Measurement of rectal tumor height from the anal verge on MRI: a comparison of internal versus external anal sphincter

Abstract

Purpose:

To determine the most accurate measurement technique to assess rectal tumor height on MRI using two different anatomic landmarks for the anal verge.

Introduction:

Accurate measurements and standardized reporting of MRI for rectal cancer staging is essential. It is not known whether measurements starting from the internal anal sphincter (IAS) or external anal sphincter (EAS) more closely correlate with tumor height from the anal verge on endoscopy.

Methods:

This retrospective study included baseline staging MRI examinations for 85 patients after exclusions. Two radiologists blinded to endoscopic results measured the distance of rectal tumors from the internal anal sphincter and external anal sphincter on sagittal T2 images. The reference standard was endoscopic measurement of tumor height; descriptive statistics were performed.

Results:

For reader 1, the mean difference in measurement of tumor height between MRI and endoscopy was −0.45 cm (SD ± 1.76 cm, range −6.0 to 3.9 cm) for the IAS and 0.51 cm (SD ± 1.75 cm range −4.7 to 4.8 cm) for the EAS. For reader 2, the mean difference in measurement of tumor height between MRI and endoscopy was −0.57 (STD ± 1.81, range −5.9 to 4.8 cm) for the IAS and 0.52 cm (STD ± 1.85, range −4.3 to 5.6 cm) for the EAS. Interobserver ICC was excellent between reader 1 and reader 2 for measurements from both the IAS (0.955 95% CI 0.931–0.97) and EAS (0.952, 95% CI 0.928, 0.969).

Conclusion:

Measurement of tumor height on MRI was highly reproducible between readers; beginning measurements from the EAS tends to slightly overestimate tumor height on average and from the IAS tends to slightly underestimate tumor height on average.

Introduction

Rectal cancer is a common malignancy in the United States. Although overall rates of colorectal cancer have declined in the past several decades, the incidence has increased for certain demographics, namely younger adults, and particularly for rectal cancer[1]. The distance as measured from the anal verge is significant, as it is one of the determinants of how a given patient is treated for rectal cancer[2] and may have prognostic implications in rectal neuroendocrine tumors[3].

Magnetic resonance imaging (MRI) has been established as the modality of choice in assessing local extent of disease at baseline as well as response to neoadjuvant therapy. Surgeons and radiation oncologists have come to rely on radiologists’ interpretations of rectal MRI for accurate staging and treatment planning(Jhaveri, 2009 #27;Klessen, 2007 #26). Tumor height from the anal verge is one of the key elements established on endoscopy and MRI; there is some data suggesting tumor height may affect time to recurrence and patterns of metastatic spread[4].

A standardized approach to MRI for staging of rectal cancer is imperative to consistency and reproducibility among radiology reports. There is evidence that it is easier to interpret for clinicians, improves quality of reporting, and reduces the need for clinicians to contact the radiologist for clarifications[5]. To that end, an expert panel from the Society of Abdominal Radiology’s Rectal and Anal Cancer Disease Focus Panel have put forth a consensus statement on appropriate lexicon in reporting of rectal MRI[6].

One of the key elements of rectal MRI reporting and an area of controversy is the reporting of tumor height from the anal verge. The anal verge is a histologic term used to describe the point at which the skin changes from hair bearing to non-hair bearing epithelium at the lower most portion of the anal canal. It is used as the reference point for endoscopic measurement of tumor height as reported by colorectal surgeons and gastroenterologists in their procedure reports. On MRI, the precise location of the anal verge is yet to be determined. Currently, the lower most point of the external anal sphincter (EAS) or internal anal sphincter (IAS) on sagittal images are often used as the MRI correlates to the endoscopic anal verge in clinical practice.

In this retrospective study, we aim to determine which structure on MRI, the EAS or the IAS, more closely corresponds to the endoscopic anal verge. Doing so could help clarify the growing literature on accurate measurement of tumor height and will help to further standardize interpretation and reporting of rectal MRI.

Methods

Patient selection

Initially, records for 180 consecutive patients with a diagnosis of rectal cancer beginning in July 2011 were reviewed. Subjects were excluded if there was no baseline MRI available for review (68 subjects) or the baseline MRI was acquired at an outside institution (16 subjects). After imaging was reviewed, an additional 11 patients were excluded if no mass was seen on MRI (7 subjects), no mass was seen on endoscopy (2 subjects), a technical problem with the exam prevented review of the study images (1 subject) or if the rectum was obscured by metallic artifact (1 subject). This resulted in a final cohort of 85 subjects.

Image acquisition

Rectal MR images were acquired on a 3-T MRI unit (Discovery MR750; GE Medical Systems, Waukesha, WI) per departmental protocol with a 32-channel phased array coil. Multiplanar T2-weighted images and diffusion-weighted images (DWI), including both b800 and 1500 s/mm², were acquired. All sequences acquired in the clinical routine were available to assist in tumor localization. The technical parameters for the sagittal T2-weighted fast spin echo images on which measurements were made are as follows: field of view 18 cm, slice thickness 4 mm, gap 1 mm, flip angle 90 degrees, bandwidth 32 kHz, and NEX 4.

Image interpretation and tumor height measurement

Two radiologists with fellowship training in body imaging (4 and 11 years experience interpreting rectal MRI) reviewed baseline MRI studies performed for staging of rectal cancer on the departmental picture archiving and communication system (PACS). Readers were blinded to the measurement of tumor height reported on endoscopy. Each reader measured the height of the tumor on the sagittal T2-weighted images from the inferior most margin of both the external anal sphincter and the internal anal sphincter. To establish tumor height, a non-bowed line was drawn from either the EAS or IAS to the top of the anorectal ring. A second non-bowed line was drawn from the top of the anorectal ring to the lower margin of the rectal tumor through the midline of the rectal lumen. The two lines were added to determine the tumor height, or distance from the anal verge. (FIGURE 1)

Figure 1:

37-year old female with rectal adenocarcinoma. Sagittal T2-weighted image through the rectum shows measurements as drawn from the internal anal sphincter (a, solid line), from the external anal sphincter (b, solid line) and from the anorectal junction to the inferior margin of the tumor (a and b, dotted lines). The T2-hyperintense fat located between the sphincters, the intersphincteric groove, is well seen in this case (c, arrowhead).

Reference standard

Endoscopic measurement of tumor height from the anal verge was the reference standard. Procedures were performed by referring colorectal surgeons or gastroenterologists. The decision to include different types of endoscopy (i.e. flexible and rigid endoscopy) was intentional, as it more closely reflects what is encountered in clinical practice.

Statistical analysis

The paired t test was used to compare IAS and EAS AV distance measurements. Linear regressions were applied to investigate the associations of the measurement from endoscopy with IAS measurement or EAS measurement. The intraclass correlation coefficient was estimated to assess interreader agreement, as well as the agreement between IAS, EAS and endoscopy measurements.

Results

The final cohort included 85 subjects, 50 male (59%) and 35 (41%) female, with a median age of 51 years (range 29–81 years). The median tumor height from the anal verge as measured via endoscopy was 8.0 cm (range 3.0 to 15.0 cm). (TABLE 1)

Table 1.

Summary of patient demographics and measurements for endoscopic measurements and MRI measurements from reader 1 and reader 2

Characteristic	N = 851
AGE	51 (29, 81)
GENDER
F	35 (41%)
M	50 (59%)
Distance from AV endoscopy (cm)	8.00 (3.00, 15.00)
Distance to tumor IAS, R1	8.30 (3.40, 13.80)
Distance to tumor EAS, R1	9.10(4.30, 14.80)
Distance to tumor IAS, R2	8.20 (2.30, 13.50)
Distance to tumor EAS, R2	9.30 (3.70, 14.20)
Difference IAS to endoscopy, R1	−0.50 (−6.00, 3.90)
Difference EAS to endoscopy, R1	0.60 (−4.70, 4.80)
Difference IAS to endoscopy, R2	−0.50 (−5.90, 4.80)
Difference EAS to endoscopy, R2	0.70 (−4.30, 5.60)

¹Statistics presented: median (minimum, maximum); n (%)

For reader 1, the median distance to the tumor as measured from the IAS was 8.3 cm (range 3.4 to 13.8 cm), whereas the median distance to the tumor as measured from the EAS was 9.1 cm (range 4.3 to 14.8 cm). The mean difference in measurement of tumor height from the IAS compared to endoscopy was −0.45 cm (SD 1.76 cm, p = 0.021). The mean difference in measurement of tumor height from the EAS compared to endoscopy was 0.51 cm (SD 1.75 cm, p = 0.009).

For reader 2, the median distance to the tumor as measured from the IAS was 8.0 cm (range 2.3 to 13.5 cm), whereas the median distance to the tumor as measured from the EAS was 9.3 cm (range 3.7 to 14.2 cm). The mean difference in measurement of tumor height from the IAS compared to endoscopy was −0.57 (SD 1.81, p = 0.005). The mean difference in measurement of tumor height from the EAS compared to endoscopy was 0.52 cm (SD 1.85, p = 0.011). There was a significant negative association for the difference between endoscopic measurement and measurement using both the IAS and EAS for both readers. There was a trend towards underestimating tumor height on MRI for tumors located higher in the rectum as measured on endoscopy. (TABLE 2) (FIGURES 2 and 3)

Table 2:

Comparison of MRI measurements to endoscopic measurements

Mean ± SD	p Value
Difference IAS, R1	−0.45 ± 1.76	0.021
Difference EAS, R1	0.51 ± 1.75	0.009
Difference IAS, R2	−0.57 ± 1.81	0.005
Difference EAS, R2	0.52 ± 1.85	0.011

Figure 2:

Boxplot of image distance differences from endoscopy.

Figure 3:

Scatter plot of image differences with endoscopy measurements on endoscopy. Plots with linear regression fitted curves and adjusted R-square estimates.

The intraclass correlation coefficients (ICC) were good for reader 1 between both the IAS and endoscopy (0.755, 95% CI 0.643–0.835) and the EAS and endoscopy (0.752, 95% CI 0.635–0.834). The intraclass correlation coefficient (ICC) was good for reader 2 between the IAS and endoscopy (0.747, 95% CI 0.623, 0.832) and moderate between the EAS and endoscopy (0.739, 95% CI 0.618–0.825). Interobserver ICC was excellent between reader 1 and reader 2 for measurements from both the IAS (0.955, 95% CI 0.931–0.97) and EAS (0.952, 95% CI 0.928, 0.969). (TABLE 3) (FIGURE 4)

Table 3:

ICC between MRI measurements to endoscopic measurements

Assessment	ICC (95%CI)
IAS Rl, Endoscopy	0.755 (0.643, 0.835)
EAS Rl, Endoscopy	0.752 (0.635, 0.834)
IAS R2, Endoscopy	0.747 (0.623, 0.832)
EAS R2, Endoscopy	0.739 (0.618, 0.825)
Interreader IAS	0.955 (0.931, 0.97)
Interreader EAS	0.952 (0.928, 0.969)
Intrareader IAS EAS, Rl	0.925 (−0.006, 0.983)
Intrareader IAS EAS, R2	0.913 (−0.01, 0.981)

Figure 4:

Bland-Altman plot between readers.

Discussion

In summary, the results of our study demonstrate that using either the IAS or EAS as the MRI correlate for the anal verge will on average result in a measured tumor height very close to endoscopy. Measurements beginning at the EAS tend to result in a measured tumor height that is slightly greater than endoscopy, and measurements beginning at the IAS tend to result in a measured tumor height that is slightly less than endoscopy. The agreement between measurements on MRI and endoscopy was good using both methods, and the interobserver agreement was excellent for both methods as well. Lastly, there was a significant trend towards underestimating tumor height on MRI as tumors were located higher in the rectum as measured on endoscopy.

From these key results, we can draw several conclusions. First, both the IAS and EAS are acceptable landmarks to begin measuring tumor height on MRI for rectal cancer, as they are highly reproducible between readers. Second, since the IAS and EAS tend to underestimate tumor height and overestimate tumor height, respectively, by similar margins, a logical compromise would be to measure from the fat plane in between the IAS and EAS, known colloquially as the intersphincteric groove. This would likely give results that on average are very close to endoscopic height, even though there will inevitably be some variability in individual cases. Third, for tumors in the upper rectum, MRI measurements tend to underestimate tumor height compared to endoscopy. This is not entirely surprising, as the upper rectum becomes more tortuous and winding and may thus be more difficult to precisely measure compared to the mid and lower rectum, which is relatively fixed by the mesorectal fascia.

Although many factors such as relationship to the anterior peritoneal reflection and anorectal junction are also important, several studies recently have tackled the question of how best to measure rectal tumor height on MRI [7–12], and give context to our findings. A few studies are worth noting specifically. One of the early studies reported a range up to 5.5 cm difference from endoscopic measurement[12], which is similar to some of the cases in our cohort. Additionally, a study by Attenberger et al [7]compared several different MR measurement techniques to assess rectal tumor height, and found that unbowed lines like those used in this study were acceptable in terms of accuracy, however the correlation with endoscopy was only fair. Lastly, Jacobs et al[10] similarly found high interobserver agreement, with a mean difference from endoscopy and MRI measurements at 2.5 cm, considerably larger than our results using either the EAS or IAS. Although these studies do address MRI for tumor height measurement in rectal cancer, they do not specifically compare methods to establish what is the best anal verge correlate on MRI.

The results of this study offer a few clinical implications. As mentioned above, the results suggest that the intersphincteric groove, the fat plane between the IAS and EAS, may be an appropriate landmark to begin measurements for tumor height on MRI. This could help to standardize measurements across institutions. Furthermore, radiologists must be aware that we may tend to underestimate tumor height for those lesions located in the upper rectum. Knowing this, a radiologist may be able to explain this discrepancy to surgical and medical colleagues when a considerable difference is seen for high tumors. However, the reasons for this final point are not fully understood, and may require further analysis to determine the underlying reasons behind it.

There are several limitations to this study. First, this was a retrospective study at a single institution, and we were not able to prospectively establish endoscopic practices by surgical colleagues. Additionally, MRI readers were blinded to the endoscopic results. Although this is an advantage in terms of preventing reader bias, it may have made lesion localization more difficult for smaller rectal tumors. Lastly, the decision to include both rigid and flexible sigmoidoscopy may be less precise than including only one type of endoscopy or another, even if it does more closely reflect clinical practice.

In summary, the results of this study show that measuring tumor height using the EAS and IAS both yield results that are on average very close to endoscopic measurements. On average, the EAS tends to overestimate tumor height by approximately 0.5 cm, and the IAS tends to underestimate tumor height by approximately 0.5 cm, and thus the intersphincteric groove may be a reasonable space in between that can serve as the landmark from which measurements can start on the sagittal T2 images. And when faced with a high rectal tumor, the interpreting radiologist should be aware that there is a significant tendency to underestimate tumor height, and this pattern may be useful to convey to referring clinicians if there is a discrepancy with the endoscopic findings.