#HEORInsights

The domain of healthcare research has witnessed remarkable advancements in the evaluation of various treatment modalities for specific diseases. One such pioneering approach, the Network Meta-Analysis (NMA), has proven instrumental in comparing multiple treatments assessed in randomized trials. Conventionally, NMAs primarily rely on Aggregate Data (AD) extracted from study publications, limiting the scope of analysis to the available summarized data. However, an innovative alternative that has gained traction in recent years is the integration of Individual Participant Data (IPD) into the NMA framework. This approach involves procuring raw information recorded for each participant in a study, allowing for a more comprehensive and nuanced evaluation of treatment effects.[1,2]

A pivotal strength of incorporating IPD into NMA lies in the enhanced quality and scope of information available for each trial. Unlike AD, IPD circumvents the limitations posed by the published data, thereby facilitating a comprehensive analysis. This comprehensive approach fosters the standardization of outcome and covariate definitions across trials, reducing the potential for reporting biases. By harmonizing diverse definitions of outcomes, IPD enables a more robust analysis of continuous scales, yielding more precise and reliable conclusions. Additionally, the collaborative nature of IPD NMA projects, involving direct engagement with trial investigators, facilitates a more thorough identification of relevant trials, thereby reinforcing the reliability of risk assessments and data analyses.[2-4]

Another compelling advantage of employing IPD for NMAs is the ability to examine and plot distributions of covariates across trials. This granular approach empowers researchers to discern systematic differences in variables that might influence measures of relative treatment effect. With access to a broader set of recorded covariates, IPD enables a more detailed comparison of how covariate distributions vary across trials, thus aiding in the assessment of their potential impact on network heterogeneity and inconsistency. Notably, this approach contributes significantly to the evaluation of the consistency assumption, a critical factor in ensuring the reliability of NMA results.[4]

Furthermore, IPD allows for the standardization and improvement of the analysis of each trial, enabling researchers to define their own estimands of interest. By standardizing effect measures across trials, IPD facilitates the utilization of advanced analytical approaches, particularly in the modeling of time-to-event data and the handling of missing data. The utilization of a multivariate NMA approach becomes feasible, enabling the simultaneous comparison of treatments across multiple outcomes, while accounting for correlations among these outcomes. This multifaceted analysis not only enhances the precision of inferences but can also alter the ranking of treatments, thereby providing a more comprehensive perspective for clinical decision-making.[4]

Using IPD in NMA can also allow adjusting for prognostic factors in the analysis of each trial. This aspect is instrumental in improving the power to detect treatment effects, especially in scenarios where stratified randomization has been employed. Moreover, the inclusion of treatment-covariate interactions facilitated by IPD, allows for a more comprehensive understanding of how treatment effects may vary across different patient populations. This nuanced analysis paves the way for tailored treatment recommendations, aligning with the evolving paradigm of personalized medicine and patient-centric care.[4]

However, the integration of IPD into NMA projects is not without its share of challenges. The substantial time and effort required to procure, validate, harmonize, and synthesize IPD often pose significant logistical hurdles. Negotiating data-sharing agreements and maintaining effective collaboration with trial investigators demand meticulous planning and clear communication to ensure the smooth execution of the research. Additionally, the inclusion of data from diverse sources may introduce data heterogeneity. Performance of IPD NMA would necessitate expertise to handle both the IPD as well as perform NMA: thus, IPD NMA requires higher and complicated statistical analytic techniques, adding to the resource-intensiveness of the technique. Mitigating concerns regarding availability bias, stemming from the unavailability of IPD from certain studies, remains a persistent challenge that requires careful consideration and methodological adjustments. Furthermore, data privacy, confidentiality, and ethical considerations arise due to the transmission of sensitive information contained in the IPD, adding another layer of complexity.[4]

To address the challenges discussed above, various innovative methodologies have emerged, such as multilevel network meta-regression, which extends the IPD NMA framework to incorporate both IPD and AD, ensuring robust and accurate estimates across diverse target populations. Additionally, matching-adjusted indirect comparison, simulated treatment comparison, and predictive-adjusted indirect comparison have been proposed as alternative approaches to account for limited IPD, although their applicability remains confined to specific study designs and populations.[4-7]

In conclusion, the integration of IPD into NMA represents a paradigm shift in evidence synthesis, offering unprecedented insights into treatment effects and patient outcomes. Despite the challenges posed by the complex nature of IPD meta-analyses, the benefits of utilizing this approach far outweigh the obstacles, fostering a more holistic and informed approach to clinical decision-making and healthcare policy formulation. As the healthcare landscape continues to evolve, the judicious utilization of IPD in NMA is poised to redefine the trajectory of evidence-based medicine, leading to more effective and tailored healthcare interventions for diverse patient populations.

Become A Certified HEOR Professional – Enrol yourself here!

Learn AI Potential For Writing Manuscripts – Enrol yourself here!

References

1. Dias S, Ades AE, Welton NJ, et al. Network meta-analysis for decision-making. John Wiley & Sons; 2018 Jan 8.
2. Riley RD, Tierney JF, Stewart LA, editors. Individual participant data meta-analysis: a handbook for healthcare research. John Wiley & Sons; 2021 May 24.
3. Riley RD, Phillippo DM, Dias S. Network Meta‐Analysis Using IPD. Individual Participant Data Meta‐Analysis: A Handbook for Healthcare Research. 2021 Apr 22:347-73.
4. Riley RD, Dias S, Donegan S, et al. Using individual participant data to improve network meta-analysis projects. BMJ evidence-based medicine. 2023 Jun 1;28(3):197-203.
5. Chaimani A. Conduct and reporting of individual participant data network meta-analyses need improvement. BMC medicine. 2020 Jun 2;18(1):156.
6. Phillippo DM, Dias S, Ades AE, et al. Multilevel network meta-regression for population-adjusted treatment comparisons. Journal of the Royal Statistical Society Series A: Statistics in Society. 2020 Jun;183(3):1189-210.
7. Phillippo D, Ades T, Dias S, et al. NICE DSU technical support document 18: methods for population-adjusted indirect comparisons in submissions to NICE.