Evaluation of the effectiveness of hip and knee implant models used in Catalonia: a protocol for a prospective registry-based study

A prospective observational design will be drawn up based on data from the Catalan Arthroplasty Register (RACat) to describe and evaluate the functioning of the different prosthesis models used in Catalonia between 2005 and 2017. Since 2005, the RACat has systematically gathered information on knee and hip replacements implanted in publicly owned institutions in Catalonia. This information is further supplemented by data drawn from three additional sources: (1) the Central Insurance Registry (RCA) which provides information about patients’ vital records (alive, change of address, death) and enables us to track them; (2) the Minimum Basic Data Set for Hospital Discharge (MBDS-HD), which serves to complete patient data, treatment episodes and diagnoses; and (3) the RACat Prosthesis Catalogue, a compendium of information regarding the specific prosthesis models that were implanted that allows us to see their characteristics based on information provided by the manufacturers.

Firstly, the models of knee and hip prostheses used during the study period will be identified and classified according to the prosthesis group to which they belong (considering prosthesis and fixation types and, for total hip replacements, the bearing surface). Subsequently, we will analyse the results and compare the behaviour of each of the models to their reference group. Lastly, aspects and limitations related to the described process will be evaluated and taken into account for the routine implementation of this methodology within the RACat (Fig. 1).

The study population to be considered will comprise all patients undergoing knee or hip arthroplasty operations at hospitals belonging to the SISCAT (Integral Public Health System of Catalonia) between January 2005 and December 2016. In the years following, after executing the study and verifying its utility, data analysis will continue on new cases and prosthesis models as they are added to the RACat.

The RACat currently has information available on more than 46,000 primary hip arthroplasties and 5000 revisions, and over 60,000 primary knee arthroplasties and 6500 revisions [28]. Since it was first founded, the registry has recorded more than 140 different models for total and partial hip replacement components, and around 130 different models for knee replacements, indicating the considerable variation in the models employed in Catalan public health institutions. Furthermore, according to the data from the MBDS-HD [2], irrespective of the original cause for surgery or type of prosthesis used, in 2015, publicly owned health institutions in Catalonia performed 9299 primary knee replacements and 7074 primary hip replacements, which accounted for 16.1% of all knee and hip replacements performed in Spain.

Consideration will be given to all primary prosthesis models employed throughout the period studied. The analyses will exclude cases where it is impossible to identify the operated joint and their laterality (if the operated joint is in the left or the right side), the gender of the patient undergoing the operation, their age, the presence of prior comorbidity or the specific prosthesis model. We will also exclude prosthesis models whose volume of implantation during the study period would not have allowed us to establish valid and reliable conclusions regarding the survival and possible deviation between observed and expected values. Before starting, a minimum frequency of 10 prostheses corresponding to the model implanted throughout the period studied will be established for the inclusion of a specific model within the analysis due to limitations connected with the range of the confidence intervals. This limit may vary depending on the distribution of frequencies of the specific models identified included in the RACat. Furthermore, the analyses will exclude revisions caused by infection, since the premature failure of these prostheses cannot be attributed to a poor outcome of the implant.

For the grouping of the models of prosthesis and detection of results potentially better or worse than expected in comparison with the reference group, three aspects will be considered: the type of prosthesis, the fixation and, for total hip replacements, the bearing surface.

As regards the type of prosthesis, we will use the same classification system previously used by both the RACat and by studies conducted within the context of other international joint replacement registries with extensive, well-established track records [6‐8], classifying prostheses according to the joint operated on, as indicated in Table 1.

Concerning the type of fixation used for the prosthesis, the following methods will be considered: cemented (all components cemented for both hip and knee prostheses). non-cemented (none of the components cemented for either hip or knee prostheses), hybrid (only the stem cemented for hip prostheses or only the tibial component for knee prostheses) and reverse hybrid (only the acetabulum cemented in hip prostheses or only the femoral component in knee prostheses).

To analyse total hip replacements, we will also consider the bearing surface between the different components of the prosthesis, such as metal on metal, metal on polyethylene, ceramic on ceramic and ceramic on polyethylene. Moreover, in total hip prostheses comprising different components, the analysis will also take into account the combinations noted in the stem and cup models.

We will consider the following patient characteristics: gender, age and comorbidity at the time of the operation, evaluated based on the Charlson comorbidity index (CCI) adapted for the International Classification of Diseases, 9th Edition, Clinical Modification (ICD-9-CM) [29, 30]. The CCI is a compound physical health measurement used to predict patient mortality on the basis of the comorbidity levels they reveal and is considered an objective measure of the general state of the patient’s health. The index is calculated by assigning points based on the weighted value of each of the 19 constituent conditions present, thereby serving to calculate a general morbidity index. Given the distribution of the points scores of the variable and the characteristics of the population studied, the scores obtained will be considered on a categorical basis (0 absence of comorbidity, 1–2 intermediate comorbidity, ≥ 3 high comorbidity).

In order to achieve the objectives, the following analyses will be performed sequentially:

Firstly, the models and combinations used over the course of the period of study will be described. For each model or combination, the total frequency implanted will be calculated, along with the relative frequency according to prosthesis group and year. Furthermore, whenever the frequency of the specific prosthesis model or combination allows, the Kaplan-Meier method will be used to calculate the crude revision risk for any cause (except infection periprosthetic fracture and luxation), at 1, 3 and 5 years, with corresponding 95% confidence interval (95% CI).

Subsequently, to identify prosthesis models whose results are better or worse than expected, two different but mutually complementary methodological approaches will be employed consecutively. As the main standard to be followed, we will use the methodology proposed by de Steiger et al. [20] within the context of the Australian Registry (AOANJRR), simplified and adapted for use in the RACat. To do this, the risk of revision will be evaluated, irrespective of the reason but excluding infection, using Cox models adjusted for gender, age and CCI. Additionally, the influence of dead as a competing event with the revision of the prosthesis will be evaluated, and in the event that this fact will be verified, we will recalculate the model, considering the patient’s death as a competitive event.

Secondly, as a complementary analysis to the above methodology, a methodological approach based on standardised replacement ratios (SRR) will be used in accordance with the methodology proposed by the NJR also adapted for the specific characteristics of the RACat [27]. In this approach, a SRR between the number of replacements observed and expected will be calculated for each prosthesis model. The expected replacements will be estimated using the sum of individual replacement probabilities, after adjusting a multi-variable Poisson regression model using age, gender and comorbidity (CCI) as the adjustment variables. A SRR greater than 1 indicates that there have been more replacements than would be expected for this prosthesis model in relation to the average for equivalent models included in the RACat. In order to ensure the validity of the analyses, only those implanted models whose volume allows for subsequent comparisons (taking a priori as the minimum reference volume n = 10) will be included. Furthermore, we will use funnel plots to graph the SRR (y-axis) against the expected number of replacements (x-axis). Confidence intervals of 95% (95% CI) and 99% (99% CI) will be constructed, deeming that a prosthesis above or below 95% CI or 99% CI has a result that could be better or worse than expected with regard to the mean for its group. This outlier status indicates that the number of replacements performed is significantly lower or higher than the number of replacements expected with regard to the mean figure for the group (2 and 3 standard deviations, respectively, for CI 95% CI and 99% CI) [21, 23, 25]. In order to control the possible effect that the hospital might have on the results of the prostheses, a sensitivity analysis will be conducted, calculating the ratio of revision for each model with an extreme SRR with regard to the rate for other prostheses for each hospital.

Once both types of analysis have been conducted, and taking into account as the reference results those obtained on the basis of the first approach, the models obtained will be compared to determine if their results are potentially better or worse than expected. As such, the results obtained on the basis of the second methodology proposed will serve firstly as an element facilitating the visualisation of the results, and secondly as a confirmation element if the first approach reveals results that could potentially be better or worse than expected in comparison with the corresponding type.

Lastly, models identified by the first and/or second approach will be evaluated by the advisor committee of the RACat, a panel of experts in orthopaedic surgery and traumatology working with the RACat. As result of this evaluation and once a consensus about the different models will be reached, we will proceed to identify those with results potentially better or worse than expected, according to both clinical and statistical criteria, having considered possible sources of bias from both perspectives.

Once the process has been completed, and with the goal of routine implementation within the RACat, all aspects of the development of this study will be considered, in order to identify possible key points. This will entail a retrospective analysis of the difficulties and limitations encountered within each of the tasks performed so we can identify the most important aspects to take into account regarding implementation.