NSAIDs and the Inflammatory Phase of Healing

When a patient walks in with an acutely painful, swollen shoulder or a flared-up elbow, the reflex — theirs and often ours — is to reach for an anti‑inflammatory. It seems intuitive: inflammation hurts, so switching it off should help. But this framing skips a step that matters clinically. Inflammation is not a design flaw. It is the opening phase of a tightly regulated biological programme that produces tissue repair — recruiting immune cells, clearing debris, and setting up the signalling that drives the proliferative and remodelling phases that follow [1]. The real clinical question is not "should we suppress inflammation?" but "what, specifically, does this drug suppress, how much of the repair process does that actually touch, and does changing it help or hinder this particular patient's problem?"

What NSAIDs actually inhibit

Non-steroidal anti-inflammatory drugs work by inhibiting the cyclooxygenase (COX) enzymes that convert arachidonic acid into prostaglandins, prostacyclins and thromboxanes [1,2]. There are two relevant isoforms. COX-1 is constitutively expressed and maintains gastric mucosal protection, renal blood flow and platelet function. COX-2 is the inducible isoform, upregulated at sites of tissue damage and driving the prostaglandin-mediated vasodilation, oedema and pain sensitisation we recognise clinically as inflammation [2]. The therapeutic anti-inflammatory and analgesic effect of NSAIDs comes principally from COX-2 inhibition; it is inhibition of COX-1 that produces the well-known GI, renal and platelet side effects [2]. Non-selective agents (ibuprofen, naproxen, diclofenac) block both isoforms to varying degrees, which is precisely why they carry both the benefit and the side-effect burden.

This matters for how we frame the drug's power: NSAIDs interrupt one enzymatic pathway — the COX/prostaglandin axis — within a much larger inflammatory and reparative network.

One pathway among many

The reparative inflammatory response is orchestrated by a broad cast of mediators well beyond prostaglandins: pro-inflammatory cytokines (IL-1β, IL-6, TNF-α), chemokines, and growth factors (TGF-β, PDGF, VEGF) that between them drive neutrophil and macrophage recruitment, angiogenesis, matrix deposition and the eventual transition to remodelling [1]. In tendon tissue specifically, this cytokine network is now well documented and only partially overlaps with the COX/prostaglandin axis that NSAIDs address [3].

Because NSAIDs act almost exclusively on the COX-derived slice of that network, they leave most of the biological machinery driving repair untouched. Clinically this produces a fairly consistent pattern in the literature: NSAIDs give a modest, time-limited reduction in pain and swelling, but they do not meaningfully redirect the underlying trajectory of soft-tissue healing in most cases [4]. If anything, the evidence on COX-2-selective agents specifically points the other way — several human, animal and in vitro studies suggest early, high-dose COX-2 inhibition can impair tenocyte proliferation and collagen/glycosaminoglycan synthesis during the initial healing window, with non-selective NSAIDs showing a more mixed but generally less detrimental profile [4]. The practical takeaway for the early inflammatory phase is that NSAIDs are, at best, a limited symptomatic tool — worth using for that purpose alone, not as a treatment expected to change healing outcomes (Table 1).

Table 1. NSAIDs in the early inflammatory phase: pros and cons

Where a genuinely potent anti-inflammatory does help: the contained, capsular problem

Corticosteroids are a different order of magnitude — a systemic anti-inflammatory and immunosuppressive intervention, not a single-enzyme blocker. Their clinical value tracks closely with how contained and how genuinely inflammatory the underlying pathology is.

Stiff shoulder (adhesive capsulitis, sometimes still called "frozen shoulder") is a good example of where this works. In its acute, painful stage the driving pathology is a synovial and capsular inflammatory-fibrotic process. Meta-analysis of placebo-controlled trials shows intra-articular corticosteroid injection produces significantly greater pain relief and range-of-motion improvement than placebo in the short-to-medium term (0–24 weeks) [5], and earlier systematic review supports its use for shoulder pain more broadly [6]. Here, dampening a genuinely active inflammatory process in a joint that isn't being asked to immediately return to full mechanical load is a reasonable, evidence-supported trade-off.

Where it backfires: a degenerative tendon is not an inflamed one

Lateral epicondylalgia (LEP, "tennis elbow") is the instructive counter-example, and it illustrates why "potent and fast-acting" is not the same as "helpful." Despite the old name "epicondylitis," the tissue pathology in chronic LEP is degenerative and failed-healing in character — disorganised collagen, neovascularisation, and an absence of the classic inflammatory infiltrate — rather than a straightforward inflammatory tendinitis. This is well captured in the continuum model of tendon pathology, where reactive, dysrepair and degenerative stages describe a structural and cellular problem, not a purely inflammatory one [7].

Injecting a potent anti-inflammatory into that kind of tissue does relieve pain quickly — corticosteroid injection reliably outperforms both physiotherapy and placebo at 4–6 weeks [8–10]. But the mechanism of relief is the problem: a landmark randomised trial found that at 52 weeks, recurrence had occurred in 72% of patients who received corticosteroid injection, compared with 8% after physiotherapy and roughly 10% with a wait-and-see approach [8]. A subsequent placebo-controlled trial confirmed the pattern: participants who received a placebo injection had substantially lower recurrence risk than those given corticosteroid (relative risk 0.23, 99% CI 0.10–0.51) — i.e., the corticosteroid group's recurrence risk was more than four times higher [9]. A systematic review across tendinopathies more broadly reached the same conclusion: corticosteroid injection is the most effective option at 4 weeks, but inferior to other conservative options from about 6 weeks onward, with worse outcomes than placebo at 1 year [10].

Two mechanisms plausibly explain why. First, and most clinically relevant, is a behavioural mismatch: the injection lowers pain without changing the tendon's structural load tolerance, so patients return to activities of daily living and sport at their prior load levels while the tissue's actual repair capacity is unchanged — a mismatch that, in a condition driven by cumulative load management, predictably drives re-injury. Second, there is a direct biological cost: a systematic review of the effects of local glucocorticoid on tendon tissue found reduced fibroblast proliferation and viability and decreased collagen synthesis across the majority of included studies, alongside loss of normal collagen organisation [11]. In a tendon whose fundamental problem is a stalled, degenerative repair process, adding an agent that further suppresses fibroblast activity and collagen synthesis is working against the tissue's actual need — even though it is very good at switching off pain in the short term (Table 2).

Table 2. Corticosteroid injection: outcome depends on the tissue

How the body already regulates its own inflammation

Everything above assumes a drug is entering the picture. It's worth stepping back, because the body already runs its own layered system for both amplifying and shutting down inflammation, and each piece of that system bears directly on the two case studies above.

Resolution isn't just "less inflammation" — it's an active programme in its own right. Acute inflammation doesn't simply fade out as the stimulus wanes; it is actively terminated by a distinct class of endogenous lipid mediators — lipoxins, resolvins, protectins and maresins, collectively called specialised pro-resolving mediators (SPMs) — synthesised from omega-3 and omega-6 fatty acids. These stop further neutrophil recruitment, drive macrophage clearance of debris, and hand the tissue over to the remodelling phase [12]. Clinically, this reframes chronic pain that has run on for months: it's often inflammation that has failed to resolve, rather than inflammation that is simply "too much" — and that resolution pathway is a separate biological target from anything NSAIDs or corticosteroids touch.

Cortisol is, in effect, the body's own corticosteroid. It binds the same glucocorticoid receptor as the injected corticosteroid discussed above and suppresses the same NF-κB-driven cytokine transcription throughout the body [13]. Its release follows a strong circadian rhythm — peaking shortly after waking and falling to its lowest around midnight — while several pro-inflammatory cytokines (IL-6, IL-1) run close to the opposite pattern, rising through the night [14]. This mismatch is the physiological basis for the classic morning stiffness seen in rheumatoid arthritis and other inflammatory joint conditions: overnight, cytokine signalling outpaces the body's own anti-inflammatory cortisol cover [14]. Practically, it's a reminder that how inflamed a joint feels on assessment is partly a function of time of day, not only of tissue state.

Prostaglandins have a direct, drug-independent effect on the tendon itself. Separate from anything an NSAID does, PGE2 measurably decreases fibroblast proliferation and inhibits type I collagen synthesis in cultured human tendon tissue [15]. In a tendon producing excess PGE2 — whether from mechanical overload or reactive tenocyte activity — the tissue is being pushed toward matrix breakdown by its own signalling, independent of whatever medication is or isn't on board. This is part of why the reactive-to-degenerative continuum described above makes biological sense: sustained PGE2 exposure is catabolic to tendon in its own right, on top of anything a corticosteroid injection adds [11].

Swelling is functional, not incidental. The visible swelling of acute inflammation is prostaglandin- and histamine-driven vasodilation and increased capillary permeability, which lets plasma — fluid, clotting factors, antibodies, immune cells — move out of the vasculature and into the injured tissue [16]. That fluid is doing real work in the repair process; it is not simply an unwanted side-effect to be drained away, which is part of why "reduce the swelling" and "help the tissue heal" are not automatically the same clinical goal.

That same fluid shift also sensitises the nerve endings sitting in it. Prostaglandins don't only drive the vascular leak behind swelling — PGE2 acts directly on the peripheral nociceptor terminals within that tissue, binding EP1 and EP4 receptors and, through PKC- and PKA-dependent signalling, sensitising the TRPV1 channel so its activation threshold drops toward body temperature; the same EP1 pathway has also been linked to sensitisation of voltage-gated sodium channels on the nerve terminal [17]. The practical result is peripheral sensitisation and hyperalgesia: an ordinarily mild stimulus — light pressure, a normal joint angle, a warm shower — now generates real pain at the injury site, which becomes measurably more pain-generating than it was before injury, not just swollen. PGE2 does this alongside bradykinin, protons, ATP and cytokines released into the same expanding extracellular space — the so-called "inflammatory soup" — and the enlarged interstitial fluid volume itself helps deliver and concentrate that mixture directly against the nerve terminal [17]. This is, mechanistically, the most defensible part of what NSAIDs are actually doing: by cutting PGE2 synthesis, they reduce this peripheral sensitisation at its source — a real, well-characterised analgesic effect, distinct from (and not evidence for) any claim that they change the tissue's healing trajectory.

Endorphins provide genuine analgesia without touching inflammation at all. Beta-endorphin, released from the pituitary and at sites of tissue injury, binds the same mu-opioid receptors as morphine and produces real pain relief through a pathway entirely separate from the inflammatory cascade [18]. This is one reason graded exercise and movement-based rehabilitation reliably reduce pain even in tissue that is still actively remodelling: circulating beta-endorphin rises measurably with exercise and produces measurable, if short-lived, analgesia [18]. The clinical implication cuts both ways: pain relief from movement doesn't mean inflammation is gone, and, equally, pain relief from an NSAID or a corticosteroid injection doesn't mean the underlying process has actually resolved.

The clinical distinction that actually matters

The dividing line isn't "is this painful and swollen" — most musculoskeletal presentations are, at some point. The dividing line is:

• Contained, genuinely inflammatory pathology with low immediate mechanical demand (e.g., the acute stage of stiff shoulder) — a potent anti-inflammatory can be a reasonable short-term tool, ideally as a bridge into mobility work rather than a stand-alone treatment.

• Chronic, load-driven, degenerative tendon pathology (e.g., established LEP, and most other tendinopathies) — potent anti-inflammatories risk masking the very signal (pain) that should be governing a graded return to loading, while doing measurable harm to the tissue's capacity to remodel. Progressive loading, education, and addressing kinetic-chain contributors remain the interventions with durable outcomes.

For the early inflammatory phase specifically, the practical message for patients is a modest one: NSAIDs may take the edge off pain for a few days, but they are not doing meaningful biological work on the injury itself, and reflexively prescribing (or self-administering) them is not a substitute for an accurate diagnosis of what kind of tissue problem is actually present.

A second cost of chronic use: what NSAIDs do to the gut

There is a further reason not to reach for anti-inflammatories reflexively or for extended periods, and it has nothing to do with the injured tissue itself: the drug's effect on the gut.

NSAIDs are directly injurious to the gastrointestinal mucosa beyond the stomach. Capsule-endoscopy studies find visible small-bowel mucosal injury in roughly 70–80% of chronic NSAID users, frequently without overt symptoms [19]. This "NSAID enteropathy" involves prostaglandin depletion, increased intestinal permeability, bile-acid-mediated mucosal toxicity and mitochondrial injury to enterocytes [19,20]. Clinically, in a meaningful subset of long-term users this manifests as chronic occult GI blood loss and iron-deficiency anaemia, and, less commonly, protein-losing enteropathy with hypoalbuminaemia [21] — a genuine nutritional cost of sustained use, distinct from bulk macronutrient malabsorption but real nonetheless.

This damage is not purely local. Barrier breakdown allows bacterial translocation and lipopolysaccharide exposure, which activates toll-like receptor 4 and perpetuates mucosal inflammation, while chronic NSAID exposure itself shifts the composition of the gut microbiota — reducing protective, butyrate-producing taxa such as Faecalibacterium prausnitzii and Bifidobacterium species and favouring pro-inflammatory groups [19,20]. In short, a drug taken to dampen inflammation can, with sustained use, degrade the ecosystem that helps regulate mucosal inflammation and barrier integrity in the first place.

That matters for chronic pain specifically, because the link between gut microbiota health and chronic pain is more than a wellness-industry talking point. Across neuropathic, visceral and headache/migraine pain models, dysbiosis is consistently associated with heightened pain sensitivity through several converging mechanisms: reduced short-chain fatty acid (SCFA) production, which normally helps regulate microglial maturation and inflammatory tone; increased pro-inflammatory cytokine and pathogen-associated molecular pattern signalling that sensitises primary afferents and dorsal root ganglion neurons; and microglial activation in the spinal cord and brain that drives central sensitisation [22]. Restoring the microbiota — via probiotics, faecal transplantation or diet — has altered pain-related outcomes in several of these animal and early clinical models, though the human evidence base is still small and mixed [22]. There is also a plausible molecular link between the two threads: gut-derived SCFAs (notably butyrate) support hippocampal and central BDNF expression [23], and BDNF signalling is itself deeply involved in pain processing, generally facilitating central sensitisation at the spinal level [24]. A microbiota depleted by chronic NSAID exposure is therefore a biologically plausible, though not yet directly proven, contributor to impaired pain-modulation capacity — on top of the nutritional cost above.

Separately, and closer to everyday musculoskeletal practice, a recent Mendelian randomisation study used genetic instruments (which largely sidestep the confounding that limits ordinary observational microbiome research) to test for causal links between specific gut microbiota taxa and 26 musculoskeletal disorders. After correction for multiple testing, robust associations remained only for rheumatoid arthritis and shoulder bursitis, with rotator cuff syndrome reaching significance under the less conservative false-discovery-rate method alone; nominal associations with osteoporosis and calcific tendinitis of the shoulder did not survive correction [25]. This is a modest, hypothesis-generating result, not a clinical directive — but it is genuine genetic evidence that a gut–musculoskeletal axis exists, rather than an assumption.

Practical takeaway: none of this changes the core message above — the choice between NSAIDs, corticosteroids, or neither still has to be matched to the specific pathology and its time-frame. It adds a second, independent reason to avoid indiscriminate or prolonged anti-inflammatory use: beyond their limited or even counterproductive effect on the injured tissue itself, sustained use can erode gut barrier integrity and microbial diversity in ways that plausibly work against the body's own pain-regulation and repair systems.

References

1. Eming SA, Martin P, Tomic-Canic M. Wound repair and regeneration: mechanisms, signaling, and translation. Sci Transl Med. 2014;6(265):265sr6.

2. Zarghi A, Arfaei S. Selective COX-2 inhibitors: a review of their structure-activity relationships. Iran J Pharm Res. 2011;10(4):655-83.

3. Millar NL, Murrell GAC, McInnes IB. Inflammatory mechanisms in tendinopathy — towards translation. Nat Rev Rheumatol. 2017;13(2):110-22.

4. Ghosh N, Kolade OO, Shontz E, Rosenthal Y, Zuckerman JD, Bosco JA 3rd, Virk MS. Nonsteroidal anti-inflammatory drugs (NSAIDs) and their effect on musculoskeletal soft-tissue healing: a scoping review. JBJS Rev. 2019;7(12):e4.

5. Wang W, Shi M, Zhou C, Shi Z, Cai X, Lin T, Yan S. Effectiveness of corticosteroid injections in adhesive capsulitis of shoulder: a meta-analysis. Medicine (Baltimore). 2017;96(28):e7529.

6. Buchbinder R, Green S, Youd JM. Corticosteroid injections for shoulder pain. Cochrane Database Syst Rev. 2003;(1):CD004016.

7. Cook JL, Purdam CR. Is tendon pathology a continuum? A pathology model to explain the clinical presentation of load-induced tendinopathy. Br J Sports Med. 2009;43(6):409-16.

8. Bisset L, Beller E, Jull G, Brooks P, Darnell R, Vicenzino B. Mobilisation with movement and exercise, corticosteroid injection, or wait and see for tennis elbow: randomised trial. BMJ. 2006;333(7575):939-44.

9. Coombes BK, Bisset L, Brooks P, Khan A, Vicenzino B. Effect of corticosteroid injection, physiotherapy, or both on clinical outcomes in patients with unilateral lateral epicondylalgia: a randomized controlled trial. JAMA. 2013;309(5):461-9.

10. Coombes BK, Bisset L, Vicenzino B. Efficacy and safety of corticosteroid injections and other injections for management of tendinopathy: a systematic review of randomised controlled trials. Lancet. 2010;376(9754):1751-67.

11. Dean BJF, Lostis E, Oakley T, Rombach I, Morrey ME, Carr AJ. The risks and benefits of glucocorticoid treatment for tendinopathy: a systematic review of the effects of local glucocorticoid on tendon. Semin Arthritis Rheum. 2014;43(4):570-6.

12. Serhan CN. Pro-resolving lipid mediators are leads for resolution physiology. Nature. 2014;510(7500):92-101.

13. Cain DW, Cidlowski JA. Immune regulation by glucocorticoids. Nat Rev Immunol. 2017;17(4):233-47.

14. Buttgereit F, Smolen JS, Coogan AN, Cajochen C. Clocking in: chronobiology in rheumatoid arthritis. Nat Rev Rheumatol. 2015;11(6):349-56.

15. Cilli F, Khan M, Fu F, Wang JH-C. Prostaglandin E2 affects proliferation and collagen synthesis by human patellar tendon fibroblasts. Clin J Sport Med. 2004;14(4):232-6.

16. Ricciotti E, FitzGerald GA. Prostaglandins and inflammation. Arterioscler Thromb Vasc Biol. 2011;31(5):986-1000.

17. Moriyama T, Higashi T, Togashi K, Iida T, Segi E, Sugimoto Y, Tominaga T, Narumiya S, Tominaga M. Sensitization of TRPV1 by EP1 and IP reveals peripheral nociceptive mechanism of prostaglandins. Mol Pain. 2005;1:3.

18. Sprouse-Blum AS, Smith G, Sugai D, Parsa FD. Understanding endorphins and their importance in pain management. Hawaii Med J. 2010;69(3):70-1.

19. Wang X, Tang Q, Hou H, Zhang W, Li M, Chen D, Gu Y, Wang B, Hou J, Liu Y, Cao H. Gut microbiota in NSAID enteropathy: new insights from inside. Front Cell Infect Microbiol. 2021;11:679396.

20. Piccirelli S, Barberio B, Tettoni E, Treppiccione C, Pezzuto E, Tabbone E, Salvi D, Bertin L, Gerardi V, Cesaro P, Savarino EV, et al. Gut microbiota in NSAID enteropathy: current evidence and future perspectives for therapeutic strategies. Pharmaceuticals (Basel). 2026;19(7):1045.

21. Allison MC, Howatson AG, Torrance CJ, Lee FD, Russell RI. Gastrointestinal damage associated with the use of nonsteroidal antiinflammatory drugs. N Engl J Med. 1992;327(11):749-54.

22. Pak R, Cho M, Pride K, Abd-Elsayed A. The gut microbiota and chronic pain. Curr Pain Headache Rep. 2024;28(5):259-69.

23. Molska M, Mruczyk K, Cisek-Woźniak A, Prokopowicz W, Szydełko P, Jakuszewska Z, Marzec K, Trocholepsza M. The influence of intestinal microbiota on BDNF levels. Nutrients. 2024;16(17):2891.

24. Mazzitelli M, Kiritoshi T, Presto P, Hurtado Z, Antenucci N, Ji G, Neugebauer V. BDNF signaling and pain modulation. Cells. 2025;14(7):476.

25. Wang Y, Sun Y, Liao H. Causal associations between 26 musculoskeletal disorders and gut microbiota: a Mendelian randomization analysis with Bayesian validation. World J Microbiol Biotechnol. 2025;41(4):106.