In Brief
| Drug | NNT (95% CI) | What matters |
|---|---|---|
| Etoricoxib 120 mg | 1.8 (1.7–2.0) | Best figure; in the EU that dose is for gout only |
| Diclofenac potassium 50 mg | 2.1 (1.9–2.5) | The sodium salt gives 6.6 in the same setting |
| Metamizole 500 mg | 2.4 (1.8–3.1) | Thinnest evidence base; agranulocytosis |
| Naproxen 500/550 mg | 2.7 (2.3–3.2) | Lasts longer than the rest |
| Ketoprofen 50 mg | 2.9 (2.4–3.7) | The average misleads: 1.8 and 4.2 |
A painkiller ranking looks convincing: five names, five numbers, a clear order. Such a list is comfortable to live with — it answers the question «which is stronger».
The trouble is that the source of those numbers does not answer that question.
What Was Actually Measured
All five figures come from Cochrane single-dose reviews, and all describe the same thing:
- pain that is acute and postoperative — most often after tooth extraction or an orthopaedic procedure;
- a single dose, not a course;
- an observation window of four to six hours;
- the outcome being the proportion of patients whose pain fell by at least half compared with placebo;
- and almost everywhere that proportion was computed from the area under the relief curve rather than measured by asking.
Hence the first limitation, and it is a hard one. None of this transfers to headache, period pain or back pain. Not because the drugs fail there, but because these reviews hold no such data. The list describes one specific situation: severe pain right after surgery, in the first hours.
Why This Is Not a Ranking
All five reviews were gathered into one overview: 39 reviews, some 460 studies, roughly 50,000 participants. And it contains a sentence that devalues any ranking built on these figures:
«No statistical comparison was undertaken.»
The reason is simple. Each drug was compared with placebo in its own trial, with its own patients, its own operation and its own measurement. You can place the results side by side; you cannot subtract one from another.
You can check this yourself by looking at the confidence intervals. If they overlap, the difference is not established.
| Pair | Intervals | Difference |
|---|---|---|
| Etoricoxib and naproxen | 1.7–2.0 and 2.3–3.2 | established |
| Etoricoxib and ketoprofen | 1.7–2.0 and 2.4–3.7 | established |
| Etoricoxib and diclofenac | 1.7–2.0 and 1.9–2.5 | overlapping |
| Diclofenac and metamizole | 1.9–2.5 and 1.8–3.1 | overlapping |
| Metamizole and naproxen | 1.8–3.1 and 2.3–3.2 | overlapping |
| Naproxen and ketoprofen | 2.3–3.2 and 2.4–3.7 | overlapping |
In total: of the ten possible pairs, the difference is established in exactly two. Etoricoxib separates from the bottom of the list — that is a real result. Everything else is an order that looks like knowledge and is not.
Metamizole shows it most clearly: its interval of 1.8–3.1 spans both first place and last. On these data it could be better than diclofenac or worse than ketoprofen.
Where the Average Misleads
For two of the five, the headline figure describes arithmetic rather than a situation.
Ketoprofen. Its 2.9 is the average of two entirely different results: 1.8 in dental pain and 4.2 in other surgery. The difference is statistically significant at P below 0.00006, and the review authors call the gap unusual. No patient lands on 2.9: they land in one subgroup or the other. More in the ketoprofen guide.
Diclofenac. Here it is not the type of pain but the salt. The potassium salt gives 2.1, the sodium salt in the same setting 6.6, and the review authors write plainly that the sodium salt «has limited efficacy and should probably not be used in acute pain». The box says one word, «diclofenac», and the result differs threefold. The full account is in the guide to the two salts.
What the Table Does Not Measure at All
NNT is computed over six hours after a single dose. Everything that happens later falls outside it by construction.
So duration of action is invisible in the list. Naproxen's half-life is markedly longer than the others': it is taken twice a day rather than three or four times. For pain lasting days that changes more than a decimal place of NNT — yet naproxen still looks mid-table. That is covered separately.
The table misses something else too: availability. Etoricoxib, with the best figure, is prescription-only in all three countries, and in the EU its 120 mg dose — the very one that produced the 1.8 — is authorised only for an acute gout attack. Metamizole is not sold in France at all. Oral ketoprofen is prescription-only in Germany, while in Italy 68 of 105 packs are sold freely.
Safety — Common to All Five
Never take two NSAIDs at once. This is the cardinal rule and the commonest mistake. The analgesia does not add up — the receptors are already occupied — while the risk to stomach, kidneys and blood pressure adds up perfectly well.
- Stomach. The risk of erosions and bleeding belongs to the whole class; COX-2 selective agents such as etoricoxib are better tolerated but not safe.
- Kidneys. All NSAIDs reduce renal blood flow. Caution with dehydration, chronic kidney disease and in older age.
- Blood pressure. The class can raise it and blunt antihypertensive drugs.
- Heart. The risk is not uniform within the class: diclofenac carries more of it than some others.
- Duration. Over-the-counter packs are sized for days, not weeks. If pain outlasts them, that is a reason to see a doctor rather than to buy another box.
Metamizole adds a risk the others do not share — agranulocytosis. The signs that mean stopping at once, and what the regulator changed in December 2024, are covered in its guide.
How to Choose When the Table Has No Answer
| Situation | What is reasonable |
|---|---|
| Dental pain | Ketoprofen — in that subgroup its figure is at the leader's level |
| Severe short-lived pain after surgery | Etoricoxib, if a doctor has prescribed it |
| Pain that will last several days | Naproxen — a duration the others lack |
| Rapid relief needed in acute pain | Watch the salt: diclofenac potassium, not sodium |
| Another NSAID already started | Do not add a second under any circumstances |
And one rule worth ten tables: a drug that has already worked for you and was well tolerated usually beats one whose review figure is a tenth of a point prettier.
The Bottom Line
The five are a good orientation and a poor ranking. What genuinely follows from them is that etoricoxib separates from the bottom of the list in acute postoperative pain, that with diclofenac the salt decides, and that with ketoprofen the type of pain does. What does not follow is who is «strongest»: that comparison was never made.
And the most important limitation deserves repeating at the end, because it is the one most often lost: all these figures were measured in acute pain after surgery, after a single dose, over six hours. Your headache took no part in these reviews.
Detailed accounts of each drug: etoricoxib, diclofenac, metamizole, naproxen, ketoprofen.
Where to get them
All five are in the catalogue. No price is shown on the cards: prescription medicines are ordered through the chat, and the administrator confirms availability and cost.
