Neer Proximal Humerus Fracture Classification

Count the displaced segments to get a one-, two-, three- or four-part fracture — then read the published kappas, which run from 0.03 to 0.57 and include an observer agreeing with themselves at 0.20.

Last updated 8 Oct 2026

Neer part count from the displaced segments

Three segments + articular surface

Greater tuberosity — displaced more than 1 cm or angulated more than 45°?Neer divides the proximal humerus into four parts — humeral head, greater tuberosity, lesser tuberosity and shaft — and “A fracture part is considered displaced if angulation exceeds 45°, or if the fracture is displaced by more than 1 cm.” The 1 cm is the threshold people argue about most, because greater tuberosity displacement of well under a centimetre is held by many surgeons to matter for the rotator cuff, and Neer’s threshold does not see it. A fracture line through a segment is not enough: the segment has to have moved.

Lesser tuberosity — displaced more than 1 cm or angulated more than 45°?The same threshold applied to the subscapularis insertion. Isolated lesser tuberosity displacement is uncommon; its main role in the count is in three- and four-part patterns, where both tuberosities separate from the head. It is also the segment most often obscured on an anteroposterior view, which is part of why adding CT did not improve agreement in the studies below.

Humeral shaft at the surgical neck — displaced more than 1 cm or angulated more than 45°?The surgical neck component. Angulation is the criterion that usually decides this one, and 45° is a large amount of angulation — a 40° varus surgical neck fracture is, by Neer’s own threshold, not displaced, and therefore a one-part fracture if nothing else has moved. Readers who expect “one-part” to mean “barely injured” should know that.

Articular surface and glenohumeral jointNeer’s system has groups beyond simple part counting. A fracture-dislocation is a part count plus a dislocation of the head out of the glenoid; the head-splitting and impression fractures are articular-surface injuries in their own right. Radiopaedia also records that Neer later added a fifth group, the valgus-impacted four-part fracture, which behaves differently from the classic four-part pattern because part of the head’s blood supply may be preserved. That group is named here and not assigned, because the finding that defines it is a measurement this page does not take.

Three-part fracture — two segments displaced from the headExample

Greater tuberosity displaced 15 mm, lesser tuberosity undisplaced, surgical neck angulated 60°, head located in the glenoid with an intact articular surface

Four segments, counted when they have moved

parts = 1 + (greater tuberosity displaced) + (lesser tuberosity displaced) + (shaft displaced)
displaced = more than 1 cm of displacement OR more than 45° of angulation

displaced segments, not fracture lines: “A fracture part is considered displaced if angulation exceeds 45°, or if the fracture is displaced by more than 1 cm.” A one-part fracture may have fracture lines through all four parts. A comminuted proximal humerus that has not moved is a one-part fracture
the head is always a part: so one displaced segment gives a two-part fracture, two give a three-part, and three displaced “with respect to the fourth” give a four-part. The arithmetic is 1 plus the count, which is why “two-part” and “two segments displaced” are different statements
the threshold is the weak point: 1 cm and 45° are measured by eye off a radiograph. In a 16-surgeon study the two-part kappa values ran 0.03 to 0.07 — the supposedly easiest category had the worst agreement, which points at the threshold rather than at the counting
CT did not help: 16 surgeons, 16 fractures, 48 image sets: Neer kappa 0.14 on radiographs, 0.09 on 3D CT and 0.07 on 2D CT. A separate study is titled “3D-radiographic analysis does not improve the Neer and AO classifications of proximal humeral fractures” and found mean kappa 0.44 then 0.49 across 7 observers and 24 cases
the groups beyond the count: fracture-dislocation and the head-splitting or impression fractures are their own groups, and Neer later added a fifth for the valgus-impacted four-part fracture. A bare part count loses all of them
derivation: Neer CS 2nd, “Displaced proximal humeral fractures. I. Classification and evaluation”, J Bone Joint Surg Am 1970 Sep;52(6):1077–1089, modified in 1975. The paper was not reachable for this page; the four segments and the 1 cm / 45° criterion are quoted from Radiopaedia and corroborated independently by a second source

Worked example

Greater tuberosity displaced 15 mm, lesser tuberosity undisplaced, surgical neck angulated 60°, head located in the glenoid with an intact articular surface
The articular surface is intact and the head is located, so neither of the first three rules fires
Greater tuberosity displaced 15 mm, which is more than 1 cm: counts
Lesser tuberosity undisplaced: does not count
Surgical neck angulated 60°, which is more than 45°: counts
Two segments displaced, so parts = 1 + 2 = three-part fracture
Reduce the greater tuberosity displacement to 8 mm and the surgical neck angulation to 40° and every segment becomes undisplaced by Neer's threshold — a one-part fracture, on the same bone with the same fracture lines. That is the threshold doing the work, and it is why the two-part kappa in the 16-surgeon study was 0.03 to 0.07
Dislocate the head instead and the answer becomes a three- or four-part fracture-dislocation, because the dislocation is tested before the count

Seven published kappas for one classification

Interobserver kappa for the Neer classification spans 0.07 to 0.49 across these three studies — a sevenfold range for one instrument — and the within-study range runs from 0.03 to 0.57. The intraobserver column is the one to dwell on: in 1997 an observer re-reading the same 26 radiographs agreed with their own earlier answer at kappa 0.20, and in 1999 the worst self-agreement was 0.27. A classification an observer cannot reproduce against themselves is not measuring the fracture. Note too that the two higher-scoring studies used a four-category version and the lowest-scoring one used the full system, so part of the spread is how many categories were on offer.
Study and reading	Observers and images	Interobserver kappa	Intraobserver kappa
Foroohar 2011, plain radiographs	16 surgeons in 4 experience groups, 16 fractures (48 image sets in all)	0.14	not measured
Foroohar 2011, 3D CT reconstruction	same	0.09	not measured
Foroohar 2011, 2D CT	same	0.07	not measured
Foroohar 2011, two-part fractures only	same	0.03 to 0.07	not measured
Foroohar 2011, whole range across all groups and modalities	same	0.03 to 0.57	not measured
Sjödén 1997, first viewing	10 observers (5 orthopaedic surgeons, 5 radiologists), 26 fractures	0.42	0.20 to 0.85
Sjödén 1997, second viewing	same	0.45 in the abstract, 0.43 in the results — the paper disagrees with itself	as above
Sjödén 1999, first assessment	7 observers, 24 cases	0.44	0.27 to 0.73
Sjödén 1999, second assessment two months later	same	0.49	individual values 0.57, 0.67, 0.73, 0.66, 0.56, 0.27, 0.48

What the part count does and does not capture

The left column is a list of ways two careful readers can produce different Neer categories from one set of films, and it is a better explanation of the kappa figures above than any claim about observer skill. The classification asks for two measurements by eye and then discards everything that is not a segment displacement.
Finding	Effect on the part count	Why it is reported separately here
Fracture lines through all four parts, nothing displaced	One-part	The count is of displaced segments. “One-part” does not mean “undisplaced simple fracture” and does not mean minor
Surgical neck angulated 40° in varus	Does not count	45° is a large threshold. A 40° varus neck is a one-part fracture by the published criterion
Greater tuberosity displaced 7 mm	Does not count	Many surgeons regard tuberosity displacement well under 1 cm as significant for the cuff; Neer’s threshold does not see it
Head dislocated out of the glenoid	Unchanged	A fracture-dislocation is its own Neer group, so this page reports it rather than folding it into a number
Head-splitting or impression fracture	Unchanged	An articular-surface injury is its own group and is tested before the count on this page
Valgus-impacted four-part pattern	Four-part	Neer added a fifth group for it. This page names it and does not assign it, because the defining measurement is not among the inputs

Counting by eye, across a threshold nobody can hold steady

Neer’s system divides the proximal humerus into four parts — the head, the greater and lesser tuberosities and the shaft — and counts how many have moved. A part is displaced if it has shifted more than a centimetre or angulated more than 45 degrees. One displaced segment gives a two-part fracture, two give a three-part, and three displaced with respect to the fourth give a four-part. Fracture-dislocations and the head-splitting and impression fractures are separate groups, and Neer later added a fifth for the valgus-impacted four-part pattern.

The first thing to get right is that the count is of displaced segments and not of fracture lines. A proximal humerus broken into four pieces that have not moved is a one-part fracture. A surgical neck angulated 40 degrees in varus is a one-part fracture. Neer’s 1970 paper was explicitly about displaced proximal humeral fractures; the undisplaced group is there as the contrast, not as a description of a minor injury.

The second thing is that this classification does not reproduce. Sixteen orthopaedic surgeons in four experience groups, reading 16 fractures on radiographs, two-dimensional CT and three-dimensional CT, produced Neer kappas of 0.14, 0.07 and 0.09 respectively — cross-sectional imaging made agreement worse — with every value in the study falling between 0.03 and 0.57. Two-part fractures, the easiest category, scored 0.03 to 0.07. Two earlier studies using a four-category version reached means of 0.42 and 0.44, and one of them reported that observers re-reading the same 26 radiographs agreed with themselves at kappas as low as 0.20. One of those papers is titled “3D-radiographic analysis does not improve the Neer and AO classifications of proximal humeral fractures”, which is about as plain as the literature gets.

The explanation is in the criterion rather than in the observers. The system asks for a centimetre of displacement and 45 degrees of angulation to be judged by eye on a projection of a complex three-dimensional shape, and then throws away everything that is not a segment displacement. Two readers who place the same greater tuberosity at 9 mm and 11 mm produce different Neer categories from one radiograph. This page counts the segments honestly and prints the kappas next to the answer, because a category with that much observer variance should never be reported as though it were a measurement.

A classification is not a score. Garden IV is not one unit worse than Garden III, a Salter-Harris IV is not twice a II, and nothing on this page adds up — these are named categories, not an ordinal measurement, and a mean or a difference taken over them has no meaning. This page states which published category the entered findings fall in and what the literature reports for that category in the cohorts it was measured in. It renders no clinical or surgical decision: not whether to operate, not which implant, not whether to take or withhold an image, and not whether a limb can take weight. Those are the treating clinician’s, and the classification is one input among many. Agreement between observers is part of what a classification means. A category two experienced readers assign differently on the same film is not a measurement of the fracture, and the published kappa values for these systems run from almost nothing to almost perfect. Each page here prints them with the number of observers and the number of radiographs they came from.

Frequently asked questions

How is a Neer part count worked out?

Count how many of the greater tuberosity, lesser tuberosity and shaft are displaced more than 1 cm or angulated more than 45 degrees, then add one for the head. One displaced segment is a two-part fracture, two is a three-part, and three displaced with respect to the fourth is a four-part. Fracture lines that have not produced displacement do not count.

Can a comminuted fracture be a one-part fracture?

Yes, and this is the commonest misreading of the system. Radiopaedia’s reproduction of the criteria states that a one-part fracture may have fracture lines involving one to four parts, with none of them displaced beyond 1 cm or 45 degrees. A proximal humerus in several undisplaced pieces is a one-part fracture; so is a surgical neck angulated 40 degrees.

How reliable is the Neer classification?

Poorly, and the figures are among the worst published for any fracture classification. Across 16 surgeons and 16 fractures the interobserver kappa was 0.14 on radiographs, 0.09 on 3D CT and 0.07 on 2D CT, with all values between 0.03 and 0.57. Two studies of a four-category version reached means of 0.42 and 0.44. In one of them, observers re-reading the same 26 radiographs agreed with their own earlier reading at kappas as low as 0.20.

Does CT improve the Neer classification?

No, in both studies that tested it. In the 16-surgeon study agreement was best on plain radiographs (0.14) and worse on 3D CT (0.09) and 2D CT (0.07). A separate study of 7 observers and 24 cases is titled “3D-radiographic analysis does not improve the Neer and AO classifications of proximal humeral fractures”. Whether CT is useful for planning an operation is a different question and not one about the classification.

Is a four-part fracture twice as bad as a two-part fracture?

The question does not have an answer, because the part count is a category and not a measurement. It counts displaced segments and nothing more: it does not record the head’s blood supply, the articular surface, a dislocation or the bone quality. A mean Neer part count over a series of patients is not a quantity, and the arithmetic difference between two and four has no interpretation.

Where is the valgus-impacted four-part fracture?

It is named on this page and not assigned. Radiopaedia records that Neer later added a fifth group for it, and it behaves differently from the classic four-part pattern because part of the head’s blood supply may be preserved. The finding that defines it is a measurement of the head’s position that is not among this page’s inputs, so the page does not claim to identify it.

Related calculators

References

Neer CS 2nd. Displaced proximal humeral fractures. I. Classification and evaluation. J Bone Joint Surg Am. 1970 Sep;52(6):1077–1089. Citation read in AO Surgery Reference’s classic-references list for the proximal humerus; the paper itself was not reachable, so the criteria below are quoted from named reproductions.
Radiopaedia. Proximal humeral fracture classification (Neer). “The Neer system divides the proximal humerus into four parts” — humeral head, greater tuberosity, lesser tuberosity and humeral shaft. “A fracture part is considered displaced if angulation exceeds 45°, or if the fracture is displaced by more than 1 cm.” One-part: fracture lines involve one to four parts and none is displaced. Two-part: one part displaced. Three-part: two parts displaced. Four-part: “all 4 parts” involved with three displaced “with respect to the fourth”. The page also records that Neer later added a fifth group, the valgus-impacted four-part fracture.
VOKA. Proximal Humerus Fractures: Etiology, Classification, Treatment, and Rehabilitation. Corroborates the displacement criterion independently: the system “is based on the relationship between four main anatomical segments: the humeral head, the greater tuberosity, the lesser tuberosity, and the shaft” and “A segment is considered a separate ‘part’ if it is displaced by more than 1 cm or angulated by more than 45 degrees”. Also that the system “was introduced in 1970 and modified in 1975”.
Foroohar A, Tosti R, Richmond JM, Gaughan JP, Ilyas AM. Classification and treatment of proximal humerus fractures: inter-observer reliability and agreement across imaging modalities and experience. J Orthop Surg Res. 2011;6:38. Sixteen orthopaedic surgeons in four groups of four (upper-extremity specialists, general orthopaedists, senior residents, junior residents) classified 16 fractures on radiographs, 2D CT and 3D CT — 48 image sets. For the Neer classification, “X-ray > 3D CT reconstruction > 2D CT scan with the kappa values being 0.14, 0.09, 0.07 respectively”; “All kappa values ranged from ‘slight’ to ‘moderate’ (k = .03 to .57) agreement”; two-part fractures “ranged from 0.03-0.07”; general orthopaedists “(0.04-0.11)” and senior residents “(0.03-0.21)”. Interobserver only; no intraobserver values.
Sjödén GOJ, Movin T, Guntner P, Aspelin P, Ahrengart L, Ersmark H, Sperber A. Poor reproducibility of classification of proximal humeral fractures: additional CT of minor value. Acta Orthop Scand. 1997;68(3):239–242. Ten observers (five orthopaedic surgeons and five radiologists), 26 proximal humeral fractures with plain radiographs and CT. “The Neer system had a kappa value of 0.42” at the first viewing. The abstract gives the second viewing as 0.45 and the results section gives 0.43 — the paper disagrees with itself, and both figures are printed on this page. Intraobserver “kappa range 0.20-0.85”.
Sjödén GO, Movin T, Aspelin P, Güntner P, Shalabi A. 3D-radiographic analysis does not improve the Neer and AO classifications of proximal humeral fractures. Acta Orthop Scand. 1999;70(4):325–328. Seven observers, 24 cases, assessed twice two months apart. “The Neer system had a mean kappa value of 0.44” and at the second assessment “the mean kappa values were 0.49”. “Intraobserver reproducibility was fair to substantial agreement for Neer (kappa range 0.27-0.73)”, with individual observer values of 0.57, 0.67, 0.73, 0.66, 0.56, 0.27 and 0.48.

Not medical advice. For healthcare professionals and education. Reference intervals vary by laboratory and assay — always use your own laboratory's. Never base a dose or a treatment decision on this page alone. Full disclaimer at calcengines.com/disclaimer/