Conclusiones

In the reviewed papers four focal infective sites were specifically addressed more than once: soft tissue (17), respiratory (16), lower gastro-intestinal (GI, 9), and urogenital (3). A further group of studies addressed clinical syndromes without discriminating between location of the primary insult (sepsis [9], acute lung injury [2]).

Review of the data suggests NSAIDs exert divergent effects on the outcome or predisposition to infection in these separate sites (Figure f3.4).

Investigational studies have predominantly sought to test either the non-inferiority of treatment with NSAID alone compared to antibiotic, or the superiority of antimicrobial plus NSAID compared to antibiotic alone in the setting of upper respiratory (URTI) and urinary tract infections (UTI). With both sites COX-inhibition results in clear symptomatic benefit, ameliorating both systemic and local features subjectively.

repeatedly demonstrated to be superior to placebo, non-inferior to antibiotics alone^344,345, equivalent or superior to alternative agents e.g. paracetamol331,334,346, and to vary within class: certain NSAIDs being more efficacious^333,334. As described earlier, there is little evidence in the literature to support an assertion that NSAID administration leads to objective (outcome category 2 or 3) benefit over antibiotic administration alone or alternate analgesic/anti-inflammatory agents in URTI or UTI.

There is no inference of an antimicrobial effect separate to their known properties.

Evidence of benefit in lower respiratory tract infections (LRTI) is less clear. A recent randomized placebo controlled trial of ibuprofen versus amoxicillin-clavulanic acid in uncomplicated acute bronchitis, found no significant reduction in symptoms by either agent compared to placebo and a greater adverse effect burden in the antibiotic arm³⁴⁷. In cases of community acquired pneumonia (CAP) two separate case control studies described either a significant independent association between acute NSAID prescription and the development of pleuropulmonary complications (37.5% NSAID vs. 7% no NSAID, OR 8.1; 95%CI 2.3-28)³⁴⁸ or, paradoxically, reduced need for intensive care treatment (5% vs. 24.4%) and a shorter hospital stay in those previously taking anti-platelet drugs (84% aspirin)³⁴⁹.

Figure f3.4 Comparative reported outcomes from NSAID administration during acute infection at different sites, including ‘severe infection’, as assessed by a pre-defined outcome matrix (Table 1). Included studies were grouped into severe infection (Panel A: sepsis, ICU, ALI), soft tissue infection (Panel B), respiratory (Panel C: URTI, LRTI including pneumonia), urogenital (Panel D), and gastro-intestinal (GI, Panel D).

Almost universally, reviewed papers describing soft tissue and lower GI pathology, relate NSAID administration to an increased incidence and/or severity of infection. As previously described, severe skin and soft-tissue infection (necrotizing cellulitis, necrotizing fasciitis, gangrenous myositis) is primarily associated with NSAID use for a primary insult, predominantly varicella zoster infection, but may include minor injury^350,351 or surgery³²⁰. Case control studies estimate the additional risk of complications from NSAID use prior to or during paediatric varicella infection between 4.9 (rate ratio, 95% CI 2.1-11.4³⁵²) and 11.5 (odds ratio, 95% CI 1.4-96.9³⁴¹), with increased risk persisting in the adult zoster population but at a lower degree of magnitude (rate ratio 1.6, 95% CI 1.0-2.2³⁵²). One included single centre retrospective case series additionally suggested an increased risk of development of streptococcal toxic shock syndrome with acute NSAID administration³⁵³.

Of the reviewed papers six suggest an association with NSAID use and complications of diverticular disease. Two case-control studies report increased risk of pericolic abscess formation, generalized peritonitis following perforation, bleeding, or fistula formation in patients with diverticular disease taking NSAIDs^354,355, with a further indicating their use is related to an increased risk of recurrence in mild colonic diverticulitis managed conservatively³⁵⁶. Three cohort studies complement these, describing increased risk of symptomatic diverticular disease, specifically bleeding in NSAID users^357-359. Despite this, paracetamol, an alternate analgesic, was seen to have a similar effect in one study³⁵⁸, and the finding of increased diverticulitis is neither universal nor pronounced (multivariable hazard ratio 1.25; 95% CI 1.05-1.47)³⁵⁹. Separately, one small single centre retrospective case-control study of 84 patients suggests an association between acute appendicitis and NSAID consumption in those over 50 years of age (OR 6.5, 95% 2.1-8.8) including a non-significant increased relative risk of death (OR 2.1, 95%CI 1.1-4.1)³⁶⁰, however this was not supported in a larger study³⁶¹.

Pathogen

The vast majority of included studies described either the prescription of NSAIDs in the context of bacterial or viral infection or their development during NSAID use. A much smaller number related explicitly to mycobacterial (1 tuberculosis, 1 leprosy), protozoan (1 malaria) or fungal (1 corneal ulcer) infection. Assessment for discrepancies in effect of NSAID on different pathogen groups was inherently conflated with infective site: viral infection being the predominant cause of respiratory tract infections, and bacteria being responsible for secondary soft-tissue and urinary

tract infections. Determination of gradated efficacy or harm in this context was felt disingenuous given the absence of clearly divergent results between categories.

Comparisons of specific sub-classes of pathogen (e.g. Gram-positive versus negative bacteria, aerobic versus anerobic) were rendered largely impossible due the paucity of studies specifying detailed microbiology. Case reports and series surrounding severe soft-tissue infection were the exception, frequently identifying Group A beta-haemolytic Streptococcus (GAS) as the cause of necrotizing fasciitis.

Other bacteria implicated in these cases included Streptococcus pyogenes, Streptococcus pneumoniae and Staphylococcus aureus. Whilst it could be hypothesized that NSAIDs led to increased susceptibility to Gram-positive bacteria, the inclusion of studies advocating the efficacy of NSAIDs in reducing the incidence of Gram-positive infections in separate settings³⁶² and the comparable microbiological profile of patients with CAP taking NSAIDs or not³⁴⁸, suggests otherwise.

Severity

Infection severity was necessarily graded in a binary manner; included papers being split into those exploring the effect of NSAIDs on either sepsis or patients admitted to critical care, or not. 2 papers addressing severe pneumonia with septic complications were additionally reviewed. In total 14 manuscripts were classified as relevant to

‘severe infection’.

Three RCT’s explored the effect of NSAID administration on sepsis. One small trial (29 patients) designed to assess safety of IV ibuprofen administration in severe sepsis demonstrated a significant anti-pyretic effect but no corresponding haemodynamic, respiratory or metabolic alterations³⁶³. The second reported no significant effect of IV lornoxicam on any parameter, including biochemical and major pro-inflammatory cytokine concentrations compared to placebo³⁶⁴. Both studies indicated that short term NSAID administration was safe with no adverse GI, renal or hepatic sequelae.

The major interventional study to date in this population, randomising patients with sepsis and at least one organ failure to either placebo or IV ibuprofen (10mg/kg [max 800mg] every 6 hours for 8 doses), found significant reduction in urinary prostanoid metabolites with corresponding decreases in physiological variables (including oxygen consumption and lactic acidosis), but no statistically significant improvement in survival (37% NSAID vs. 40% placebo)²⁷⁶. This is supported by a small

treatment for patent ductus arteriosum were noted to have lower circulating C-reactive protein and IL-6 after day 4 compared to those not on NSAIDs³³⁶. No comment on outcome was made in this paper.

The null findings of RCTs contrast strongly with reviewed observational reports of NSAID use in severe infection. Prescription of anti-platelet regimes for cardiovascular, cerebrovascular or peripheral artery disease was associated with decreased risk of mortality in various subgroups of a mixed intensive care population (ORs 0.04-0.34, 93.5% of patients taking aspirin), even including those with high bleeding risk³⁶⁵. Patients with severe sepsis receiving anti-platelet agents for at least two days during their ICU stay (96.7% including aspirin) were reported to have a significantly reduced risk of in-patient mortality (In-ICU: OR 0.56, 95% CI 0.37-0.84, In-hospital: OR 0.57, 95% CI 0.39-0.83)³⁶⁶. This is supported by a large retrospective cohort study relating aspirin usage in the 24-hour period around the time of the systemic inflammatory response syndrome or sepsis recognition with reduced in-hospital death (10.9% aspirin vs. 17.2% propensity matched non-users, and 27.4%

vs. 42.2% respectively)³⁶⁷. There was also evidence within the reviewed papers that this beneficial effect may extend to other NSAIDs, however this was abrogated if they were used in conjunction with aspirin³⁶⁸.

Observational data regarding a potentially preventative or protective role for COX-inhibiting drugs against severe infection was mixed. As previously discussed, one case-control study indicated that prior anti-platelet therapy (predominantly aspirin) may reduce the requirement for ICU admission and length of hospital stay in patients with CAP³⁴⁹, however pre-hospital aspirin therapy was not found to be independently associated with a reduction in either the development of sepsis, acute lung injury (ALI)/acute respiratory distress syndrome (ARDS) or in-hospital mortality in a separate prospective cohort study³⁶⁹. Despite concerns that acute and/or chronic NSAID use during evolving bacterial community-acquired infection may lead to the development of sepsis and septic shock raised in one article³⁴⁸ this was not borne out in a multi-centre case-control study³⁷⁰.

Two included studies investigated the effect of existing aspirin therapy on the development of ALI specifically, the most common cause of which is severe sepsis. A small, single-centre cohort study reported reduced incidence of ALI (12.7% vs. 28%;

OR 0.37, 95% CI 0.16-0.84) in patients admitted to ICU with known risk factors (34%

presenting with pneumonia, 27% meeting sepsis criteria) on anti-platelet therapy (99% aspirin), however there was no effect on ICU or in-hospital mortality³⁷¹. A

similar reduction in incidence of ALI was also noted in a multicenter international cohort study, however this did not reach significance on multivariate analysis and after adjusting for propensity to receive aspirin (Cochran-Mantel-Haenzel pooled OR 0.7; 95% CI 0.48-1.03)³⁷².