Datasets come first
Data quality is the problem most practitioners name first. Every Zynthari analysis starts with an analysis-ready dataset.
Why start with the data
A model is only as good as the dataset under it. Sponsor files rarely arrive ready: units change between sites, covariates are named a dozen ways, and samples below the limit of quantification are coded as text.
Fixing that by hand takes days, and it is hard for anyone to check later. So we build the dataset with you first, and keep a log of every change.
What every intake checks
- Units
- Concentration, dose and time units are identified and made consistent. Every conversion is logged.
- Covariate mapping
- Sponsor columns are mapped to standard variables such as weight, age, sex and creatinine clearance. You confirm the mapping.
- Transformations
- Time after dose, dose records and flags are derived by standard scripts, never edited by hand.
- Below the limit of quantification
- BLQ samples are flagged and handled by the rule in your analysis plan.
- Values against the protocol
- Dose times, infusion lengths and sampling times are compared with the protocol, with the page cited. On the roadmap
QC under written SOPs
Each dataset is checked under written SOPs: record counts are reconciled against the source files, derivations are checked a second time, and every change is logged with its reason.
The output is the dataset your analysis needs: a NONMEM-ready file, or CDISC ADaM PK datasets (ADPC, ADNCA, ADPPK).
Data checks on CSV, SAS XPT and CDISC files are in pilot.