor distributions between datasets may partially explain the lower performance of the score in the validation dataset .

What modifications to the HCV-MOSAIC score might improve its predictive capacity for HCV reinfection?

Based on current evidence, several modifications might enhance the predictive capacity of the HCV-MOSAIC score for HCV reinfection:

• Incorporation of group sex as an additional risk factor, given its significant effect on the ROC curve

• Recalibration of risk factor weights (beta coefficients) specifically for reinfection prediction

• Development of population-specific cut-off values

• Inclusion of additional behavioral or clinical predictors not currently in the model

• Development of time-varying models that account for changes in risk behaviors over time

Research investigating these modifications would be valuable for improving the tool's performance in identifying individuals at risk for HCV reinfection .

How can the HCV-MOSAIC score be integrated into clinical and public health screening algorithms?

The HCV-MOSAIC score shows potential for integration into clinical and public health screening algorithms in several ways:

The relatively good sensitivity of the score (70.4% at cut-off ≥2.0 in the training dataset) makes it potentially useful for case-finding in settings where ongoing HCV transmission remains a concern and reinfection incidence is high .

What are the implications of the HCV-MOSAIC score for designing behavioral interventions to reduce HCV transmission?

The HCV-MOSAIC score has several implications for behavioral intervention design:

• Targeted education: The six risk factors in the score identify specific behaviors that should be addressed in prevention messages.

• Risk reduction counseling: Individualized feedback based on score components could help prioritize which risk behaviors to modify.

• Harm reduction approaches: For behaviors that may be difficult to eliminate (e.g., drug use), the score highlights the importance of harm reduction (e.g., avoiding sharing of injection equipment).

• Group-level interventions: Given the significant effect of group sex on the ROC curve, interventions targeting group sexual encounters may be particularly effective.

• Combination prevention: The score supports a comprehensive approach addressing both sexual and drug-use transmission routes.

By identifying the behaviors most strongly associated with HCV transmission, the HCV-MOSAIC score provides a framework for developing and evaluating behavioral interventions in high-risk populations .