Ginger extract

Ginger has a stronger human evidence base than almost anything else in this section, and the entry has to hold two things at once: the trials are better than most in this section, and the field cannot agree on what they show.

The trial

Two hundred and sixty-one people with knee osteoarthritis and moderate-to-severe pain were randomized, after a washout, to a standardized concentrated extract of two ginger species — Zingiber officinale and Alpinia galanga, sold as EV.EXT 77 — or placebo, twice daily for six weeks, with acetaminophen allowed as rescue.

On the primary intention-to-treat endpoint, the proportion of people whose knee pain on standing fell by at least 15 mm, 63% responded on ginger against 50% on placebo (p = 0.048). The secondary measures agreed: pain on standing fell 24.5 mm versus 16.4 mm (p = 0.005), pain after walking fifty feet 15.1 versus 8.7 mm (p = 0.016). The WOMAC composite moved in the same direction and missed significance (12.9 versus 9.0 mm, p = 0.087). Quality of life was identical between arms.

That is a properly conducted multicentre trial with a positive result, which is more than most of this section can say. It is also a thirteen-percentage-point difference in responder rate at p = 0.048, over six weeks, on a symptom.

The two meta-analyses

In 2015, five randomized placebo-controlled trials covering 593 patients were pooled. Pain favoured ginger by a standardized mean difference of −0.30 (95% CI −0.50 to −0.09, p = 0.005), with low inconsistency between trials (I² = 27%), and disability by −0.22 (−0.39 to −0.04). The authors called ginger modestly efficacious and reasonably safe, rated the evidence moderate quality, and noted that most of the included trials had inadequate intention-to-treat analyses.

In 2020, a second systematic review pooled four trials from an overlapping set and reached the opposite conclusion. What it did not reach was the opposite estimate. Capsules against placebo gave a pain difference of −7.88 mm at one month, p = 0.0001, with the four trials showing no heterogeneity at all (I² = 0%) — the cleanest pooling anyone has managed here. Function did not separate (p = 0.24), and those same four trials disagreed with each other almost completely on it (I² = 99%). Topical ginger produced nothing. The authors called the pain effect minimal, confined to that one domain of the WOMAC, described the evidence base as heterogeneous and of poor methodological quality, and concluded there was insufficient evidence to support oral ginger over placebo.

So the two reviews agree on the number and disagree on what to do with it. That is not primarily a disagreement between review teams either. It is a statement about how few trials there are, and how much weight four of them can carry.

The crossover trial against ibuprofen

Before Altman, a three-period double-dummy crossover compared ginger extract, ibuprofen and placebo in the same patients, three weeks each. The ranking came out ibuprofen > ginger > placebo on both the pain visual analogue scale (p < 0.00001) and the Lequesne index (p < 0.00005), which is the sentence usually quoted.

The next sentence is the one that matters: in the crossover analysis proper, ginger did not differ significantly from placebo. The advantage appeared only in exploratory tests confined to the first treatment period, before anyone had crossed over — which is where a crossover trial is least protected against the things it was designed to control.

What it costs to take

Tolerability is the consistent finding across the whole literature. In the pooled analysis, people taking ginger were more than twice as likely to withdraw because of adverse events (RR 2.33, 95% CI 1.04–5.22). In Altman's trial, gastrointestinal adverse events occurred in 59 patients on ginger against 21 on placebo — mostly mild, and nearly three times as many.

Ginger also has antiplatelet activity in laboratory work, which is the standard basis for caution alongside anticoagulants. No study quantifying a clinical bleeding risk turned up in the search behind this entry, so that is a mechanism rather than a number.

What it might be doing

6-gingerol, the principal pungent constituent, reduces catabolic and oxidative markers in osteoarthritic chondrocytes and is protective in an animal model of osteoarthritis. Ginger's constituents inhibit COX and lipoxygenase pathways, which is the pharmacological reason its trials compare it with ibuprofen rather than with a placebo food. And in the one trial to measure it over a useful period, three months of 500 mg ginger powder daily lowered TNF-α and IL-1β against a starch placebo.

Set alongside the other supplements in a 2022 pooled analysis, ginger improved pain and no other clinical measure — and that pooled pain effect is a single trial holding it up. Its standardized mean difference of −3.76 becomes −0.20, with a confidence interval crossing zero, the moment that one study is dropped. In the same analysis curcumin improved pain and function against placebo but not against active comparators, vitamin D improved both, vitamin E did nothing, and omega-3 improved function but not pain.

That analysis also cut its results by trial quality, and the two directions are worth putting side by side. Curcumin's benefit is present in the trials scoring well on the Jadad scale and absent in the weaker ones. The herbal formulations do the opposite: a large function benefit in the trials scoring under three, nothing at all in those scoring four or more.

Practical notes

The positive symptom result belongs to a specific standardized extract of two species, not to ginger generally, and not to the root you cook with — no trial of dietary ginger against a joint endpoint exists in any species. Powdered rhizome at 500 mg a day, which is what the cytokine trial used, is a different exposure again.

What would change this entry

An adequately powered trial with a proper intention-to-treat analysis — the exact weakness the 2015 review names in most of the studies it pooled. Twenty-five years after the Altman trial, nobody has run it, and until somebody does the two meta-analyses will keep disagreeing about the same five papers.