Is Amalgam Removal Safe? What the Evidence Actually Says

Do mercury fillings actually make you sick, or is that a myth?

Two randomised controlled trials were built to answer exactly this, and neither found detectable harm. Casa Pia randomised 507 children to amalgam or composite over seven years, finding no significant difference in memory, attention, visuomotor function or nerve conduction velocity at any point (P values from .29 to .91) ^[1]. The New England Children's Amalgam Trial followed 534 children for five years: the difference in five-year full-scale IQ change was 1.0 point (95% CI -0.6 to 2.5), not significant ^[2]. A Casa Pia kidney analysis found no effect on urinary albumin or microalbuminuria ^[3].

In adults, a cohort of 20,000 New Zealand Defence Force personnel found no association with chronic fatigue syndrome or kidney disease, and a multiple sclerosis hazard ratio of 1.24 (95% CI 0.99 to 1.53) the authors called limited evidence ^[6]. A 2021 systematic review found no difference in systemic effects between amalgam and resin-based materials ^[5].

The exposure is real. Urinary mercury in the amalgam children peaked near 3.2 µg/L in year two, up from 1.5 at baseline, returning to baseline by year seven ^[9]. A randomised adult trial put the daily absorbed dose at up to 3 micrograms for an average number of fillings and 7.4 for a heavy load, against a World Health Organization tolerable figure of 30, so even a heavy load sits at roughly a quarter of it ^[10].

How does an amalgam filling actually stay in the tooth?

It does not stick. It is wedged in, and that single fact explains most of what follows.

Amalgam has no chemical bond to tooth. It is held by shape alone: the cavity is cut with walls that lean inwards, sometimes with small grooves, so the set metal locks in and cannot lift out. Sound tooth is removed to create that lock, over and above whatever the decay destroyed.

The consequence is measurable. Where the same premolars were tested sound, then prepared, then restored, cutting the cavity dropped cusp stiffness to 0.58 of the intact tooth. Filling it with amalgam returned 2.6 per cent of what the preparation had taken away. Bonded composite in the same teeth returned 77.8 per cent ^[14]. A second laboratory put unbonded amalgam's contribution at 5 per cent, against 39 to 61 once the amalgam was bonded in ^[15]. In root-treated molars loaded to failure, intact teeth withstood 2,485 newtons, prepared and unrestored teeth 534, and amalgam-restored teeth 1,706 ^[16].

Read that carefully, because it is easy to overstate in the other direction. Amalgam is not attacking the tooth. It is simply not helping it. The filling sits in the cavity as a passive block and the walls flex around it much as they would if it were empty.

A large old filling marks a tooth that has already lost a great deal, and the material in the middle was never designed to give any of it back.

The weakness comes from the missing tooth, not from the metal.

Unbonded amalgam^[14]2.6%

Bonded amalgam^[14]62.5%

Bonded composite^[14]77.8%

Percentage of the stiffness lost to cavity preparation that each restoration gave back. Laboratory measurement on extracted premolars, each tooth serving as its own control.

What happens to mercury levels when an old amalgam is drilled out?

They rise sharply and briefly, then settle below where they started. When all amalgams were removed in one session from 12 adults with a mean of 18 amalgam surfaces, plasma mercury rose within 48 hours, peaking on average 32% above pre-removal levels ^[11].

Technique changes that number. Among 28 patients treated with water coolant and high-volume evacuation, 18 had rubber dam isolation and 10 did not; only the group without a dam showed significant rises in plasma (p = 0.012) and urine (p = 0.037) mercury ^[13]. Isolation is not a wellness gesture. It is the measurable difference between a controlled procedure and an uncontrolled one, and the protocol sits on the holistic and biomimetic dentistry page.

It is worth being exact about why this argues against removal as detoxification, because the one-year figures on their own could be read the other way. Clearing every filling does lower the long-run number. The difficulty is that the number was never in a range that was doing anything, at roughly 3 micrograms a day for an average load and 7.4 for a heavy one against a tolerable figure of 30 ^[10]. So a healthy mouth accepts a real spike, and loses sound tooth at every site, to reduce a measurement that was not causing harm. Where a filling has actually failed, that same isolation makes the replacement a controlled procedure, and exposure stops being the question that decides anything.

Plasma peak above pre-removal level, first 48 hours^[11]32%

Plasma below pre-removal level one year later^[13]52%

Urine below pre-removal level one year later^[13]76%

A brief spike, then a long decline. Rubber dam isolation targets the spike.

Should I get my old silver fillings taken out if they aren't causing problems?

No, not for the material's sake. But "not causing problems" is the part worth checking rather than assuming, because the failure that costs a tooth is usually the one with no symptoms. A restoration that is intact, sealed at its margins, symptom-free and clean on a radiograph is doing its job, and that last item is the one you cannot check for yourself. The 2021 systematic review put it plainly: old amalgams should be replaced with adhesive restorations when clinically necessary, not because of concerns about the material ^[5].

The reason is biological, not ideological. No filling comes out without a rim of sound tooth coming with it, which is the argument for preserving tooth structure.

Two numbers sit behind that caution. When 544 fillings were removed and the cavity underneath inspected, 22.6 per cent had no decay at all ^[23]: roughly one replacement in four opened a tooth that did not need opening. And across 1,337 decisions to replace an existing restoration, 70 per cent ended with more surfaces of the tooth restored than before ^[22]. Replacement is rarely like for like. It is usually a step up.

Will I feel better or healthier after having my amalgam fillings removed?

If you are otherwise well, there is no good evidence that you will. A randomised trial assigning 82 patients to removal, to removal plus a detoxification programme, or to no removal found inorganic mercury falling to 27% of pre-removal levels within 60 days ^[10]. In a separate group followed a year after full removal, plasma mercury sat 52% and urinary mercury 76% below pre-removal levels ^[13].

Those numbers are not in dispute. What has not been shown is that the levels they replaced were causing symptoms ^{[5] [6]}. A falling biomarker is a measurement, not a recovery. True allergy to mercury is a recognised reason to change material, but it is uncommon, and a diagnosis rather than an assumption.

If amalgam is being phased out worldwide, doesn't that mean it is unsafe?

No. This is the most misread fact in the subject. The Minamata Convention on Mercury is an environmental treaty aimed at protecting health and the environment from man-made mercury emissions. It entered into force in August 2017, was ratified by 127 governments as of February 2021, and commits each to at least two of nine measures to phase down, not ban, amalgam ^[4]. World Dental Federation policy runs on the same two tracks: safe use for the patient, phase-down for the environment ^[12]. The regulated problem is mercury reaching waste water and disposal streams. That is a reason to stop placing new amalgam, not a verdict on the filling in your tooth.

What does a failed amalgam look like, and why is it usually found late?

If the filling was never holding the tooth together, then what fails is usually the tooth. Patients are told a filling has failed as though the word explains itself, and the reason it matters is timing: the failure that ends a tooth is rarely the one you notice.

Amalgam is durable. Median survival is around twelve to thirteen years and many last far longer ^[17]. When these restorations do end, three things account for most of it: decay restarting at the edge, fracture of the filling, and fracture of the tooth around it. In a seventeen-year practice study, cusp fracture alone accounted for five per cent of every restoration placed, and the highest extraction rate of all was in premolars carrying complete amalgams ^[17].

That third category is the silent one.

In the largest study of cracked teeth conducted, 2,858 teeth across 209 practices, fifty-five per cent of teeth with a visible crack had no symptoms at all ^[18]. Among those that did hurt, the commonest complaint was not what the textbooks predict: pain to cold in 37 per cent, pain on biting in only 16 ^[18]. And what is visible understates what is there. When cracked teeth were opened, 89 per cent had at least one internal crack as well ^[19].

This is why the timing goes wrong. A tooth can be cracking for years while feeling normal, and the first unmistakable signal is often a piece of cusp breaking away. By then the choice is no longer between a small repair and a larger one.

Now the counterweight, because alarm is not accuracy. In that same cohort followed three years, only three per cent of cracked teeth fractured and twelve per cent showed any progression ^[20], survival exceeded 98 per cent, and around eighty per cent of teeth placed under observation were still simply being observed ^[30]. A crack is not a countdown. Most are stable.

What genuinely changes the outlook is depth. Where a crack sat beside a gum pocket deeper than six millimetres, two-year survival after root canal treatment fell to 74 per cent against 97 where the pocket was shallower ^[21].

A crack that has travelled below the gum is the one that ends teeth.

What you can see, and what only an examination can settle

Some of this is genuinely visible to you. How many walls are left, since a filling bounded by four standing walls is a different proposition from a rim of tooth around a slab of metal. How thick those walls are, because a wall thinned to a shell no longer behaves like one. Crack lines, often clearest when the tooth is dry. And a change in shape, since a cusp that has flattened or chipped has been saying something for a while.

What none of that settles is what to do. Whether a crack is confined to enamel or has reached dentine, whether a solid-looking wall is undermined from beneath, whether a pocket has formed beside a crack, whether decay is tracking under the margin: none can be judged from outside. They need magnification, a radiograph and probing.

Here the conservative answer refuses to be simple. Sometimes the least destructive thing is to leave a twenty-year-old amalgam exactly where it is. Sometimes it is to replace it now, because waiting until the cusp breaks turns a moderate repair into a major one. Both are conservative positions, and only an examination tells you which applies.

What happens when it is replaced, and why the first cut decides the rest

This is the part almost nobody explains, and it is where a tooth is either preserved or quietly spent.

When the amalgam comes out, the cavity does not come with it. What remains is the shape cut to hold the filling: walls leaning inwards, undercuts, grooves. A ceramic inlay or onlay is a solid object that seats from above, so it cannot drop into a space narrower at the top than the bottom. Those undercuts must be dealt with, and there are two ways.

Cut them away. Widen the walls until they diverge and the ceramic will seat. Quick and certain. It also removes sound tooth from a tooth that has already lost a great deal, and removes it for the convenience of the restoration rather than because disease was there.

Or fill them in. Bond composite into the undercuts, rebuild the internal shape to something a ceramic can seat onto, and leave the walls standing. Nothing sound is removed. The composite becomes a substructure and the ceramic sits on it.

The evidence for the second is better than most assume. Sixty-four compromised molars were restored with lithium disilicate in four preparation designs and fatigued in a chewing simulator ^[24]. The design leaving the weakened wall standing produced a significantly less destructive failure pattern than either cusp-coverage design. The design where the wall was cut back to make a wider inlay was the one with significantly lower fracture strength ^[24].

In that study, cutting the wall back gave the worst of both outcomes: less tooth, and a weaker result.

Cutting the wall away does not buy strength.

A gap I have to be straight about. No published study has directly compared blocking out undercuts against cutting the walls away. The step is standard practice taught everywhere, resting on the design comparison above, on adhesive principle, and on the fatigue work behind bonded substructures. It does not rest on a head-to-head trial, because none exists.

And the principle has a limit. A wall can be too thin to keep. Where roughly two millimetres remained, restoring it as an inlay gave 380 newtons against 470 for a partial onlay covering the cusp instead ^[29]. Thin walls are better covered than preserved. So the rule is not "never cut." It is narrower: do not remove sound wall merely to give the restoration a path of insertion, and decide cusp coverage on the wall's real thickness rather than on what makes the ceramic easier to fit.

Sealing the dentine at the moment it is cut

One more step, and it belongs in the same visit rather than at the fitting.

The moment an old amalgam is removed, dentine is exposed for the first time in perhaps twenty years, and it is at its most bondable right then. Left bare while an impression is taken, a temporary worn and a laboratory works, it spends weeks collecting temporary cement and bacteria, and the bond made later is made to a compromised surface.

Immediate dentine sealing is the decision to bond it now instead. In extracted molars, dentine sealed at preparation reached 58 megapascals, indistinguishable from a fresh bond, while the same teeth sealed only at cementation reached 12 ^[25].

In this situation it does a second job. The composite blocking out the undercuts and the sealed dentine beneath are the same piece of work, done while the surface is fresh. When the ceramic is finally fitted it bonds to composite and enamel, not to dentine that has spent a fortnight under a temporary.

The honest limits. That is laboratory measurement. In patients, a meta-analysis of the clinical trials found no difference in post-operative sensitivity between sealing early and late ^[26]. What can be said is narrower: the technique reliably improves the bond on the bench, and the series reporting the best long-term results use it. Across 1,146 partial lithium disilicate restorations followed an average of seven and a half years, all placed with immediate dentine sealing, survival was 97.3 per cent ^[27].

There is also direct reassurance about the surface itself. Dentine stained by decades of contact with amalgam bonds perfectly well: in 118 large restorations placed specifically to replace amalgams, survival was 96.6 per cent at just over three years, and the failures were related neither to inadequate adhesion nor to decay ^[28].

1. 1

   Assess before anything is touched

   Walls, wall thickness, crack lines, pocket depth beside any crack, and radiographs. This decides whether the filling should come out at all.

2. 2

   Remove under isolation

   Rubber dam and high-volume suction, the step measured to control mercury exposure during removal.

3. 3

   See the real cavity

   Only once the metal and any decay are gone is the true amount of remaining tooth visible. Any definite plan made before this is provisional.

4. 4

   Seal the dentine immediately

   Bond the freshly exposed dentine in the same visit, while it is at its most bondable.

5. 5

   Block out, do not cut out

   Fill the undercuts left by the old preparation rather than widening the walls to remove them.

6. 6

   Decide coverage on thickness

   Cover a cusp that is genuinely thin. Keep a wall that is genuinely sound. Thickness decides, not convenience.

If amalgam is safe, why place bonded materials instead?

Because the preparation can stop where the disease stops. A bonded restoration is held by adhesion rather than by shape, so no sound tooth has to be sacrificed to lock it in. That is what adhesive dentistry changed about restorative practice.

It is worth saying plainly that this is not a claim of superiority in every respect.

In the trials Cochrane pooled, composite restorations had almost double the failure rate of amalgam (RR 1.89, 95% CI 1.52 to 2.35) and a higher risk of secondary caries (RR 2.14), though no greater fracture risk ^[4]; in Casa Pia the need for further restorative treatment from year five ran about 50% higher in the composite group ^[1]. Those composites were placed in the 1990s and materials have improved since, but technique still decides the result, which is why composite work is placed under isolation and in increments.