Protocol and dataset: Registration-Based Three-Dimensional Measurement of Removable Partial Denture Frameworks

Purpose Removable partial denture frameworks must fit the mouth precisely. Poor fit contributes to loss of retention, uneven loading of abutment teeth, plaque accumulation, mucosal inflammation and eventual fracture. Since digital design and manufacture entered routine practice, three-dimensional surface registration has become the usual way of quantifying how well a framework fits: two digitised surfaces are aligned computationally, and the distance between them is reported as a measure of accuracy. This project examines that measurement process rather than the frameworks themselves. Its purpose is to determine whether the numbers this literature produces mean the same thing from one study to the next. The question arises because the word “accuracy” is applied to at least three different comparisons. A framework may be measured against its own computer-aided design file, which quantifies how faithfully the manufacturing process reproduced the intended geometry. It may be measured against the master cast on which it was built, which quantifies adaptation to a physical object. Or it may be measured against the patient’s own tissues, which quantifies clinical fit. These answer different questions, and a framework can score well on one and poorly on another. Alongside this, the algorithm used to align the two surfaces can itself alter the reported distance, because an unconstrained best-fit distributes deviation across the whole surface and can conceal local misfit. Design A systematic review of primary studies published between January 2018 and April 2026 that quantified removable partial denture framework geometry using three-dimensional registration. Five databases were searched: PubMed/MEDLINE, Embase, Scopus, Web of Science and Cochrane CENTRAL. Screening and full-text assessment were conducted by two reviewers. Every included study was classified by the reference against which deviation was computed, by the alignment strategy applied, by the scanning approach, by the outcome measure, and by the acceptability threshold used and where that threshold came from. Completeness of methodological reporting was recorded as a primary outcome: whether each study stated its sampling density, iteration limit, convergence criterion, reference region and scanner accuracy. Risk of bias was appraised with instruments appropriate to each design, at item level for laboratory studies and at domain level for clinical studies. Analyses were conducted within reference-standard strata. Values obtained against one type of reference were at no point compared with values obtained against another. Findings Forty-eight studies met the inclusion criteria. Seventeen measured deviation from a CAD design, 22 from a master cast or die, eight from intraoral tissue, and one from both in separate arms. All were described in the source literature as measuring accuracy. Alignment was unconstrained best-fit across the whole registered surface in 19 studies and was not characterised beyond an algorithm name in nine. The parameters that govern where a registration algorithm settles were almost never reported: a convergence criterion appeared in one study and an iteration limit in two. An acceptability threshold was stated in 35 studies. Only four cited a primary clinical investigation. Tracing the most widely propagated threshold to its origin showed that it derives from descriptive measurements of frameworks that were clinically accepted in the 1980s and 2000s; neither source study derived or validated a criterion separating acceptable from unacceptable fit. An observed distribution has become a threshold through repeated citation. No combination of reference standard, alignment class and outcome measure contained enough methodologically concordant studies to support pooling. What this means Reported deviation values in this field are not interchangeable, and quantitative synthesis across them is not currently defensible. This is a limitation of measurement reporting rather than a criticism of the underlying research: within-study comparisons remain valid, and findings that recur across studies using different protocols are strengthened by that diversity. The project proposes a minimum set of items that studies of this kind should report, so that future work can be compared within defined strata. Materials deposited here The review protocol, the complete study-level extraction dataset, database-specific search strategies, the list of reports excluded at full text with reasons, item-level and domain-level risk-of-bias appraisals, and a study identification ledger reconciling every record assessed. Note on timing This protocol and dataset are deposited retrospectively. The protocol was agreed in writing between the reviewers before screening began and was not amended thereafter, but it was not registered with PROSPERO or deposited publicly at that time. It should not be read as a prospective registration. The review is unregistered, and this is stated as a limitation in the manuscript.