AMNOG dossiers can require hundreds or even thousands of statistical tables for a single clinical study. Different populations, endpoints, subgroups and extensive safety analyses can substantially increase the number of required outputs. The challenge is therefore not limited to conducting the statistical analyses. Sponsors need processes that can generate, review and update large numbers of analyses consistently, while producing the outputs required for the German G-BA dossier.

This becomes particularly challenging when global statistics and programming teams first produce English-language analyses that are subsequently transferred into German G-BA tables. Every additional handover increases coordination requirements and makes late changes more difficult to implement.

Why AMNOG dossiers can require thousands of statistical analyses

The statistical requirements of an AMNOG dossier extend well beyond the primary and secondary endpoints of a clinical study. Patient-relevant outcomes, including morbidity, mortality, health-related quality of life and adverse events, need to be analysed and presented according to the requirements of the German benefit assessment.

Subgroup analyses and potentially multiple relevant patient populations add further complexity. Even one additional population may require substantial parts of the analysis package to be repeated.

The overall volume of analyses can therefore result from several dimensions:

• relevant patient populations,
• efficacy and quality-of-life endpoints,
• extensive safety analyses,
• subgroups,
• different endpoint operationalisations and analysis time points,
• and the table formats required for the German assessment.

For complex Phase III programmes, the combination of these requirements can result in several thousand outputs.

Operational planning should therefore consider more than the number of endpoints in a study. The relevant question is how many combinations of population, endpoint, analysis and subgroup ultimately need to be generated and quality-controlled for the AMNOG dossier.

Global clinical study analyses and G-BA tables are not the same

Global statistical analyses from a clinical study cannot necessarily be transferred unchanged into the German AMNOG dossier. The German benefit assessment has specific requirements for analyses, presentation and tabulation.

In many organisations, this results in two consecutive processes. Global biostatistics and programming teams first produce the English study outputs. Relevant results are then selected and transferred into the tables required for the German AMNOG dossier.

This separation has important operational consequences. Results need to remain consistent across different outputs, German and English tables need to align, and changes to the underlying analyses need to be reflected across all affected tables.

Manual or semi-automated transfer processes are particularly challenging at scale. As the number of outputs increases, tracing and implementing changes consistently across all affected tables becomes increasingly difficult.

Why collaboration between Market Access, Biostatistics and Statistical Programming matters

A Statistical Analysis Plan (SAP) specifies the planned statistical analyses. For a large AMNOG analysis package, however, a formal specification alone may not resolve every question that arises during implementation.

Wording can be interpreted differently, and study data can contain characteristics that were not fully anticipated when the original specifications were written. Direct communication between statisticians and programmers can help resolve these questions before they lead to multiple review cycles.

AMNOG projects introduce another interface. Local Market Access teams understand the requirements of the German benefit assessment but may have limited direct access to global programmers. In some organisations, even communication with global biostatistics takes place through several organisational layers.

This creates an interface between three perspectives:

1. Market Access defines which evidence is required for the AMNOG assessment.
2. Biostatistics translates these questions into appropriate statistical analyses.
3. Statistical Programming generates the required outputs from the clinical study data.

The greater the organisational separation between these functions, the greater the potential coordination and review burden.

Late changes can affect thousands of AMNOG tables

The limitations of large AMNOG analysis packages become particularly visible when requirements change shortly before submission or during the assessment process. A revised population or comparator-related change may affect considerably more than a handful of tables.

If a patient population is redefined, baseline characteristics, efficacy analyses, safety analyses and subgroup analyses may all need to be rerun. A change that appears relatively limited from a methodological perspective can therefore trigger a substantial regeneration of the analysis package.

One example from AMNOG project work illustrates the potential scale. Following an IQWiG assessment, a patient population had to be redefined. Approximately 10,000 tables consequently needed to be regenerated. Using an automated analysis workflow, the analyses and tables could be updated within three days, while differences between the outputs were systematically identified.

The example illustrates an important distinction between manual and reproducible analytical processes: For large AMNOG analysis packages, responsiveness depends not only on how quickly the original tables can be produced, but on how quickly the entire package can be regenerated in a controlled manner.

Automation of AMNOG analyses is about more than speed

Automation in AMNOG analyses is often discussed primarily as an efficiency measure. Speed matters, but it addresses only part of the problem.

A reproducible analytical workflow can support four central requirements:

1. Consistency: Analyses and tables are generated from the same defined data and analytical processes.
2. Reproducibility: Changes can be systematically applied to the affected outputs.
3. Traceability: Differences between versions of the analysis package can be identified systematically.
4. Scalability: Large numbers of outputs can be generated without treating each additional table as a separate manual process.

Traceability becomes particularly important when an analysis package contains several thousand tables. After a change, knowing that the outputs have been regenerated is not enough. Project teams need to understand which results changed, where those changes occurred and which differences are relevant to the dossier.

Data visualisation and structured comparisons between analysis versions can therefore become part of the quality-control process. They can help statistics and Market Access teams identify relevant changes without manually comparing hundreds or thousands of old and new tables.

German and English AMNOG tables should come from the same analytical process

Another potential source of error arises when English analyses and German G-BA tables are produced sequentially in separate processes.

An integrated approach can generate the required German and English outputs from the same underlying analyses. This removes the need for a separate manual transfer of results between different table formats.

The distinction lies in the workflow:

In a sequential approach, global analyses are generated first. Relevant results are subsequently selected for Germany, transferred into local tables, reviewed and adjusted again when the analyses change.

In an integrated approach, the required outputs are generated from a common analytical process and can be reproducibly regenerated when changes occur.

For complex AMNOG dossiers, this distinction can become more important than the speed of producing the original set of tables.

An end-to-end process connects statistical analysis with the AMNOG dossier

An end-to-end approach to AMNOG analyses connects statistical specification, programming, tabulation and dossier requirements within a common workflow.

This does not mean that statistical and Market Access responsibilities become interchangeable. Their respective expertise remains distinct. The objective is to reduce unnecessary handovers and define requirements jointly at an early stage.

The process therefore starts before the German tables are generated. Early planning should address questions such as:

• Which populations will be required for the German benefit assessment?
• Which endpoints and operationalisations need to be represented?
• Which subgroup analyses are required?
• Which outputs will ultimately be needed for the dossier?
• How will changes and analysis versions be tracked?
• How can German and English tables be generated consistently?

The technical implementation of the analyses consequently becomes part of early dossier planning rather than a separate task at the end of the project.

Why flexible statistical workflows become even more important with JCA

The Joint Clinical Assessment (JCA) further increases the importance of scalable statistical processes. Different populations and comparators can create multiple PIC configurations, for which relevant outcomes then need to be analysed.

European and national evidence requirements also do not arise completely independently of one another. Sponsors therefore need to consider how statistical evidence can be planned and operationalised to support both European analyses and subsequent national benefit assessments efficiently.

This is not only a question of generating tables. It also concerns collaboration between global and European functions, local Market Access teams, biostatistics and Statistical Programming.

For indications involving multiple populations, numerous endpoints and extensive subgroup analyses, an analytical process that is already difficult to manage under AMNOG may face additional pressure as JCA requirements are added.

AMNOG analyses require responsiveness, not just a completed table package

The quality of a large AMNOG analysis package should not be assessed solely by whether all required tables are available at the planned submission date. The process also needs to remain responsive when requirements change.

New assessment questions, revised populations or changes in the relevant comparator situation can require substantial re-analysis. Processes that depend heavily on manual transfers and multiple organisational handovers can quickly come under pressure in these situations.

An integrated and reproducible analytical process reduces these dependencies. Biostatistics, Statistical Programming and Market Access can implement changes on a common analytical basis and systematically understand their impact.

For AMNOG dossier preparation, one question therefore becomes increasingly important: Can the team not only produce thousands of tables, but also change thousands of tables in a controlled and traceable way when the requirements change?

Frequently asked questions about statistical analyses for AMNOG dossiers

Why can an AMNOG dossier require thousands of statistical tables?

An AMNOG dossier can require thousands of tables because analyses may need to cover multiple patient populations, patient-relevant endpoints, safety outcomes and subgroups. The total number of outputs is driven by the combinations of populations, endpoints, analyses and subgroup characteristics rather than by the number of endpoints alone.

Can global clinical study tables be used directly in an AMNOG dossier?

Global clinical study tables cannot necessarily be used unchanged for the German AMNOG dossier. The German benefit assessment has specific requirements for the analyses, presentation and tabulation of clinical evidence. Additional analyses or differently structured outputs may therefore be required.

Why are manual transfers into German G-BA tables a problem?

Manual or semi-automated transfers create an additional interface between the original statistical analyses and the German dossier tables. At scale, this increases the effort required to maintain consistency, perform quality control and implement changes across hundreds or thousands of outputs.

What are the benefits of automated workflows for AMNOG analyses?

Automated AMNOG workflows can improve scalability, consistency, reproducibility and traceability. Their main advantage is not limited to producing the initial tables faster. A reproducible workflow also allows large analysis packages to be regenerated systematically when populations, specifications or other requirements change.

Why should German and English tables be generated from the same process?

Generating German and English tables from the same analytical process reduces separate transfer steps and helps maintain consistency between the outputs. When an underlying analysis changes, the affected outputs can be regenerated from a common analytical basis rather than manually updated across separate table sets.

How can changes during an AMNOG assessment affect the statistical analysis package?

A change to a relevant patient population or analysis requirement can affect large parts of an AMNOG analysis package. Baseline characteristics, efficacy, safety and subgroup analyses may all need to be recalculated. For complex studies, one methodological change can therefore affect hundreds or thousands of outputs.

What roles should be involved in planning AMNOG analyses?

AMNOG analysis planning requires close collaboration between Market Access, Biostatistics and Statistical Programming. Market Access brings the requirements of the German benefit assessment, Biostatistics defines appropriate analytical methods, and Statistical Programming translates these specifications into reproducible outputs.

Why does JCA increase the importance of scalable statistical workflows?

JCA can introduce multiple population and comparator configurations that need to be addressed with relevant outcome analyses. Sponsors may therefore need to manage European and subsequent national evidence requirements within increasingly complex and time-sensitive analytical programmes. Scalable and reproducible workflows can make these requirements easier to operationalise.