Secondary Logo

Journal Logo


Curve Prevalence of a New Classification of Operative Adolescent Idiopathic Scoliosis: Does Classification Correlate With Treatment?

Lenke, Lawrence G., MD*; Betz, Randal R., MD; Clements, David, MD; Merola, Andrew, MD§; Haher, Thomas, MD§; Lowe, Thomas, MD; Newton, Peter, MD; Bridwell, Keith H., MD*; Blanke, Kathy, RN*

Author Information


The King classification system has remained the principal means of classifying thoracic adolescent idiopathic scoliosis (AIS). 8 Despite marked changes in the three-dimensional preoperative assessment 1,3,9 and the use of segmental spinal instrumentation in the last decade, 2,4,14,15 this coronal plane-only system has prevailed, 8 but it does have its shortcomings. It is neither comprehensive nor uniplaner, and the reliability of the system is suspect. 5,10 As a direct reflection of a multicenter scoliosis group (Harms Scoliosis Study Group) and difficulty with reproducible idiopathic scoliosis curve classification, a new system has been developed. 11 Six goals in formulating this surgical classification system were specific from the onset: 1) comprehensive with all types of AIS curves classified, 2) two-dimensional with applicability to three-dimensional assessment, 3) treatment based, 4) to separate out specific curve types by objective radiographic criteria, 5) highly reliable, and 6) logical, easily understood, and useful to scoliosis surgeons.

The purpose of this article is twofold: 1) define the prevalence of curve classification with this new system and 2) test the ability of this new classification system to correlate with regions of the scoliotic spine to be fused. Both of these will be assessed via a multicenter retrospective review of 606 consecutive AIS cases.


Curve Classification Triad.

There are three components to this new curve classification system: curve type, lumbar spine modifier, and sagittal thoracic modifier. Each of these three components should be identified separately and then combined together to create the complete classification triad.

Classification begins by reviewing the long cassette upright posteroanterior and lateral radiographs as well as right and left side-bending radiographs. The spinal columns are divided into three regions: proximal thoracic (PT), main thoracic (MT), and thoracolumbar/lumbar (TL/L). One must also keep in mind the regional apexes with a MT apex being located between the body of T2 inclusive to the T11–T12 disc, a thoracolumbar apex located from the body of T12 to the body of L1 including the T12–L1 disc, and lumbar apex extending from the L1–L2 disc to the body of L4 inclusive.

Curve Types 1–6.

Regional curves are separated into major (largest Cobb) and minor curves. Specific objective criteria in the coronal and sagittal planes determine whether the minor curves are structural or nonstructural. Structural criteria in the coronal plane include inflexibility on side-bending ≥ 25°; and in the sagittal plane PT (T2–T5) and thoracolumbar (T10–L2) kyphosis ≥ +20° (Table 1). Thus, each PT, MT, and TL/L minor curve is either designated as structural or nonstructural based on these criteria. The largest Cobb measurement is considered the major curve and thus is always structural in these operative cases. A template can then be created whereby six curve types are designated: Type 1, MT; Type 2, double thoracic (DT); Type 3, double major (DM); Type 4, triple major (TM); Type 5, TL/L; and Type 6, TL/L–MT (Table 2).

Table 1
Table 1:
Radiographic Surgical Structural Criteria for Minor Curves
Table 2
Table 2:
Curve Types 1–6

These curve type designations are meant to be treatment based by having the regions of the spine that are designated structural requiring instrumentation and fusion, whereas those nonstructural regions will not. Although not directly providing fusion levels, the curve type designation does implicate appropriate regions of the spine to be included in the instrumentation and fusion and those regions that should be left unfused (Table 3).

Table 3
Table 3:
Treatment Options as Designated by Curve Type

Lumbar Spine Modifier.

The lumbar spine is a mobile region and serves as the foundation of the spine and pelvis. The degree of lumbar deformity is an important determinant of spinal balance and success with scoliosis instrumentation and fusion. 7,12–15,17,19,20 Therefore, we have developed a lumbar spine modifier to classify the severity of the lumbar deformity in each scoliosis curve and to complement the specific curve Types 1–6.

Lumbar spine modifiers A, B, and C are based on the association of the center sacral vertical line (CSVL) to the lumbar spine on long cassette upright radiographs. For the lumbar spine modifier A, the CSVL lies between the lumbar pedicles up to the stable vertebra. The curve must have a major MT apex (curve Types 1–4), which excludes any TL/L curves (Types 5 and 6) (Figure 1).

Figure 1
Figure 1:
Lumbar modifiers A, B, and C are demonstrated on successive radiographs.

For lumbar modifier B, a major thoracic curve also exists, but the CSVL falls on the apex of the lumbar spine between the medial border of the lumbar concave pedicle and the concave lateral margin of the apical vertebral body or bodies (if the apex is a disc) because of the mild lateral deviation from the midline of the lumbar spine (Figure 1). The curve must also have a major MT apex (curve Types 1–4).

For lumbar modifier C, the CSVL falls completely medial to the concave lateral aspect of the TL/L apical vertebral body or bodies (if the apex is a disc). Thus, lumbar modifier C may exist with any of the curve Types 1–6, with curve Types 5 and 6 always having lumbar curve modifier C because of the necessary deviation from the midline of the apex of the major TL/L curve for the curve Types 5 and 6 designation (Figure 1).

Sagittal Thoracic Modifier (−, N, +).

Thoracic sagittal alignment is crucial in the formation of scoliosis, 6 the preoperative assessment of surgical indications, 3 the specific operative approach, 2 and instrumentation techniques used to correct the scoliosis. 1,3,7 Currently, instrumentation techniques are often principally guided by the thoracic sagittal profile to optimize sagittal alignment during concomitant coronal plane correction. 9,16,22 For these reasons we have devised a simple thoracic sagittal modifier to complement the six curve types and three lumbar spine modifiers.

The sagittal thoracic modifier is based on the upright lateral radiograph and measured from the superior endplate of T5 to the inferior endplate of T12 (T5–T12). When this measurement is <+10°, the sagittal modifier is designated as a “−,” or hypokyphosis; when the measurement is between +10° and +40°, it is designated as “N,” or normal kyphosis; and for measurements >+40°, the designation is “+,” or hyperkyphosis.

Complete Curve Classification.

Complete curve classification thus combines the specific curve Types 1–6 along with the lumbar spine modifier (A, B, and C) and the sagittal thoracic modifier (−, N, or +) to form the specific curve classification (e.g., 1A−, 1AN, 1A+, 1B−. . ., 6CN, 6C+). Although this produces a total of 42 possible curve classifications, if one follows the rule of assigning the appropriate curve Types 1–6 and then adding the appropriate lumbar spine and sagittal thoracic modifier separately, the specific curve classification follows logically.

Multicenter AIS Review.

A total of 606 consecutive AIS cases were reviewed from five scoliosis centers (Philadelphia, PA; New York, NY; Denver, CO; San Diego, CA; and St. Louis, MO). All cases were consecutive and accrued between 1992 and 1998. Each case was given a complete curve classification by the new system by the lead author. In addition, the surgical treatment was noted, specifically the regions (PT, MT, TL/L) of the spine that were included in the instrumentation and fusion. In this matter the ability of this new system to correlate with regions of the spine that should be included in instrumentation and fusion could be assessed because this new classification system was not specifically used when these surgeries were performed.


Curve Type

The prevalence of the six curve types noted was as follows: Type 1, MT (n = 309, 51%); Type 2, DT (n = 118, 20%); Type 3, DM (n = 69, 11%); Type 4, TM (n = 19, 3%); Type 5, TL/L (n = 74, 12%); and Type 6, TL/L–MT (n = 17, 3%) (Table 4).

Table 4
Table 4:
Curve Type Prevalence as a Consecutive Series of 606 AIS Operative Cases

Lumbar Modifier

The most common lumbar modifier was A (n = 247, 41%), followed by lumbar modifier C (n = 228, 37%) and then B (n = 131, 22%).

Sagittal Thoracic Modifier

The most common sagittal thoracic modifier noted was normal (“N”; n = 458, 75%), followed by hypokyphosis (“−”; n = 84, 14%), and least was hyperkyphosis (“+”; n = 64, 11%).

Curve Classification

The five most common curve classifications noted were as follows: 1AN (n = 114, 19%), 1BN (n = 66, 11%), 2AN (n = 63, 10%), 5CN (n = 62, 10%), and 1CN (n = 50, 8%). These five curve classifications accounted for 58% of all the curve classifications noted. There were five curve classifications that were not seen in our retrospective review: 2C+, 3A−, 3B−, 4A−, and 4B−. All other curve classifications were noted at least once (Table 5). All 606 curves were classifiable by this system.

Table 5
Table 5:
Prevalence of Complete Curve Classification Found in Consecutive 606 Adolescent Idiopathic Scoliosis Review

Ability to Predict Regions of the Spine to Be Fused

By designating specific curve Types 1–6, the regions of the spine that are structural and should be considered included in instrumentation and fusion are quite evident. Specifically, for curve Type 1 (MT) only the MT curve should be fused; for curve Type 2 (DT) both the PT and MT curves should be fused; for curve Type 3 (DM) the MT and TL/L curves should be fused; for curve Type 4 (TM) all three curves, PT, MT, and TL/L, should be fused; for curve Type 5 (TL/L) only the TL/L curve should be fused; and for curve Type 6 (TL/L–MT) both the MT and TL/L curves should be fused.

In review of all 606 curves treated, an average of 90% of the regions of the spine deemed structural by this new system were included in the instrumentation and fusion of these cases. For Type 1, 89% (275 of 309) of the MT curves were fused; for Type 2, 95% (112 of 118) of the PT and MT regions were fused; for Type 3, 90% (62 of 69) of the MT and TL/L regions were fused; for Type 4, 75% (14 of 19) of the PT, MT, and TL/L regions were fused; for Type 5, 93% (69 of 74) of the TL/L regions were fused; and for Type 6, 83% (14 of 17) of the MT and TL/L regions were fused (Table 6).

Table 6
Table 6:
Ability of the Curve Type to Correlate With Specific Regions of the Spine to Be Fused


We elected to begin our new classification with six curve types based on regional structural criteria of the spinal column. Simplistically, curve Types 1–4 (MT, DT, DM, and TM) all have an MT region that is also the largest curve (major) in most situations. However, an important component is that the surgical structural criteria of the PT and TL/L region may be found with hyperkyphosis in the PT and TL/L sagittal regions (T2–T5 and T10–L2 kyphosis ≥+20°) and/or with inflexibility on side-bending (≥25° of residual Cobb measurement). The surgical structural criteria of the sagittal plane were considered important to create a treatment-based classification system because hyperkyphosis in the PT and TL/L regions usually mandates consideration for including these regions of the spine in the instrumentation–fusion of the MT region. 13

Types 5 and 6 curves have the thoracolumbar or lumbar curve as the largest (major) curve. For the TL/L (Type 5) curve, the MT region is nonstructural, whereas the TL/L-MT (Type 6) curve has an MT region that is surgically structural because of inflexibility on side-bending (≥25°). These six curve types represent the entire spectrum of various surgical structural curves encountered in AIS. Importantly, we did not find any surgical cases in the 606 retrospective case reviews that were not classifiable by this new system.

The need for differentiating lumbar curve modifiers A, B, and C based on the position of the lumbar spine to the CSVL becomes apparent when specific treatment algorithms are being developed. Also, this has importance for grading the results with respect to lumbar spine positioning after surgical correction. 12 In this manner, if one advances from a lumbar modifier A to a modifier B to a modifier C, the lumbar curve obviously deviates more from the CSVL and correspondingly has more angular malalignment. Thus, the goal of surgical intervention, whether only the thoracic curve, thoracic and part of the lumbar curve, or the thoracic and the majority of the lumbar curve are instrumented and fused, should be to create as optimal a lumbar spine position as possible. In this respect, translational correction of the lumbar spine, either via instrumented correction maneuvers or via spontaneous correction after selective instrumentation and fusion of the MT curve, can be determined and quantified using the A, B, and C lumbar spine modifier nomenclature. This will be important when the results of surgical treatment are critically analyzed. 12,21 Lumbar modifier A was the most prevalent (n = 247, 41%), followed by modifier C (n = 228, 37%), and then modifier B (n = 131, 22%).

We also elected to have a sagittal thoracic modifier (−, N, or +) because of the great importance placed on this region of the spinal column to curve assessment and treatment. We divided the sagittal thoracic regional alignments into three groups (hypokyphosis “−,” normal kyphosis “N,” and hyperkyphosis “+”) based on previous data of “normal” alignment. 2 The normal “N” thoracic kyphosis modifier was the most prevalent (n = 458, 75%), followed by modifier “−” (n = 84, 14%), and lastly, modifier “+” (n = 64, 11%). In our reliability studies, the additional variable of −, N, or + did not lessen interobserver or intraobserver reliability. This was, indeed, the most reliable component of the system (kappa = 0.938 and 0.970 for intraobserver and interobserver reliability, respectively). 11 However, all these curves were premeasured by a single surgeon. Those few curves that have their thoracic sagittal Cobb measurements near the border of the three different types of sagittal modifiers (i.e., those with T5–T12 Cobb measurements of 9°, 10°, or 11°, or 39°, 40°, or 41°) may certainly pose measurement reliability difficulties. However, approximately 85% of patients had their preoperative sagittal thoracic Cobb between T5 and T12 measure >2° away from these two break-off points of 10° (e.g., ≤7° or ≥13°) and 40° (e.g., ≤37° or ≥43°). So these individual measurements should not change the reliability significantly.

The 1A (−, N, or +) curves are all MT curves with nonstructural PT and TL/L regions as was previously designated as King Type III curves. The 1AN curve was the most frequently encountered curve in our review (n = 114, 19%) (Figure 2). The 1B (−, N, or +) curves are MT curves previously classified as a King Type II or III curve, depending on exactly how far from the midline the lumbar curve deviated and the amount of structural characteristics present. 8,12 This curve was found to produce poor classification reliability by the King system because it was often graded as an intermediate curve between a true Type II and a true Type III curve. 10 However, it was the second most common curve pattern seen (n = 66, 11%) in our review (Figure 3). The 1C (−, N, or +) curves are MT curves previously classified as King Type II curves with the lumbar apex completely deviated from the midline. These curves often cause the most difficulty with coronal decompensation after selective thoracic fusion, but selective thoracic fusion may still be possible because the lumbar curve is flexible enough to side-bend to <25° and lacks a thoracolumbar junctional kyphosis. 4,13,17,19,23

Figure 2
Figure 2:
Classification 1AN is presented here on the upright anteroposterior and lateral radiographs. The MT region is the major overhang curve (53°) with the PT region above and the TL/L region below, nonstructural in the sagittal plane, as well on side-bending radiographs (17° and 6°, respectively).
Figure 3
Figure 3:
Curve Type 1BN. The MT region is the major curve (59°) with the PT region above and the TL/L region below nonstructural in both the lateral and side-bending radiographs (0° and 6°, respectively).

One of the important curve types presented by this new system is the 1C curve, which is an MT curve with a secondary nonstructural lumbar curve in which the apex completely crosses the midline. This 1C pattern has a lumbar component that side-bends to <25° and also lacks a thoracolumbar junctional kyphosis of >20° from T10 to L2. In our case review of 606 cases, there were five 1C+ curves, 10 1C− curves, and 50 1CN curves for a total of 65 1C curves. In reviewing how these 1C curves were treated by the different surgeons, 40 of 65 were treated with a selective thoracic fusion: 22 by the anterior route and 18 by the posterior route. The other 25 patients had posterior treatment of both curves. So according to the surgeons, it is possible to perform a selective thoracic fusion in these 1C curve types approximately 62% of the time. Obviously, as one approaches a larger lumbar curve that side-bends to very near 25°, there is going to be some debate as to whether this can adequately be treated with a selective fusion. These borderline cases need to be evaluated much more carefully to ensure that a selective fusion will be successful. This involves not only evaluating the side-bend measurement but also the ratio of thoracic to lumbar Cobb measurements, apical deviation, and apical rotation. It also requires careful evaluation of the clinical appearance of the patient as well as the patient’s overall spinal balance. Lastly, it requires surgical techniques to limit thoracic correction to maintain overall spinal balance and a willingness for both the surgeon and the patient/family to accept a lumbar curve with residual scoliosis that is mobile.

The 1CN curve pattern was the fifth most common pattern seen (Figure 4). The 2AN (DT) and 5CN (TL/L) were the third and fourth most common curve types noted in our review (Figures 5 and 6). Altogether, the Type 1 (MT) curves were the most common curve type (n = 309, 51%) of the six different curve patterns.

Figure 4
Figure 4:
Curve Type 1CN. The MT curve is the major curve (57°). Lumbar C position is noted. The left side-bender shows the nonstructural PT region above (24°) and TL/L region below (18°).
Figure 5
Figure 5:
Curve Type 2AN. The MT curve is the major curve (53°). On left side-bender the PT region is noted to be structural by bending out to only 29°. The lumbar curve is nonstructural (7°).
Figure 6
Figure 6:
Curve Type 5CN. The TL/L curve is the major curve (52°), with nonstructural PT and MT regions above.

It must be noted that just because a regional component of the curve is deemed “surgical structural” by our objective radiographic criteria, this does not absolutely mandate inclusion of a less “surgical structural” curve in the instrumentation and fusion of a more “surgical structural” curve based on radiographic and/or clinical analysis. As described above for the main thoracic curve pattern, it is apparent that the ratios of these surgical structural curves (thoracic: lumbar Cobb, apical vertebral translation, apical vertebral rotation, 18 and side-bending flexibility) may be more important than the absolute values, as has been thoroughly investigated when attempting to distinguish between a true King Type II versus a DM curve. 13 These theories will be further evaluated and tested in the future with curve classification-specific treatment algorithms that follow logically from this classification system.

Although rare, it certainly occurs that the sagittal plane minor structural criteria of kyphosis ≥20° in the proximal (T2–T5) and TL/L regions (T10–L2) would be the sole determinant of a surgical structural curve. Our 606 case reviews had 118 total Type 2 DT curves. The sagittal plane was structural from T2 to T5 without the coronal plane being structural in 13 of 29 of all DT curves with T2–T5 kyphosis ≥20°. In cases of Type 3, 4, or 6, those having a structural thoracolumbar junction without having coronal plane structural criteria present was found in 14 of 39 patients (36%) of those with T10–L2 kyphosis ≥20°.

An acknowledged potential criticism of our new classification system is the possibility of 42 separate curve classifications produced. However, the five most common classifications (1AN, 1BN, 2AN, 5CN, and 1CN) accounted for 58% of all curve classifications. We found that 37 of 42 possible classifications were noted in our case review with the following five not seen: 2C+, 3A−, 3B−, 4A−, and 4B−. It is unknown if these would be seen in a larger review; however, we certainly can anticipate that these curve patterns could all exist in AIS.

Finally, the ability of this classification system to correlate with operative treatment was evaluated by reviewing the surgical cases, assessing the curve types, and determining whether the surgically structural regions of the spine were fused and the nonstructural regions unfused in these 606 cases. Overall, 90% of all cases had the predicted surgically structural regions of the spine arthrodesed, and the nonstructural regions left mobile. This potentially confirms the structural radiographic criteria used to separate out structural versus nonstructural curves. However, just because the regions of the spine were included in the instrumentation–fusion as predicted by this new classification system, this does not guarantee that the “best” surgical results were produced. Ideally, comparison of the surgical results of these curves having the predicted regions of the spine included in the instrumentation–fusion versus those curves not following these guidelines must be performed to validate the benefits of this new treatment-directed classification system. Future multicenter prospective analyses are being performed to investigate this and will require a variable radiographic scoring system.


A new comprehensive classification system for AIS found all 606 consecutive AIS cases classifiable. Type 1 MT curve was the most common curve type noted in 51% of all cases, and the complete curve classification 1AN (n = 114, 19%) was the most commonly seen. Finally, this new classification system appears to be correlate with treatment of surgically structural regions of the spine in 90% of cases by the objective radiographic criteria used.

Key Points

  • A new comprehensive classification system of AIS found all 606 consecutive AIS cases classifiable.
  • The main thoracic curve pattern, Type 1, accounted for 51% of all cases.
  • This classification system correlated with treatment of the surgically structural regions of the spine in 90% of all cases by objective radiographic criteria.


1. Bernhardt M, Bridwell KH. Segmental analysis of the sagittal plane alignment of the normal thoracic and lumbar spine and thoracolumbar junction. Spine 1989; 14: 717–21.
2. Betz RR, Harms J, Clements DH, et al. Comparison of anterior versus posterior instrumentation for correction of adolescent thoracic idiopathic scoliosis. Spine 1999; 24: 225–39.
3. Bridwell KH, Betz RR, Capelli AM, et al. Sagittal plane analysis in idiopathic scoliosis patients treated with Cotrel-Dubousset instrumentation. Spine 1990; 15: 921–6.
4. Bridwell KH, McAllister JW, Betz RR, et al. Coronal decompensation produced by Cotrel-Dubousset “derotation” maneuver for idiopathic right thoracic scoliosis. Spine 1991; 16: 769–77.
5. Cummings RJ, Loveless EA, Campbell J, et al. Interobserver reliability and intraobserver reproducibility of the system of King et al for the classification of adolescent idiopathic scoliosis. J Bone Joint Surg Am 1998; 80: 1107–11.
6. Dickson RA. The etiology and pathogenesis of idiopathic scoliosis. Acta Orthop Belg 1992; 52 (suppl): 21–25.
7. Kalen V, Conklin M. The behavior of the unfused lumbar spine following selective thoracic fusion for idiopathic scoliosis. Spine 1990; 15: 271–4.
8. King HA, Moe JH, Bradford DS, et al. The selection of fusion levels in thoracic idiopathic scoliosis. J Bone Joint Surg Am 1983; 65: 1302–13.
9. Lee CK, Denis F, Winter R, et al. Analysis of the upper thoracic curve in surgically treated idiopathic scoliosis: a new concept of the double thoracic curve pattern. Spine 1993; 18: 1599–608.
10. Lenke LG, Betz RR, Bridwell KH, et al. Intraobserver and interobserver reliability of the classification of thoracic adolescent idiopathic scoliosis. J Bone Joint Surg Am 1998; 80: 1097–106.
11. Lenke LG, Betz RR, Harms J, et al. Adolescent idiopathic scoliosis: a new classification to determine extent of spinal arthrodesis. J Bone Joint Surg Am 2001; 83: 1169–81.
12. Lenke LG, Betz RR, Harms J, et al. Spontaneous lumbar curve coronal correction after selective anterior or posterior thoracic fusion in adolescent idiopathic scoliosis. Spine 1999; 24: 1663–71.
13. Lenke LG, Bridwell KH, Baldus C, et al. Preventing decompensation in King type II curves treated with Cotrel-Dubousset instrumentation: strict guidelines for selective thoracic fusion. Spine 1992; 17: 274–81.
14. Lenke LG, Bridwell KH, Baldus C, et al. Cotrel-Dubousset instrumentation for adolescent idiopathic scoliosis. J Bone Joint Surg Am 1992; 74: 1056–67.
15. Lenke LG, Bridwell KH, Blanke K, et al. Radiographic results of arthrodesis with Cotrel-Dubousset instrumentation for the treatment of adolescent idiopathic scoliosis: 5 to 10 year follow up. J Bone Joint Surg Am 1998; 80: 807–14.
16. Lenke LG, Bridwell KH, O’Brien MF, et al. Recognition and treatment of the proximal thoracic curve in adolescent idiopathic scoliosis treated with Cotrel-Dubousset instrumentation. Spine 1994; 19: 1589–97.
17. Lonstein J. Decompensation with Cotrel-Dubousset instrumentation: a multi-center study. Presented at the Anniversary Meeting of the Scoliosis Research Society, Minneapolis, Minnesota, September, 1991.
18. Nash CI, Moe JH. A study of vertebral rotation. J Bone Joint Surg Am 1968; 51: 223–9.
19. Richards BS, Birch JG, Herring JA, et al. Frontal plane and sagittal plane balance following Cotrel-Dubousset instrumentation for idiopathic scoliosis. Spine 1989; 14: 733–7.
20. Richards BS. Lumbar curve response in type II idiopathic scoliosis after posterior instrumentation of the thoracic curve. Spine 1992; 17 (suppl): 282–6.
21. Roye DP Jr, Farcy JP, Rickert JB, et al. Results of spinal instrumentation of adolescent idiopathic scoliosis by King type. Spine 1992; 17 (suppl): 270–3.
22. Shufflebarger HL, Clark CE. Fusion levels and hook patterns in thoracic scoliosis with Cotrel-Dubousset instrumentation. Spine 1990; 15: 916–20.
23. Thompson JP, Transfeldt EE, Bradford DS, et al. Decompensation after Cotrel-Dubousset instrumentation of idiopathic scoliosis. Spine 1990; 15: 927–31.

operative adolescent idiopathic scoliosis; new classification system; major curve; minor curve; structural criteria]Spine 2002;27:604–611

© 2002 Lippincott Williams & Wilkins, Inc.