How accurately does a tube shift localise an impacted canine?

When a maxillary canine hasn't erupted, the usual way to find out whether it lies buccal or palatal is two radiographs read by the principle of parallax [1]. Four studies, three in patients and one on a laboratory typodont, have scored that read against a known position. Vertical parallax placed the canine correctly in 65% to 97.7% of readings, and the studies split on whether the horizontal or the vertical tube shift does better [2, 3, 4, 5]. Cone-beam CT (CBCT) can localise the canine, typically at a higher effective dose than conventional radiography [1].

What the question actually is

A tube shift, also called the parallax technique, compares two radiographs taken with the X-ray tube in two positions. For a canine, horizontal parallax pairs an upper standard occlusal with a periapical radiograph centred on the canine region, or uses two periapicals, one centred on the upper central incisor and one on the canine region. Vertical parallax pairs an occlusal or periapical taken with the beam aimed downwards with a panoramic radiograph [1].

The question here is narrower than how the rule works. It asks how often the pair puts the canine on the correct side, and what "correct" was measured against.

What is a tube shift checked against?

Three reference standards recur in these studies.

• The surgeon's finding. In patient studies, the true position is what was recorded at operation. One study added study models taken after the canine erupted [2, 3, 4, 6].
• A known laboratory position. One study set a simulated ectopic canine in a typodont built from human teeth, in nine positions, and photographed each one [5].
• CBCT. Several newer studies treat the CBCT reading as the answer and score panoramic or intraoral readings against it [7, 8, 9].

Each has a weakness. A surgical reference exists only for canines that went to surgery, so those samples are selected, and a systematic review of CBCT against conventional radiography rated most comparative studies at high risk of bias in subject selection [2, 4, 10]. CBCT is an imperfect reference too, though the evidence on this is early. In the two in vitro accuracy studies that review included, CBCT accuracy ranged from 50% to 95% [10, 11]. In a study of 17 patients checked against surgical exposure, CBCT readings differed significantly from the surgical finding only on canine shape and bony coverage [6].

None of the studies here reads the same pair of radiographs against both surgery and CBCT. None of these studies measures how much an accuracy figure moves with the choice of reference.

What the evidence shows

How accurate are horizontal and vertical parallax for impacted maxillary canines?

Sources: [2, 3, 4, 5].

Two of the four favour the horizontal shift. In the Belfast study, six orthodontists located 83% of 43 ectopic canines correctly with horizontal parallax and 68% with vertical parallax. They also marked their answer "unsure" in 12% of vertical readings and 5% of horizontal ones [3]. On the typodont, horizontal parallax scored 83% and vertical parallax 65% [5].

The records study went the other way, reporting 97.7% for vertical parallax and 92.6% for horizontal [4]. Its surgical confirmation covered only the cases that had surgery, and the abstract doesn't say how many that was.

The studies differ in reference, in size, and in how many cases were confirmed. The Royal College of Surgeons of England guideline sides with the Belfast result and states that horizontal parallax is more reliable than vertical. It adds that steepening the tube angle has been suggested as a way to improve the vertical shift [1, 12].

Vertical parallax has a practical draw. It uses the panoramic radiograph, which is often already taken, so it needs only one more exposure, the occlusal [12].

Does parallax work as well for buccal canines as for palatal ones?

It does not. Two studies with a surgical reference agree on the direction.

• With vertical parallax from a panoramic and an anterior occlusal radiograph, almost 90 per cent of palatal canines were located correctly and fewer than half of buccal canines [2].
• Sensitivity for palatal canines was 88% with horizontal parallax and 69% with vertical parallax. For buccal canines it was 63% with either technique [3].

The buccal figure in the second study rests on nine buccal canines, so read it as a direction more than a precise rate.

Judging position by magnification on one panoramic located one buccal canine in 10 [2]. Against CBCT, the magnification method placed 68.00% of buccal and 69.57% of palatal canines correctly, and the angulation method 28.57% of buccal and 84.91% of palatal canines [8]. In the overall cohort, a panoramic angle threshold of 50.4° did not reach significance for separating buccal from palatal canines, with a sensitivity of 52.6% and a specificity of 70.2% [9]. Both studies concluded that one panoramic radiograph is not reliable enough to localise the canine on its own [8, 9].

Where do two-view parallax and cone-beam CT disagree?

On the typodont, CBCT located the canine correctly in 94% of readings, against 83% for horizontal and 65% for vertical parallax [5]. That is a laboratory result on nine positions.

In patients, the disagreement shows up when the same canines are read twice. Seven judges read 25 impacted canines on a panoramic, an occlusal and two periapicals, then on CBCT. Their readings differed on buccal or palatal position in 16%, on the mesiodistal position of the cusp tip in 21%, and on root resorption of the adjacent teeth in 36% [13]. Across six studies in a systematic review, agreement between conventional radiographs and CBCT on canine position ranged from kappa 0.20 to 0.82, with observed agreement of 64% to 84% [10].

Where both are checked against a reference, CBCT comes out ahead for buccal or palatal position and vertical position [6, 14]. In one of those studies the 2D set was a panoramic and a lateral cephalogram, which performed similarly to CBCT for overall inclination and the mesiodistal position of the apex [14]. The other doesn't name its 2D views in the abstract [6]. Neither is a clean parallax comparison.

The British Orthodontic Society guideline sums up the position: three-dimensional imaging localises the canine more accurately, and in most cases conventional radiographs and CBCT agree [15].

How often a different reading changes the plan varies by study. Treatment decisions differed between the 2D set and CBCT in 27% of canines in one study [13], 24% of decisions in a survey of 112 orthodontists [16], and 9.8% of cases among 132 canines reviewed by an interdisciplinary team [17]. In 39 canines referred for surgery, surgical planning did not differ significantly between the panoramic radiograph and CBCT [18].

Where it's unsettled

• Which shift is better. Two studies favour horizontal parallax and one favours vertical, with different references and samples [3, 4, 5].
• The buccal canine. The shortfall is consistent, but the buccal groups are small [2, 3].
• The reference itself. Surgical samples are selected, CBCT is imperfect, and no study here measures parallax against both [10, 6].
• Whether a different reading matters. Plan changes after CBCT range from 9.8% to 27% across studies with different raters, cases and 2D sets [13, 17].

What this means at the chairside

When are two radiographs enough to localise the canine, and when does the case call for CBCT?

The British Orthodontic Society sets the sequence for an ectopic canine [15]:

1. Take at least two conventional radiographs that permit parallax. That can be two intraoral radiographs, or one intraoral radiograph and a panoramic.
2. Consider whether they are enough to make a treatment plan.
3. If they are, no further imaging is needed. If they aren't, consider localised CBCT.

UK guidance states there is no indication for routine CBCT [1]. Small field of view CBCT may be appropriate in selected cases where conventional radiographs don't give enough information to plan, and for equivocal cases or those with potential complications [1, 15]. It typically carries a higher effective dose than conventional radiography, and where there is doubt about taking one, a dental and maxillofacial radiology specialist's opinion should come first [1]. The American Dental Association and the American Academy of Oral and Maxillofacial Radiology recommend reviewing images already taken and using CBCT judiciously [19]. There is no strong evidence for CBCT as the first image [11].

Individual studies have named the features that justify further views or CBCT. These are the authors' proposals, and the guideline test above stays the rule.

• A canine that looks buccal on the panoramic radiograph and can't be palpated justifies further views [2].
• An interdisciplinary team judged CBCT indicated in 47% of cases. A buccal canine (odds ratio 5.3 against palatal) and horizontal angulation (odds ratio 10.9 against vertical) were among the 2D predictors [17].
• One group proposed small volume CBCT when canine inclination on the panoramic radiograph exceeds 30°, when root resorption of adjacent teeth is suspected, or when the canine apex can't be made out [20].

Read against the accuracy data, parallax located palatal canines more reliably than buccal ones [2, 3]. The guideline question stays the same for both: do the two views let you plan?

Parallax reads depth from two flat images. Try your eye on real radiographs.