Longevity and registry data
The best sources of information on knee replacement failure rates are national registers. One of the oldest is the Swedish Knee Arthroplasty Register (myknee.se). It has been collecting data on patients in Sweden since 1975. The largest registers are in the UK (njrcentre.org.uk) with records of over 1 million total knee replacement operations since April 2003, and in Australia (aoanjrr.sahmri.com) with over 600,000 knee replacements recorded since September 1999.
All national registers provide publicly available reports both online and by publishing an annual report. They include records of surgery in state and privately funded hospitals.
Although these reports appear to present a wide array of data on the success of knee replacement surgery, they in fact focus on one principal measure of success: the rate or probability of revision surgery according to the number of years since surgery.
Every record of an operation includes details such as the patient’s gender, age, other health issues, type and severity of arthritis and the implant design that was used. Every record of a revision operation also includes the reasons that the revision surgery was required.
With large datasets, the registries can divide the data into a wide range of comparative groups so that potentially contributing factors and reasons for failure can be better understood. It is important to understand that ‘big data’ like this does not prove cause or effect. But it does allows us to consider what factors could influence the likelihood that a design is likely to become a long-lasting success for a given patient.
Factors that are known to influence the revision rate include:
- Primary disease – osteoarthritis is less likely to need further surgery than inflammatory arthritis
- Age – younger age groups tend to have a higher rate of revision
- Year of operation – operations in the last 2 decades are associated with lower rates of revision than previous years (but this is not changing much in recent year-to-year comparisons)
- Gender – gender differences do exist but are more complex and depend on other factors
- Type of implant – partial knees and posterior stabilised knees tend to have higher rates of revision than other types, but this may be influenced by their patient demographics
- Material – new materials can provide better results in some designs but they are not always better than other designs using established materials
- Method of fixation – overall, using cement to fix a knee implant is still associated with a lower revision rate but many newer designs are being used without cement
- Brand of implant – this is the area that has generated the most interest over the years. The National Joint Register (njrcentre.org.uk) was in fact initiated after a particular design of hip was recalled for an excessive number of failures. The registry was seen as a mechanism to detect implants that were not performing in line with state-of-the-art.
What to look for
The rate or probability of revision is based on known experience. It is reported as a percentage of the total number of patients in a group at a series of time points. A good starting point is to compare the published revision rate for a type of knee replacement for your own age group and gender. You can also compare the overall rates of revision for particular designs of knee. Although designs of knee are not usually separated to gender or age groups, it is possible to see what the average age and proportion of males to females in the reported group is.
By necessity the probability of revision is based on certain assumptions designed to address what are known as confounding factors. One example is how to account for the people no longer available for monitoring. Whether their knee still does or, if they have since passed away, would have continued to function is unknown. The calculation also takes into account how many patients were in the dataset to start with. How likely would the calculated value be the same if the dataset were much larger? The level of certainty or confidence in a value is displayed as a percentage range, usually in brackets next to the main value: based on known information, there is a 95% chance that the ‘real’ rate of revision is somewhere between the lower and upper values.
If the study group is small or the time point being reported does not include many patients (this could be because the device is new or not mainstream and widely available), the confidence interval will be ‘wide’ (they differ considerably from the stated probability). If a group includes many hundreds of knees the level of confidence will be high (the lower and upper percentages will be very close to each other).
It is worth appreciating that if a device that has a higher probability of failure than another device you are interested in, the confidence interval can help you understand whether or not the difference is meaningful. On the whole when comparing two designs, if one of the stated probabilities is between the lower and upper values of the other, the available data does not prove there is a statistically significant difference between the designs. If, however, the probabilities and confidence intervals are separated, it is safe to say that one has a demonstrably lower rate of failure than the other.
The National Joint Registry for England, Wales, Northern Ireland and the Isle of Man also publishes patient information leaflets about the NJR and its data. These can be found on the registry website www.njrcentre.org.uk.