In complementary medicine, we have many natural therapies that have existed for hundreds, even thousands of years: acupuncture, herbal medicine, homeopathy, spinal manipulation. Since "hard" evidence for or against these treatments is often scarce (not least because the research funds are not available), promoters are often keen to point out that complementary therapies are first natural and therefore harmless, and second, they have stood the test of time. Therefore their effectiveness and safety can be taken for granted.

This sounds plausible but it can also be misleading. Recently, the World Health Organisation warned of the hazards of alternative medicine. It reported that in 2002, almost 10,000 adverse reactions were reported in China - more than twice the number recorded during the previous decade.

How is this possible? To start with, the claim that all complementary treatments are natural is quite simply untrue. What, for instance, is natural about sticking needles into people's skin as in acupuncture, or manipulating their spinal column as in chiropractic? But other complementary therapies are indeed natural. In herbal medicine, for example, we use the products of nature as medicine. But does this mean that all herbal medicines are automatically safe?

Obviously not. Many potent poisons, after all, are plant-based. Think of hemlock. The correct conclusion therefore must be that some, but not all, complementary therapies are natural and that some, but not all, natural treatments are safe. In some cases, natural can mean dangerous.

And what about the "test of time"? Does the fact that a treatment has been around for centuries provide us with any guarantee of its value? Let's consider another example from history. Bloodletting was used for centuries, not just in England but in most other cultures. It was employed as a "cure-all", and if the patient died during the course of it, the physicians of the time concluded they had not drained enough blood. Today we understand more about the functions of blood, and know that bloodletting must have hastened the death of thousands, if not millions. In this particular case, the test of time has proved to be a poor indicator of efficacy.

And what about safety? Are all treatments with an ancient history devoid of risk? Instinctively, many of us think of time-tested herbal treatments as safe. If not, would we not have noticed ages ago that people dropped like flies when taking them? Perhaps, but the trouble is that things are often much more complicated. A herbal potion could cause less tangible harm which only emerges years later. We know, for example, that some herbal treatments have a potential for damaging our genes; many plant extracts have been shown to produce DNA damage in test-tube tests. Such effects do not necessarily occur in our body, but we cannot exclude the possibility. Gene damage would become noticeable only in generations to come. And who would then think of a herbal cause?

But there are more obvious reasons why a therapy that has stood the "test of time" could still produce harm. For example, a herbal extract for sale in UK health food shops might be manufactured differently from that used by our ancestors. They may have used it as a tea, while today's commercial product could be a hi-tech pill.

This seemingly small difference could mean that certain toxic substances of the plant enter the commercial product but not the traditional drink. Some people believe that this is what happened with kava, the herbal anxiolytic recently banned in the UK. The traditional drink was probably safe but the commercial extracts might have caused liver damage. And there is dosage: typically, a tea is low-dose while an extract could contain dangerously high concentrations of active ingredients.

Traditionally, a herbal remedy was taken on its own. It was certainly not combined with the drugs we use today. Therefore, herb-drug interactions are an important consideration. The lesson we should learn from all this is one of caution and common sense. Natural is not necessarily safe and a long history of use is a far cry from providing guarantees. There are many people out there who, for one reason or another, will try to convince you of the opposite. My advice is not to believe them.

· Edzard Ernst is professor of complementary medicine at the Peninsula medical school at the Universities of Exeter and Plymouth.