Duration of Prophylactic Strategies
In the absence of reliable evidence about how long to employ these prophylactic strategies, it is the opinion of this work group that patients and physicians discuss the duration of prophylaxis.

We recognize the diversity of opinion concerning the clinical importance of DVT as an isolated event or as a surrogate outcome for PE or post-thrombotic syndrome, (for further discussion, please see the Methods section), and understand that for clinical, and sometimes for even medico-legal reasons, DVT prevention may be the clinician’s immediate concern. There is moderate evidence to suggest that pharmacological agents and/or mechanical compression devices reduce DVT rates in patients undergoing elective knee or hip arthroplasty. This is why we are suggesting prophylaxis. Readers of this guideline should recognize, however, that the available, published evidence does not establish whether these prophylactic strategies affect  rates of all-cause mortality, fatal PE, symptomatic PE, or symptomatic DVT in patients undergoing elective hip or knee arthroplasty.

We also note that the present recommendation for prophylaxis is of a “Moderate” (rather than “Strong”) grade partly because it is based on a surrogate outcome we do not consider “critical” (we considered  major bleeding, pulmonary emboli, and all cause mortality as “critical,” and symptomatic DVT, any DVT, and proximal DVT as not critical). The “critical” outcomes are all patient-oriented. The non-critical outcomes are not.

The inability to recommend a specific prophylactic strategy is a direct result of the network meta-analyses we performed. We performed numerous such analyses with sensitivity analyses that included separately analyzing data from patients who underwent hip and knee arthroplasty, analyzing these data combined, evaluating the impact of study quality on the results, and by comparing the  results of each prophylactic strategy to placebo (or no treatment) and, when placebo/no treatment data were not available, comparing the results of  each strategy to results obtained with enoxaparin (as discussed in the Methods section, this use of two comparators allows us to check the logical consistency of our models). The results of these analyses did not consistently suggest that any one strategy is preferable to another (please see Figure 38 - Figure 55 and Table 32 - Table 34; and, for the results of our sensitivity analyses, see Appendix XV).

We also analyzed data on other outcomes but, due to lack of data,  network meta-analysis was not possible for them. In total, then, our analyses of the different prophylactic strategies is comprised of 112 high-or medium quality randomized controlled studies that enrolled patients undergoing elective hip and/or knee arthroplasty (see Appendix XIII, Table 53). As with the network meta-analyses, the data did not suggest that any specific prophylactic strategy was superior or inferior.

Part of the reason that current data do not permit a conclusion about specific prophylactic strageties is that, in our final network meta-analyses, no pharmacological agents showed a statistically significant effect in preventing all-cause mortality, symptomatic pulmonary emboli, symptomatic DVT, and major bleeding, when data from  hip and knee studies were analyzed separately or when they were combined. This may be because these events are rare. In addition, infection rates and re-operations (for any reason) were not reported. Reoperations due to bleeding were reported, but were often part of the study authors’ definition of major bleeding.

Many of the commonly used agents such as sodium warfarin and various low molecular weight heparinoids did not show efficacy for preventing VTED. This may be partially explained by the lack of comparison studies with placebo controls and by the rarity of the events of interest. In the final model with PE as the outcome, there were 181 events among 42,390 patients across 25 trials, and only 3 of these trials had a placebo or no prophylaxis arm.

There were a limited number of studies that evaluated mechanical compression devices.  In one study on total hip arthroplasties,48 there was a lower risk of major bleeding in the mechanical group. However, this study was only of moderate quality, partially because only 37% of the compression group had this device alone, with the remainder of the patients receiving low dose aspirin (81 mg/day) as well. There were also difficulties with the comparability of the control and intervention groups (that some of the studies we examined were not of high quality is another reason why the present recommendation is of  “Moderate” strength).

In some analyses of mechanical compression device studies, less bleeding was found in comparison to no treatment. This may not appear intuitively logical, but might be occurring because of problems with randomization and the patient populations which may not be generalizable to the standard population of patients typically undergoing total hip and knee arthroplasties.  The effect may also be occurring for some presently unknown physiological reasons.  Other potentially confounding factors with these studies are enumerated below.

Conclusions about specific prophylactic strategies are also difficult because, in addition to the above-mentioned challenges posed by the rarity of the events of interest and the lack of reporting of critical outcomes, the available studies:
  • Enrolled a select group of patients and did not necessarily include patients who had a high risk for VTED or bleeding and may not be representative of a typical patient population
  • Used different drug doses (e.g. Enoxaparin at 30 mg bid vs. 40 mg per day).
  • Used different timing of administration of agents (short-term vs. longer-term dosing)
  • Used different routes of administration
Comparing different prophylactic strategies is difficult because there is a paucity of placebo-controlled trials because of early acceptance of prophylaxis being the standard of care.

Also, we are unable to recommend specific pharmacologic agents and/or mechanical devices because the results of our analyses with DVT as the outcome were not robust on sensitivity analyses. Due to the rarity of the critical outcomes of interest and the limited number of placebo-controlled trials, we had to rely on the analysis of DVT (i.e., any DVT), a surrogate measure, to evaluate the relative efficacy of the prophylactic strategies. However, the results of these analyses depend on the structure of the model used, as agents shown to significantly reduce the occurrence of DVT in one model are often not statistically significant in an alternate model (see Table 97 in Appendix XV).

Some clinical practice guidelines make recommendations about the duration of pharmacologic prophylaxis. The available evidence is partially from manufacturer-funded trials, and is of only one agent. The latter is particularly problematic because the potential differences in the risks and benefits of various pharmacological agents may become more prominent as the duration of prophylaxis increases.  We are, therefore, reluctant to make such a recommendation until more is known about the relative risk/benefit profiles of these different agents. Rather, the work group recommends that patients and physicians discuss the appropriate duration of prophylaxis for each individual situation. This physician-patient discussion is low cost and consistent with current practice.

As of April 1, 2011, several of the analyzed agents are not approved for marketing or the treatment of any medical condition in the United States. The United States Food and Drug Administration’s (FDA) current policy regarding disclosure of marketing applications can be found in “Current Disclosure Policies for Marketing Applications” on the FDA website.

We excluded some studies we retrieved for this recommendation. The reasons for doing so are shown in Appendix XIV, Table 62).