Clinical investigation: cervix
Prediction of radiotherapy outcome using dynamic contrast enhanced MRI of carcinoma of the cervix

https://doi.org/10.1016/S0360-3016(02)02972-3Get rights and content

Abstract

Purpose: To investigate whether analysis of MRI enhancement data using a pharmacokinetic model improved a previously found correlation between contrast enhancement and tumor oxygenation measured using Po2 histograph. To evaluate the prognostic value of gadolinium enhancement data for radiotherapy outcome, and to study the efficacy of combined enhancement and MRI volume data.

Methods and materials: Fifty patients underwent dynamic gadolinium-enhanced MRI as part of their initial staging investigations before treatment. Gadolinium enhancement was analyzed using the Brix pharmacokinetic model to obtain the parameters amplitude and rate of contrast enhancement. Pretreatment tumor oxygen measurements (Eppendorf Po2 histograph) were available for 35 patients.

Results: Both standard and pharmacokinetic-derived enhancement data correlated with tumor oxygenation measurements, and poorly enhancing tumors had low tumor oxygen levels. However, only the pharmacokinetic-analyzed data correlated with patient outcome and patients with poorly (amplitude less than median) vs. well-enhancing tumors had significantly worse disease-specific survival (p = 0.024). For the 50 patients studied, no relationship was found between enhancement and volume data. Combining MRI volume and enhancement information highlighted large differences in outcome (p = 0.0054). At the time of analysis, only 55% of patients with large, poorly enhanced tumors were alive compared with 92% of patients with small, well-enhanced tumors.

Conclusion: These preliminary results suggest that pharmacokinetic modeling of dynamic contrast-enhanced MRI provides data that reflect tumor oxygenation and yields useful prognostic information in patients with locally advanced carcinoma of the cervix. Combining MRI-derived enhancement and volume data delineates large differences in radiotherapy outcome.

Introduction

The presence of hypoxia in solid tumors has a detrimental influence on treatment response and clinical outcome, and the ability to detect poorly oxygenated tumors before definitive treatment would be an extremely useful prognostic tool. It would enable identification of poor-risk patients who might benefit from additional treatment modalities, such as chemotherapy or surgery, and could lead to efficient use of hypoxia-modifying therapies that have been shown to improve the outcome of radiotherapy (RT) in a number of tumor sites (1).

Presently, the Eppendorf Po2 histograph is regarded as the reference standard for assessing tumor oxygen levels (2) and is the only method for measuring hypoxia that has proven prognostic value in a number of tumor sites 3, 4, 5. However, the method has limited clinical application, because it is invasive, can only be used on accessible tumors, and is not widely available. There is, therefore, interest in investigating alternative methods of measuring tumor hypoxia. The techniques of interest include measuring the level of binding of hypoxia-specific probes 6, 7, quantifying the expression of hypoxia-inducible factors 8, 9, and the comet assay (10). Noninvasive imaging techniques are also attractive, because the equipment required should be available in cancer centers and could be applied to inaccessible tumors. There are a number of possibilities based on cross-sectional imaging. These include contrast-enhanced dynamic CT (11), blood oxygen level-dependent MRI (12), and dynamic contrast-enhanced MRI (13).

Mayr et al. (13) demonstrated the prognostic value of assessing tumor enhancement, using gadolinium and dynamic MRI, in cervical carcinoma treated with RT. High levels of tumor enhancement before therapy or early in the course of therapy were associated with good local control. Furthermore, Yamashita et al. (14) showed histologically that in cervical tumors, poorly enhancing areas of dynamic MR images contain fewer capillaries and more abundant fibrous tissue. Both authors suggested that high levels of contrast enhancement might reflect better tumor oxygenation. To investigate this suggestion, we initiated a prospective study of patients with locally advanced cervical carcinoma. Pretreatment tumor measurements were made of both gadolinium enhancement using a standard model to analyze the dynamic MRI data (13) and tumor oxygenation using an Eppendorf Po2 histograph. We found a significant positive correlation between the level of contrast enhancement and tumor oxygenation (15). The conclusion from that study was that noninvasive dynamic MRI might provide a method for measuring hypoxia in human tumors. However, it is also possible to analyze dynamic MRI data using pharmacokinetic models, and the two most widely used are those derived by Tofts 14, 16 and Brix 17, 18.

Dynamic MRI, based on the uptake of gadolinium diethylenetriamine pentaacetic acid (DTPA), has also been shown to improve the depiction of cervical carcinoma (19) and to enhance the accuracy of assessing bladder or rectal involvement (20). It allows volume measurements to be made on a pixel-by-pixel basis. Tumor extent is known to be an important prognostic factor in carcinoma of the cervix (21). Evidence from a number of tumor oxygenation studies suggests that the level of hypoxia is independent of tumor volume 22, 23. The ability to obtain accurate volume measurements and information about tumor biology from a single noninvasive investigation, therefore, has immense prognostic potential.

The hypothesis behind the present study was that dynamic MRI parameters reflect tumor oxygenation and thus should predict patient outcome after RT. Specifically, the work had three aims: (1) to investigate whether analysis of MRI enhancement data using a pharmacokinetic model improved the correlation found between contrast enhancement and oxygenation data; (2) to evaluate the prognostic value of gadolinium enhancement data for RT outcome; and (3) to study the efficacy of combined enhancement and MRI volume measurements.

Section snippets

Patients

Between May 1996 and December 1999, 50 patients with locally advanced carcinoma of the cervix (FIGO Stage IB-IVA) underwent dynamic contrast-enhanced MRI as part of their staging investigations. Patients with Stage IB or IIA had bulky tumors with evidence of nodal disease or were considered medically unfit for surgery. Of the 50 patients, 3 had adenocarcinoma and 5 adenosquamous cell carcinoma. Tumor oxygenation data were also available for 35 of these patients. It was not possible to obtain

MRI dynamic data

Intertumor (Table 1) and intratumor (Fig. 1) variations were found in the MRI parameters. Histograms of amplitude and kep values could be classified into four main groups: those with low values for both amplitude and kep; those with high amplitude and low kep; and those with a bimodal distribution of amplitude and kep. Of particular note was the presence of a bimodal distribution in certain cases, suggesting distinct tumor populations. Table 2, Table 3 summarize the distribution of the

Discussion

A number of other small studies have examined the prognostic value of dynamic contrast-enhanced MRI in cervical cancer. Although at first sight their findings appear to be discordant, it is clear that differences exist in the methods of choosing the ROI and analyzing the data. Hawighorst et al. (18) performed a histologic analysis of the whole tumor specimen and chose their ROI to coincide with the area of highest angiogenicity (measured as intratumor microvessel density). They observed that

Conclusion

The results of the present study have shown that pharmacokinetic modeling of dynamic contrast-enhanced MRI provides data that reflects tumor oxygenation and yields useful prognostic information for patients with locally advanced carcinoma of the cervix. Combined MRI-derived enhancement and volume data can separate large differences in RT outcome.

References (29)

  • M. Hockel et al.

    Association between tumor hypoxia and malignant progression in advanced cancer of the uterine cervix

    Cancer Res

    (1996)
  • J.A. Raleigh et al.

    A clinical study of hypoxia and metallothionein protein expression in squamous cell carcinomas

    Clin Cancer Res

    (2000)
  • S.M. Evans et al.

    Detection of hypoxia in human squamous cell carcinoma by EF5 binding

    Cancer Res

    (2000)
  • R. Airley et al.

    Glucose transporter Glut-1 expression correlates with tumor hypoxia and predicts metastasis free survival in advanced carcinoma of the cervix

    Clin Cancer Res

    (2001)
  • Cited by (0)

    Supported by the Christie Hospital Endowment Fund and a European Commission Biomed concerted action (contract number BMH4-98-3006).

    View full text