Restorations on a radiograph

A restoration on a radiograph is any filling, inlay, crown or bridge unit that appears inside the tooth outline as material the beam passes through differently from tooth. How brightly it reads depends on the material and on how thick it is. Reading one means reading its margin, because the margin is where caries under a restoration would show.

Synonyms: fillings on a radiograph; amalgam on a radiograph; crown on a radiograph

What is a restoration on a radiograph?

It is the first thing on most bitewings and the thing that sets the terms for everything else on the image. A restoration occupies space that used to be tooth, so it removes a surface you would otherwise read directly and replaces it with a border you have to read instead.

That border is the whole job. Everything below is about how much the border can tell you, and where it stops telling you anything.

How radiopaque is a restoration, and when does it stop standing out?

Not as reliably as the phrase "bright filling" suggests.

Laboratory measurements on tooth-coloured CAD/CAM materials found that at 1 mm thickness, every CAD/CAM specimen tested matched or fell slightly below enamel, and that radiopacity varied with the thickness of the specimen [1]. In the same study, polymer-infiltrated ceramic was fairly radiolucent despite containing radiopacifying elements and a high filler content [1]. A separate comparison of 13 restorative CAD/CAM blocks found one lithium disilicate block whose radiopacity did not differ significantly from dentin; the evaluated materials had significantly different radiopacity values from one another [2].

So a thin ceramic restoration can read at roughly the density of the tooth it sits in. If yours doesn't stand out, the material may be the reason [1]. The radiolucent restorative materials entry covers the set that goes the other way and reads darker than tooth.

Material under the restoration behaves differently again. Every pulp-capping material in one in-vitro comparison was more radiopaque than dentin, and only some of them were more radiopaque than enamel [3]. The authors of that study note that where enamel is superimposed, a material that is not more radiopaque than enamel can complicate radiographic assessment [3].

All of the above are disk specimens measured against an aluminium step wedge. They tell you what the materials do to a beam. They are not measurements of restorations in a mouth.

How do you tell a restoration margin from caries at that margin?

Partly by where you are looking, and the effect is large.

In an in-vitro study of 82 posterior teeth with prepared class II amalgam boxes and simulated lesions, direct digital radiography reached a sensitivity of 73% and specificity of 95% at the buccal and lingual line angles, and a sensitivity of 29% and specificity of 90% at the mid-gingival floor. For conventional radiography the figures were 63% and 93% at the buccal line angle, 61% and 93% at the lingual line angle, and 44% and 95% at the mid-gingival floor. Lesions at the buccal or lingual line angle were more easily detected than those in the mid-gingival region [4].

Read that as a map of where the radiograph is worth trusting. A shadow at the line angle is a finding. A clean-looking mid-gingival floor is much weaker evidence that nothing is there.

Overall performance is less settled than any single figure suggests, and the studies do not agree:

• In vitro, on 47 molars and 49 premolars with class II amalgams, bitewing sensitivity was 73% for molars and 80% for premolars, with specificity 90% and 95%. Negative predictive value was 58% for molars and 86% for premolars [5].
• In the 82-tooth study above, total sensitivity was 58% and specificity 93% for digital radiography, and 56% and 93% for conventional; overall accuracy was 76% and 75% respectively [4].
• In a third in-vitro comparison, on 66 extracted teeth of which 48 carried amalgam restorations, conventional bitewing radiography detected secondary caries with a sensitivity of 0.65 and specificity of 0.81 [6].

Those are three laboratory estimates of the same quantity, made on different teeth and verified in different ways, and they do not converge on one sensitivity. Specificity was 90% and 95% in the first, 93% in the second and 0.81 in the third [4][5][6]. In the first of those studies, positive predictive value was 95% for molars and 92% for premolars [5] — so a shadow the radiograph does call at the margin was usually a real lesion, while a negative predictive value of 58% for molars [5] shows the reverse did not hold nearly as well.

The finding itself has its own entry — secondary caries — and so does the other thing that changes a margin's shape, a restoration overhang. At the cervical margin, cervical burnout is the confusion that arrives before either of them.

What does a crown hide?

Enough that the study designed to test it went looking for a better modality.

In an in-vitro study of 120 teeth — 30 with class V composite, 30 with class V amalgam, 30 with thermoplastic polymer full crowns and 30 with nickel-chromium metal crowns — cone-beam CT outperformed intraoral radiography for detecting artificial buccal recurrent caries under the restorations [7]. Both intraoral film and intraoral digital images were included, and both were beaten.

That is one study on extracted teeth with artificially created lesions, so take the direction rather than the size of it: under full coverage, the intraoral radiograph is working with less than it is used to.

There is a related finding worth keeping straight, because it is easy to over-read. On CBCT, having a restoration in the neighbouring tooth lowered accuracy for proximal caries in the tooth being read: 0.826 for artifact-free scans against 0.526 next to an amalgam, 0.613 next to a porcelain crown and 0.601 next to a metal-ceramic restoration; intraoral radiographs of the teeth alone scored 0.657 [8]. That adjacency effect was measured on CBCT only. It says nothing about what a neighbouring restoration does to a bitewing.

What it's confused with

Which appearances need a closer look?

This is where the radiograph hands the decision back to the examination, and the guidance is explicit about the order.

ADA/FDA patient-selection guidance states that radiographic screening for the purpose of detecting disease before clinical examination should not be performed, and that clinical examination, patient history, review of any prior radiographs and caries risk assessment should precede radiographic examination [9]. It also directs that care be taken to examine all radiographs for any evidence of caries, bone loss from periodontal disease, developmental anomalies and occult disease [9].

Two points in that guidance bear directly on restorations. Large or deep restorations are named among the clinical situations for which radiographs may be indicated [9]. And for the adult patient, the guidance states it is important to examine patients for recurrent dental caries [9].

On interval: for a recall patient with clinical caries or an increased risk of caries, the guidance recommends a posterior bitewing examination at 6 to 12 month intervals if proximal surfaces cannot be examined visually or with a probe [9].

Common errors

• Treating a low-contrast restoration as an imaging problem. A thin ceramic can sit at roughly the density of enamel or dentin by material property [1][2].
• Reading a clean gingival floor as a clear gingival floor. Sensitivity at the mid-gingival floor was 29% on direct digital radiography in vitro, against 73% at the line angles [4]. The projection is the limit, not the tooth.
• Quoting a laboratory sensitivity as chairside performance. Every figure on this page was measured in vitro, on extracted teeth [4][5][6].
• Carrying a CBCT finding across to a bitewing. The restoration-adjacency effect on accuracy was measured on CBCT [8], and does not transfer.