A revised framework for clinical sample size
The Center for Medical Device Evaluation of China’s National Medical Products Administration published its revised registration review guidance for HER2 gene amplification reagents on October 13, 2026. The announcement is dated October 12. It specifies no separate effective date or transition period, and the document is advisory rather than a legally binding regulation.
The 2017 guidance specified at least 1,000 clinical samples, including at least 400 with HER2 immunohistochemistry (IHC) results of 2+. After those requirements were met for one tumor type, each additional tumor type required at least 300 samples, including at least 120 IHC 2+ samples.
The 2026 minimum sample-size section no longer uses those fixed counts. It recommends a single-group performance-goal approach to estimate the numbers of HER2 amplification-positive and amplification-negative samples separately. The recommended clinically acceptable performance criterion, P₀, is at least 90% for both positive and negative agreement. Both groups should consist mainly of IHC 2+ samples.
The 90% figure is a parameter in the sample-size framework, not a standalone approval threshold. Applicants should justify the expected agreement, significance level, statistical power and other parameters, account for sample exclusions, and ensure representation of relevant patient characteristics. Other, more appropriate sample-size methods may be used with justification.
Removing fixed counts from this section does not establish a general reduction in study workload. Different tumor indications require separate trial designs and sample-size estimates. Final sample numbers depend on the statistical assumptions, intended population and available sample composition. Protocols built around the previous primary-tumor-plus-additional-tumor structure warrant reassessment against this approach.
Cancer indication and companion drug define the evidence
The guidance covers fluorescence in situ hybridization (FISH) assays for HER2 amplification in tissue sections from surgical resections or biopsies, including single- and dual-signal assays. It applies to registration and change-registration applications and focuses on breast and gastric cancer. Other cancers require additional evaluation suited to the claimed scope. Bright-field methods such as CISH and SISH may use applicable portions while supplementing evidence for their own methodologies.
Clinical evaluation is illustrated using trastuzumab. For other companion drugs, applicants should provide evidence under the relevant companion diagnostic guidance. An assay co-developed with a new anticancer drug may use the associated drug trial data, provided the report identifies the investigational assay, its role in the trial and its relationship to drug efficacy.
The comparator should be the original companion diagnostic for the anticancer drug claimed by the applicant. The assay’s cutoff should be set with reference to that comparator’s sensitivity to support agreement. Establishing the companion drug and intended population is therefore necessary to assess comparator suitability. A shared molecular target alone does not establish that clinical evidence can be transferred between intended uses.
Borderline results require analytical and clinical evaluation
The revised guidance calls for analytical performance and cutoff studies for each claimed cancer type. Cutoffs should be established with reference to the latest relevant clinical guidance for that cancer and verified using the assay’s proposed interpretation rules. Samples should include amplified, non-amplified and borderline tissues, such as HER2/CEP17 ratios around 1.8–2.2 or average HER2 copy numbers around 4–6 per cell. These are study sample ranges, not a universal diagnostic threshold independent of the full interpretation rules.
Borderline samples also belong in precision studies, alongside amplified and non-amplified samples. Evaluation should consider sources of variation such as batches, days, operators and readers. FISH performance depends on region selection, signal counting and interpretation as well as probe signals, making reproducible reading procedures part of the evidence.
Clinical analysis should report positive, negative and overall agreement with 95% confidence intervals for both all samples and the IHC 2+ subset. The proportion of indeterminate results and its effect on agreement should be assessed. Discordant samples may undergo confirmation with a third-party reagent or another appropriate method, but those confirmation results should not be incorporated into the original statistical analysis. This preserves the meaning of the initial comparison.
Align tissue handling with the proposed instructions
Where both surgical and biopsy specimens are claimed, both should be included in the clinical study. The guidance specifies paraffin-embedded tissue fixed in 10% neutral buffered formalin and calls for evaluation of collection, fixation, section thickness, transport and storage conditions. Sample stability studies should cover the intended cancer types.
The instructions for use should explain these conditions and the rules for slide validity, region selection, initial cell counts, indeterminate results and HER2 amplification heterogeneity. Comparing actual study procedures with proposed labeling can identify evidence gaps that a sample-size revision alone would miss.
Alternative methods remain available where they meet applicable regulatory requirements and are supported by detailed research and validation. The revised framework provides scope for justified study design, with each indication supported by its own study design and evidence.
Sources
- Primary official source人表皮生长因子受体2基因扩增检测试剂注册审查指导原则（2026年修订版） (opens in a new tab)Center for Medical Device Evaluation, National Medical Products Administration
- Supporting source国家药监局器审中心关于发布《人表皮生长因子受体2基因扩增检测试剂（荧光原位杂交法）注册审查指导原则（2026年修订版）》等2项注册审查指导原则的通告（2026年第28号） (opens in a new tab)Center for Medical Device Evaluation, National Medical Products Administration
- Supporting source人表皮生长因子受体2基因扩增检测试剂盒（荧光原位杂交法）注册技术审查指导原则（2017年第209号附件3） (opens in a new tab)China Food and Drug Administration