European Commission logo
Joint Research Center (JRC)
TFA-POL

Markov state transition model to estimate the health impacts and related costs to the public of different trans fatty acids (TFA) intake scenarios at the EU level

HealthPublic PolicyEuropean Unionpublic healthtrans fatty acids (TFA)

Quality & Transparency

Quality

Model uncertainties

Models are by definition affected by uncertainties (in input data, input parameters, scenario definitions, etc.). Have the model uncertainties been quantified? Are uncertainties accounted for in your simulations?

yes
As input values of the model assume deterministic values, the model also performs a probabilistic sensitivity analysis by applying probabilistic distributions to all variables. Each of the distributions used in this analysis were chosen following the current trends and literature recommendations. In addition, there are uncertainties related to the current levels of iTFA intakes (starting point of model calculations) for the EU population. Therefore, deterministic and probabilistic sensitivity analyses for 3 alternative initial iTFA intake scenarios in addition to the base case were carried out, as well. Also, for 1000 outcomes in the probabilistic sensitivity analysis, the probabilities for saving costs and DALYs were calculated for each policy option to test the robustness of the overall outcome of comparing the cost-effectiveness of policy options.

    Sensitivity analysis

    Sensitivity analysis helps identifying the uncertain inputs mostly responsible for the uncertainty in the model responses. Has the model undergone sensitivity analysis?

    yes
    After the paper publication, the model was subjected to a comprehensive sensitivity auditing and uncertainty analysis conducted by an expert group in 2017, which provided suggestions that could improve the development of future similar models.

      Have model results been published in peer-reviewed articles?

      yes
      An article describing the model and the respective results was published in The American Journal of Clinical Nutrition (2016).

        Has the model formally undergone scientific review by a panel of international experts?

        Please note that this does not refer to the cases when model results were validated by stakeholders.

        no

          Model validation

          Has model validation been done? Have model predictions been confronted with observed data (ex-post)?

          yes
          During the impact assessment process, the (SANTE-led) inter-service group (ISG), in which the JRC participated, the JRC model and publication were extensively discussed, in particular the input sources, uncertainties and assumptions made. An external consultant gathered additional data and further ‘validated’ inputs and assumptions used in the JRC model with MS, industry and NGOs. The JRC model was made available to the consultant as it was decided to use the JRC model for the impact assessment and add additional analyses to it, in particular including costs to the food industry (which wasn’t part of the JRC study, which took a public-spending perspective).

            Transparency

            To what extent do input data come from publicly available sources?

            This may include sources accessible upon subscription and/or payment

            Entirely based on publicly available sources

            Is the full model database as such available to external users?

            Whether or not it implies a specific procedure or a fee

            no
            The TFA intake, defined as E%, as a starting point for the model (“today”) was calculated as described in the published paper’s Supplemental Tables 1–3. All other data sources are publicly available and referenced in the paper (table 1).

              Have model results been presented in publicly available reports?

              Note this excludes IA reports.

              yes

              Have output datasets been made publicly available?

              Note this could also imply a specific procedure or a fee.

              no
              Results are available in the published paper.

                Is there any user friendly interface presenting model results that is accessible to the public?

                For instance: Dashboard, interactive interfaces...

                no

                  Has the model been documented in a publicly available dedicated report or a manual?

                  Note this excludes IA reports.

                  no
                  Details about data, assumptions, limitations, model structure and results are available in the published paper.

                  Is there a dedicated public website where information about the model is provided?

                  no

                    Is the model code open-source?

                    no

                    Can the code be accessed upon request?

                    yes