The Cochrane Collaboration has assessed the analgesic efficacy of many non-opioid analgesics in the setting of postoperative pain. These metaanalyses describe the effects of single doses of these drugs by the NNT50, i.e. the number of patients needed to treat to achieve 50 % pain relief in one patient compared to placebo. Lower NNTs suggest better efficacy.

For the worldwide commonly used coxibs the following NNT50 were found:

These results suggest analgesia comparable or better than non-selective NSAIDS such as ibuprofen 400 mg (NNT 2.5), ketorolac 20–60 mg (NNT 1.8) or naproxen (NNT 2.7) [33].

Similar results were achieved with regard to duration of action, where again coxibs were comparable to naproxen and superior to other NSAIDs [33].

Coxibs have been used widely as adjuncts to opioids in the setting of multimodal analgesia. Here a very well designed study showed that after laparoscopic surgery a combination of parecoxib/valdecoxib compared to placebo combined with PCA morphine improved analgesia and reduced opioid consumption as well as number and severity of opioid adverse effects, all by around 30 % [11]. Similar studies with similar results have been performed after orthopedic surgery such as hip [47] and knee joint replacements [18]. A recent meta-analysis shows no advantage of NSAIDs over coxibs here [30], although other analyses showed some minor differences [9, 28].

Overall the current data suggest that coxibs and non-selective NSAIDs at appropriate doses have no clinically relevant differences in efficacy and are exchangeable. This conclusion is in line with the respective key message of a highly regarded document on scientific evidence in acute pain management [27].

There was a wide spread belief in the literature that short-term use of nonselective NSAIDs would not lead to GI adverse effects [20], although we all have seen morbidity and even mortality as a consequence of hemorrhage and perforation. Studies contradicted this impression; a study exposing elderly patients to non-selective NSAIDs for only a few days was even discontinued in view of the unexpectedly high rate of ulcers [15].

A summarizing analysis of a number of trials assessing exposure of elderly patients to nonselective NSAIDs and coxibs for 5–7 days found considerable rates of GI ulcers on gastroscopy with ketorolac (30 %) and naproxen (20 %) compared to parecoxib (1 %). The latter rate is comparable to placebo (2 %) [39]. A meta-analysis on short-term use in soft tissue injury comes to similar findings of a significant advantage of coxibs over nonselective NSAIDs [19].

These data confirm a superior GI safety of coxibs even with short-term use; these advantages might be even greater in patients with empty stomach kept nil by mouth in the postoperative setting, where there might be also a greater risk of stress-induced ulceration.

As outlined above, the withdrawal of rofecoxib has lead to a farreaching discussion on the potential CV toxicity of coxibs. The lack of platelet inhibition and effects on endothelium could have explained a class effect of coxibs with increased rates of thromboembolic events including myocardial infarction and stroke [10]. However, these concerns could not be confirmed for the other coxibs and there is now a consensus, that CV toxicity is a class effect of all anti-inflammatory compounds with no correlation to COX-2 seletivity [48]. This has been confirmed by revised statements of regulatory authorities such as FDA and EEA. A meta-analysis found even, that the coxibs celecoxib and parecoxib/valdecoxib show reduced CV adverse events in comparison to nonselective NSAIDs [32]. All these data are anyway the results of long-term use and therefore a specific look at postoperative use is justified.

Our group performed a meta-analysis of CV adverse events with perioperative use of parecoxib and its active metabolite valdecoxib compared to placebo [40]. 32 studies were identified including over 8500 patients randomized and found no statistically significant difference in CV events, even when stratified for cardiac risk factors.

In conclusion, neither long-term nor short-term use of the coxibs celecoxib and parecoxib/valdecoxib increases CV events in comparison to nonselective NSAIDs; cardiac risk factors should not influence the choice between these two therapeutic groups [38]. However, it needs to be noted that all anti-inflammatory compounds are contraindicated after cardiac surgery; the increased risk shown here in studies is most likely the result of ‘shear stress’ on platelets in the roller pumps of extracorporeal circulation [1].

As coxibs do not affect the isoenzyme COX-1, which is exclusively expressed in platelets, these compounds should have no effect on platelet function and thereby reduce the risk of perioperative hemorrhage. This is in contrast to non-selective NSAIDs, which have a profound inhibitory effect on platelet aggregation.

These theoretical considerations have been confirmed in in vitro studies of platelet function, where coxibs have an effect comparable to placebo even with long-term exposure, while nonselective NSAIDs lead to significant impairment [35].

These findings could be translated into clinical studies, where celecoxib and placebo had comparable effects on all parameters of blood loss after total knee joint replacement [18] and major plastic surgery [43]. Previous studies had shown an increased rate of bleeding with nonselective NSAID use e.g. in an RCT after ENT surgery [31] and a metaanalysis after major surgery in comparison to placebo [29].

An RCT comparing a nonselective NSAID with a coxib after hysterectomy and breast surgery respectively found significantly reduced blood loss (in the range of 1/3 to 2/3) with coxib use [16]. This was also shown in an RCT in tonsillectomy, where ketoprofen increased postoperative bleeding in comparison to celecoxib [34].

In conclusion, coxibs have no effect on platelet aggregation and are here comparable to placebo, which translates into reduced perioperative hemorrhage compared to non-selective NSAIDs.

Nonselective NSAIDs can induce bronchospasm in patients with hypersensitivity reactions, often called aspirin-induced asthma. A systemic review confirmed that coxibs do not carry this risk [50].

Although concerns about detrimental effects of NSAIDs on bone healing are widely held, the scientific situation remains unclear here.

Even with regard to non-selective NSAIDs in the setting of animal models, the results are inconclusive and often contradictory. The data are more confusing with regard to the coxibs [2]. In human studies, it is difficult to show any clinically relevant effects of non-selective NSAIDs on bone healing and fracture repair [8]. The exception seems to be spinal fusion; there are limited data suggesting a dose-dependent effect of ketorolac [12, 36] and diclofenac on impaired union [26]; in contrast coxibs had an effect comparable to non-use in one of these studies [36]. However, these studies are all retrospective and a meta-analysis concludes that in high quality studies no statistically significant effect of any NSAID on rates of union could be identified [8]. Another metaanalysis suggests that high-dose ketorolac might be the only harmful compound here [24].