When should you retake a dental radiograph?

Every error in radiographic technique ends at the same decision: expose again, or read what you have. It is a trade-off rather than a rule. A retake costs the patient another exposure, and guidance is explicit that any exposure has to be justified by clinical benefit ². A flawed image costs something too: a finding missed, or a finding invented. The literature gives you rates and causes, but what counts as an acceptable retake rate is not settled — the 2021 review that pooled the rates proposes a reference value on limited data and calls on the field to establish recommended levels ¹.

What are you actually deciding?

Start with the vocabulary, because the studies do not share it. A 2021 review of 26 studies counts rejects ¹. A single-institution analysis of 9,495 intraoral and 2,339 extraoral radiographs counts repeats ⁴. A retrospective series of 50,000 intraoral radiographs counts retakes ⁵. The three overlap heavily and are not the same measurement, which matters later when the numbers disagree.

The decision itself has two halves, and they pull in opposite directions.

The first is dose. An ADA expert panel's position, published in 2024, is that clinicians should only perform radiographic imaging "after reaching the professional judgment that there is a clear clinical benefit from the imaging exam, and that this benefit outweighs the risks associated with exposure to ionizing radiation" ². Nothing in that sentence exempts the second exposure. A retake is an exposure, so it has to clear the same bar the first one did.

Alongside it sit the two named principles. ALARA — "As Low as Reasonably Achievable" — means "all X-ray exposures are justified in relation to their benefits; necessary exposures are kept as low as reasonably achievable" ². ALADA is the companion: keeping dose "as low as diagnostically acceptable" ³. Read ALADA carefully, because it is the one that does the work here. Diagnostically acceptable is a threshold, not a standard of perfection. It sets a floor the image has to clear, and says nothing about the distance above it.

The second half is diagnosis. ADA policy, adopted in 1995 and amended in 2007 and 2016, states that images "must be correctly identified and be of diagnostic quality" ². That is the obligation on the other side of the trade: an image that cannot support the call is not a cheaper option than a retake, it is a worse one.

So the question in front of you is not whether the image is good. It is whether this image is diagnostic for the task you took it for, and whether a second exposure is justified by what the first one failed to show.

How often are radiographs retaken, and why?

Rates are not scarce. Agreement is.

The 2021 review pooled reject rates across 26 studies ¹:

Read the right-hand column, not the left. A periapical reject rate reported somewhere between 2.58% and 34.42% is not a number with noise on it; it is several different situations being counted under one heading. The review's own account is that the included studies were "largely heterogeneous in terms of the operator, patient age, and factors associated with reject rate" ¹. Operator type is one of the things the studies differed in. The review reports that "Operator type and related work experience also influenced the reject rate", while adding that "With such limited data, it remained to be elucidated if their reject rates were significantly different" ¹. In the single-institution analysis, the repeat rate for intraoral radiographs was highest for internees at 44.7% and undergraduate students at 28.2% ⁴.

The causes are more consistent than the rates. Across the reviewed studies, "the commonest errors leading to reject were positioning error and cone cut" for intraoral imaging ¹. The review's conclusions name "positioning error and patient movement" as the two common reasons for the rejects "of intra-oral and extra-oral imaging, respectively" ¹. The single-institution analysis puts numbers on it: positioning error accounted for 38.7% of reported intraoral errors, improper angulation for 26.1%, and improper receptor placement for 11.2% ⁴. In the 50,000-image series from one dental college, which used photostimulable phosphor plates, cone cut was the most common operator error, at 988 images ⁵.

Periapicals get retaken more than bitewings do. The pooled periapical mean sits about five points above the bitewing mean ¹, and in the 50,000-image series periapicals were 80.8% of intraoral retakes ⁵. The review's explanation is anatomical: periapical radiography requires the receptor to go deep enough to capture the root apices, where it "most likely will touch the palatal vault or the floor of the mouth, causing discomfort and gagging", while bitewing radiography "covers the tooth crowns only and hence the receptor usually is more tolerated by the patient" ¹.

Is there an acceptable retake rate?

This is the part that is unsettled, and the honest answer is that the number you want has not been settled on.

The 2021 review's own recommendation is that reject analysis "should be encouraged, so that the international DMFR community could establish recommended reference levels/values for quality assurance programs to follow" ¹. That is a statement about future work. The review does offer a figure — "with the limited data available in the existing literature, the overall reject rate for intra-oral radiography stood at approximately 9% and this should be a reference value for institutions to consider achieving" ¹ — but note what that is. It is one review's proposal, on its own account of limited data, not a threshold issued by a professional body.

Individual audits do grade against published standards. One audit of 506 student-taken digital periapicals, graded against Health Protection Agency standards, found 45.7% at Grade 1, 37.6% at Grade 2, and 16.7% at Grade 3 — "requiring a repeat radiograph; and were diagnostically unacceptable" ⁶. That tells you a standard can be applied to a stack of images. It does not tell you the rate a practice should be hitting, and one audit of one undergraduate clinic is not a benchmark.

Whether digital receptors made things worse is also open. The 2021 review reports that the studies it examined "suggested that the reject rate was higher for digital imaging than film, and particularly higher for intra-oral sensor than photostimulable phosphor plate", while recording that "the underlying reasons were not investigated in depth" ¹. The sharpest single comparison is old: a 1998 study in which three radiography technicians exposed 50 teeth with film and with a CCD sensor found "six per cent of dental films required retakes compared with 28% with the sensor" ⁷. A 1998 CCD sensor is not a current one, so treat that gap as a historical observation rather than a current performance figure.

There is one more finding, and it is the one that should change how you use all of the above. The review lists as a limitation that many studies reported common errors rather than common reasons for reject, and states plainly that "errors do not necessarily lead to rejects" ¹.

That is the whole point. An error is a property of the image. A reject is a decision about the image.

How do you decide at the chairside?

Name the question before you judge the image. You did not take a radiograph; you took it to see something. Proximal surfaces on these four teeth. The apex of that root. The crestal bone in this sextant. Whatever it was, it is the standard the image gets measured against, because ALADA sets a diagnostic floor rather than a target of perfection ³.

Then ask where the flaw sits relative to that question.

A cone cut that clips a surface you need is a different object from the same cone cut over a surface you do not. Overlapped contacts that obscure the two surfaces you are checking are not the same as overlap at the far end of the image. Elongation or foreshortening matters if you were measuring; it may not if you were looking for a radiolucency. The same question applies to a herringbone artifact or a panoramic ghost image. Each of those entries describes what the flaw looks like and why it appears; what no entry can tell you is whether it lands on your question, and that is the part that decides the retake.

If the flaw does not touch the question, the image is diagnostic for the task and a second exposure buys nothing you can justify ². Record what is wrong with it, and read it for what it can answer — not for what it can't.

If the flaw does touch the question, the image is not a cheaper answer. It is an absent one, and reading it anyway is how a finding gets missed or invented. That is the case where a retake is the correct clinical decision rather than an admission of a bad first attempt.

Two things follow. First, a retake needs a reason you could say out loud, and "it isn't a nice image" is not one — the justification test applies to the second exposure as much as the first ². Second, if you retake, change something. The commonest errors leading to reject are positioning error and cone cut ¹; in the single-institution analysis they were positioning, angulation and receptor placement ⁴. Repeating the same geometry generally reproduces the same image at twice the dose.

The dose side of this is not hypothetical, and clinicians already behave as though it is real. CBCT has the lowest pooled reject rate of the modalities reviewed, at 2.77% across five studies with a range of 1.64–20.25%, and one explanation the review offers is that "CBCT imaging had higher radiation dose than plain radiography, so operators tended to avoid rejects and thus retakes" ¹. If that potential reason holds, the threshold for accepting an imperfect image already moves with what a repeat would cost. Worth knowing that about yourself.

One caution to close on. Deciding a flaw is harmless is a diagnostic judgement, and it can be wrong in the same direction twice — you can talk yourself into reading an image you should have repeated, and you can also mistake a technique artifact for something real, which is a separate question with its own answer. Both errors are quieter than a retake, and neither shows up in a reject-rate audit.