Complications of hip and knee replacement surgery: a pharmaco-epidemiological approach

Abstract

In the general discussion, several clinical and methodological aspects of this thesis were discussed. We compared the observed study findings between the two data sources (i.e. Denmark or United Kingdom), and described explanations for the (dis)concordances. The discussion was concluded with some clinical and methodological recommendations. Our studies show that THR and TKR surgery briefly are accompanied with an increased risk of several potentially fatal complications, including VTE, AMI, stroke, and GI bleeding. Although evidently elevated, the risk of clinical harms are generally limited to the first 2 to 6 weeks (with the exception of VTE). Given the well-established benefits of THR and TKR surgery, we still promote the surgery in patients who require them, and we consider this as a relatively safe procedure with a high benefit-risk ratio. Although the average excess risk of adverse outcomes following THR and TKR surgery was low in absolute figures, we call for careful risk assessment in patients undergoing this surgery. A 20-fold increase for a low-risk patient may be acceptable, but for an individual with an already high baseline cardiovascular risk, this may be a discussion worthy issue. Risk prediction tools could be helpful for such risk assessments, and we therefore promote the development of these tools. The consistent findings on the potential benefits of bisphosphonates in preventing implant failure are promising, and this should be further investigated in randomized clinical trials. In the second part of our discussion, we share some methodological recommendations and points of interest to take into consideration when performing observational studies. In particular, we strongly promote the use of time-dependent cohort studies (or even casecontrol studies) whenever possible. Our examples showed strong deviations in study results when the data were analyzed using a time-fixed cohort design. It is crucial to feel comfortable with the strengths and limitations of the available databases. Aspects such as under recording (differential or non-differential), availability of covariates, and calendar time span should be put against the type of research question and point estimate (i.e. relative or absolute risk estimate). Dealing with confounding can be difficult and, in our examples, may be limited by the quality and quantity of available confounders. Residual confounding may therefore still be an issue, regardless of the utilized technique to deal with confounding. We feel it is therefore important to focus on causality / timing pattern analyses, rather than to look at overall risk estimates solely. In conclusion, from a clinical point of perspective, THR and TKR surgery briefly boosts the risk of several potentially fatal complications. Although the excess risk is generally limited to 2 to 6 weeks following surgery, we promote careful risk assessments, as such relative increases may be substantial in patients already at elevated risk of developing one or more of these negative outcomes. The use of prophylactic drug regimens (e.g. prolonged thromboprophylaxis, proton pump inhibitors to prevent GI bleeding, antiplatelets to lower stroke rates, and bisphosphonate use to lengthen implant survival) should be further explored in randomized clinical trials. From a methodological point of view, we found that simple choices in study design (e.g. time-dependent versus time-fixed cohort) may vastly influence study findings. We could not find a major impact of other methodological differences (e.g. confounder handling techniques and data source), but we cannot exclude this possibility in other research settings in observational studies. Hence, the robustness of design choices in epidemiological studies need to be further explored.