How accurately does a bitewing show periodontal bone loss?

A bitewing shows periodontal bone loss, but it tends to show less of it than is really there [5]. Against bone measured at surgery, radiographic bone level underestimated loss in every one of 17 comparisons in a meta-analysis of infrabony defects, by a mean of 1.10 mm [1]. A 2026 systematic review put the underestimation at 0.5 mm to 2.8 mm [2]. Most of the surgical comparisons behind those numbers studied advanced disease or infrabony defects rather than bitewings in early disease. They bound the bitewing's accuracy more than they measure it directly.

Advanced loss is obvious on a radiograph. For incipient disease, a 1991 bitewing study noted there was no consensus on what counts as bone loss at all [3].

What the question actually is

"Accuracy" here means how far the bone level you read on the radiograph sits from the bone level found with the flap open. That surgical measurement is the reference standard in most of the studies below [1][5][6][7][8][9]. One review also counted clinical attachment level as a reference [2].

What is a normal bone level on a bitewing, and where do you measure from?

On a bitewing, you measure from the CEJ to the alveolar crest apical to it. Both studies below measured it that way [3][4].

One 1991 study measured bitewing sites that had no clinical attachment loss. It found a CEJ-to-crest distance between 0.4 and 1.9 mm (95% confidence limits) [3]. The study's indexing lists adolescents and molars, so read it as a young-patient range.

In an adult sample of 84 Swedish men aged 26 to 54, the mean CEJ-to-crest distance at premolar and molar bitewing sites was 1.00 to 1.12 mm across the three groups studied [4]. Those men were described as apparently healthy, not screened for periodontal health. Their mean is a population figure, not a cutoff for healthy bone.

If you've been taught that the crest should sit within about 2 mm of the CEJ, that fits the upper end of the 1991 range.

How do you tell horizontal from vertical bone loss on a bitewing?

Picture two lines between neighbouring teeth. One joins their CEJs, and the other joins the two alveolar crest levels. If the lines are parallel, the bone loss is horizontal. If they aren't, a vertical defect is present [22].

How far off parallel counts as vertical is a judgment call. In a 2026 study, 70 periodontists from 22 countries classified radiographs edited so the crest sat 0° to 35° off the CEJ line. Calls of vertical rose with the angle, with a marked shift between about 20° and 30°. A 35° angle reached the study's 70% agreement mark at seven of eight sites [22]. The periodontists called mandibular sites vertical more often, and a tilted CEJ line led to more false calls of vertical. Individual periodontists differed significantly [22].

Horizontal loss is the more common pattern. On 150 panoramic radiographs of patients with chronic periodontitis, 92.2% of affected teeth had horizontal bone loss and 7.8% had a vertical defect. That study counted a defect as vertical when it met the root at 45° or less and was at least 3 mm deep [23]. Among 416 people seeking dental care, 30.2% had at least one vertical defect of 3.0 mm or more on full-mouth intraoral radiographs. Mesial defects were more common than distal ones [24].

The AAP staging chart for the 2017 classification lists mostly horizontal bone loss under stages I and II, and vertical bone loss of 3 mm or more as a stage III complexity factor. A complexity factor may shift the stage to a higher level [17]. Angular defects also went with more tooth loss. In 201 people examined 10 years apart without systematic periodontal therapy, 13% of teeth with an even pattern of bone loss were lost, against 22%, 46% and 68% of teeth with angular defects of increasing depth [25].

Much of the accuracy evidence in the next section was measured on vertical defects: infrabony defects in the meta-analysis [1] and interproximal vertical defects in [8].

What the evidence shows

How closely do radiographic measurements match clinical or surgical bone levels?

Against surgical measurement, the direction is settled: the radiograph reads short. A meta-analysis of 11 studies (17 comparisons) of infrabony defects found radiographic bone level underestimated the surgical bone level by a mean difference of −1.10 mm (95% CI −1.34 to −0.85). Clinical attachment level underestimated it by −1.22 mm (95% CI −1.49 to −0.95). The two didn't differ significantly (−0.05 mm, P = .76) [1]. Clinical attachment level isn't a way around the radiograph. It missed by about as much.

The size of the miss is less settled. A 2026 systematic review of 30 studies (1,645 adults) found intraoral radiographs had a sensitivity of 60% to 90% against flap surgery or clinical attachment level. Correlation with clinical measurement was moderate (r = 0.5 to 0.8 across 13 studies), and radiographs underestimated bone loss by 0.5 mm to 2.8 mm [2]. The review pooled periodontitis with peri-implantitis.

The individual studies show the same pattern:

Look at the standard deviations. In [7] and [9] they're larger than the mean, so a single site can be off by far more than the average suggests. Readers disagree too. In one study the average underestimation varied between clinicians from 2.2 to 3.2 mm on film and 2.1 to 3.0 mm on digital [6]. Another called interobserver variation "substantial" [5]. In [9], intraoral radiographs correlated with the surgical measurement at r = 0.53 to 0.67.

The two clinical methods that came closest were bone sounding and probing to bone. Bone sounding without a flap missed by 0.6 mm on average [6]. Probing to bone differed from the surgical measurement by a mean of 0.1 mm [8].

What changes the measurement: angulation, view type, image processing?

View type. In one study that compared views against open-flap measurement, periapical radiographs were significantly more accurate than bitewings (p < 0.001). Bitewings underestimated bone loss by 11–23%, periapicals by 9–20% [5]. In that study, bitewing and periapical images were enlarged by about 8% in the maxillary arch and 4–5% in the mandibular arch [5]. For how the two views compare more broadly, see bitewing vs periapical radiograph.

Bitewing orientation matters as well. A 2020 study reviewed 320 dental-school records of patients with posterior bone loss who had both types. Vertical bitewings showed the level of bone loss in 52.81% of sets against 3.75% for horizontal bitewings. They showed furcation areas in 43.44% against 1.25% [10]. Horizontal bitewings more often showed the whole anatomical crown [10]. This measured what was visible, not accuracy against surgery.

Angulation. The beam angle is the obvious suspect, and elongation and foreshortening cover what it does to the picture. For the bone-level number itself, the sources here are thin. What is established is narrower. Detecting small changes at the crest by subtraction radiography depends on reproducing the same beam-to-patient geometry between serial radiographs [11].

Receptor and processing. In 25 periodontitis patients, direct digital and film radiographs of the same sites gave significantly different bone levels. Digital images showed more sites with bone loss [12]. The processing evidence is contested: the in vivo filter study found no gain, and an in vitro receptor study found gains with specific settings:

• On digitized radiographs checked against surgery, basic filters didn't make bone-loss measurements more valid than the unchanged image. One filter ("mean value") underestimated significantly more [13].
• In that study, radiographic underestimation of the CEJ-to-defect distance ranged from 0.3 ± 2.0 to 0.8 ± 1.9 mm [13].
• In dry pig mandibles (in vitro), inverted digital images underestimated defect depth compared with film [14].
• In cadaver specimens (in vitro), receptors with higher contrast resolution and a dedicated periodontal filter gave more accurate measurements [15].

The same film-versus-digital question runs through caries detection, with its own evidence.

Where it's unsettled

Does a radiograph underestimate early bone loss, and why?

The early-loss evidence is thin and it doesn't all point the same way.

A 1982 study cut defects of increasing size into the posterior maxilla of one human skull. There, a standardized radiographic method underestimated bone loss below about 30% and overestimated it above about 60% [16]. That's an in vitro result on a single skull, using the Bjorn technique on standardized film.

A 2017 surgical-comparison study in periodontitis patients found the opposite gradient. Underestimation was larger in premolar and molar regions and increased with defect depth, and deep defects in those regions were the most underrated [6]. The settings differ: stepwise defects in dry bone against real periodontal defects measured at surgery. The two results don't cancel out. Read each within its own setting.

What these sources don't settle is the reason. They show that the radiograph reads short and where it reads shortest. None of the studies retrieved for this article states a tested mechanism for why early loss escapes the radiograph, so this article doesn't give one.

Three more gaps are worth keeping in mind:

• Bitewing-specific data are scarce. The meta-analysis [1] and most primaries [6][7][8][9] studied infrabony defects or advanced disease.
• Angulation is unquantified for a single image. The retrieved sources establish its importance for serial comparison [11]. They don't give a figure for how much a given angle error shifts one reading.
• The headline range is a pooled one. The 0.5 mm to 2.8 mm figure [2] mixes periodontitis with peri-implantitis.

What this means at the chairside

Every surgical comparison cited here found the radiograph read short [1][5][6][7][8][9], and some sites are off by much more than the average.

Use the radiograph alongside the clinical examination, not instead of it. The 2026 review called periapical and bitewing radiographs "the most accurate and appropriate tools for oral health professionals to use as an adjunct to clinical evaluation" [2]. In infrabony defects, clinical attachment level underestimated surgical bone level by about as much as the radiograph [1].

Staging leans on attachment loss first. In the 2017 AAP/EFP classification, initial stage is determined from clinical attachment loss. Radiographic bone loss is used when attachment loss isn't available [17]. Severity is judged with reference to root length [18], and radiographic bone loss is scored by root third [17]:

As indirect evidence of progression, grade uses percentage bone loss divided by age: <0.25 for grade A, 0.25 to 1.0 for grade B, >1.0 for grade C [17].

Guidance leaves the choice of view to clinical judgment. The ADA/FDA selection criteria say imaging "may consist of, but is not limited to, selected bitewing and/or periapical images of areas where periodontal disease (other than nonspecific gingivitis) can be identified clinically." On those images you assess the level of supporting alveolar bone, the interproximal bony crest, root length and shape, bone loss in furcations, and calculus deposits [19].

The 2026 ADA/AAOMR update, as reported by ADA News, calls a 2D full-mouth radiographic series plus a clinical examination the gold standard for evaluating periodontal disease. Frequency depends on clinical findings and treatment response. It states there is currently no evidence to support CBCT in managing periodontal disease, except for treatment planning of complex cases [20].

Selection has a cost. A 1995 study of 490 new adult patients compared periapicals chosen from clinical signs, plus posterior bitewings, against a complete set. Of 2,415 teeth with proximal or furcal bone loss, 152 sites of bone loss (6%) were missed by the selected set [21]. Where posterior bone loss is present, one study found vertical bitewings showed the bone level far more often than horizontal ones [10].

You know where the crest should sit. Try reading real bitewings. → Start practising