Useful Parameters in Dynamic Contrast-enhanced Ultrasonography for Identifying Early Response to Chemotherapy in a Rat Liver Tumor Model

INTRODUCTION

Colorectal cancer (CRC) is the third most common cancer worldwide and the second most common cause of cancer-related death in developed countries.^[1] Metastatic CRC is treated by chemotherapy using cytotoxic and/or molecular targeted drugs, among which irinotecan is widely used.^[2,3]

The standard method for evaluating treatment response on computed tomography (CT) and magnetic resonance imaging (MRI) is Response Evaluation Criteria in Solid Tumors version 1.1,^[4] but as these are based solely on a reduction in tumor size, accurate early evaluation of the efficacy of chemotherapy can be difficult.^[5]

Dynamic imaging modalities such as ultrasound (US), CT, and MRI have been used to evaluate the therapeutic effect and to quantify changes in perfusion parameters early after initiation of chemotherapy.^[6-8] As a result, various mathematical models have been applied to analyze the chemotherapeutic response.

Compared with other modalities, functional dynamic contrast-enhanced ultrasonography (DCE-US) is a less invasive and more repeatable technique for measuring tumor perfusion; therefore, it has the potential for evaluating tumor response to chemotherapy.^[9] The previous studies have reported a significant association of several parameters with tumor response to chemotherapy,^[10-14] but with variable results, and no consensus exists regarding the best method for evaluating early response.

Several studies have suggested that structural changes (e.g., necrosis or relative microvessel density [MVD]) induced by chemotherapeutic agents cause alterations in intensity that is apparent on DCE-US before tumor growth can be detected.^[15,16] The previous pathological studies have confirmed MVD as an independent predictor of disease-free and overall survival.^[17] Indicators of tumor vascular density (relative blood volume and relative blood flow) can be determined from the time-intensity curve (TIC), which portrays the kinetics of microbubble contrast agent flow through the tumor.^[18,19] Wei et al. demonstrated excellent correlations between myocardial blood flow and TIC parameters in a mathematical model of DCE-US,^[20] using only time- and intensity-related parameters, which correlated with flow velocity and microvessel cross-sectional area, respectively. Therefore, TIC enables the evaluation of the correlation between parameters and structural changes in tumors simpler in the mathematical model. However, to the best of our knowledge, no investigation has applied a mathematical model for evaluating early tumor response on DCE-US. Therefore, we evaluated the correlation in an experimental animal study with clinically used contrasts.

The purpose of this study was to determine the parameter that best correlates with tumor growth on the TIC of DCE-US in a mathematical model and verify whether this parameter could assess early response to chemotherapy in a rat colorectal liver metastasis model.

MATERIAL AND METHODS

Experimental protocol and US techniques

Figure 1 shows the experimental protocol. An intravenous catheter for the injection of contrast medium and saline/ irinotecan was placed through the tail vein, under general anesthesia. DCE-US was used to measure maximum tumor diameter before injection of saline/irinotecan (day 0) and again on days 3 and 10 after injection. Maximum tumor diameter (mm) was measured and DCE-US data were obtained using the following settings: LOGIQ7 diagnostic US system (GE Healthcare, Milwaukee, WI, USA) with a 9L probe, 2D mode as used for B-mode US of the breast, gain 36, dynamic range 51, and rate 31 frames/ second, harmonic mode as used for CEUS of the breast, gain 74, dynamic range 51, rate 18 frames/second, and low mechanical index 0.18 using harmonic imaging. A skilled operator scanned US in a constant procedure and under similar conditions. DCE-US was performed in the plane of maximum tumor diameter. Sonazoid (0.3 mL/kg; Daiichi Sankyo, Tokyo, Japan) was used as the contrast medium, administered intravenously at a flow rate of 2.4 mL/min through a multiprogramming syringe pump (FP-1000, Melquest, Toyama, Japan). DCE-US scanning was performed at 0–85 s after injection from the arterial phase to the parenchymal phase. Imaging parameters (gain, depth, and the injection speed) were the same for all subjects. Besides that, we performed DCE-US to keep the scanning plane as identical as possible under respiratory fluctuation.

Close

Figure 1:

Experimental protocol. Maximum tumor diameter was measured on days 0, 3, and 10, and DCE-US was performed on days 0 and 3. Saline was injected in the first step (n = 6) and irinotecan in the second step (n=14). DCE-US: Dynamic contrast-enhanced ultrasonography.

Export to PPT

Measurement of tumor diameter and parameters on the TIC

Relative change in tumor size was calculated between the tumor sizes recorded on days 0 and 10. The region of interest (ROI) was set in the plane of maximum tumor diameter, and the mean signal intensity of the ROI was measured automatically for every frame. The TIC data were processed by a workstation in LOGIQ7. A fitted curve was obtained for the TICs of the following function, the differentiated function, and the slope at t = 0:

$\text{F (t) = PI }\times (1-{\exp }^{-\text{kt}})+\text{B}$ $\text{F'(t)=PI}\times \text{K}\times {\exp }^{-\text{kt}}$ $\text{F'(0)=PI}\times \text{k=slope}.$

The following four parameters were measured from the fitted curve: Peak intensity (PI, dB), k value, wash-in slope (PI × k, dB/second), and time to peak (TTP, second), as shown in Figure 2. PI was defined by calculating the differential value in signal intensity from the baseline before and after contrast enhancement on DCE-US. The k value was defined as the flow velocity.^[20] Relative change of each parameter was calculated on days 0 and 3.

Close

Figure 2:

DCE-US and the fitted curve constructed from the raw data. DCE-US was conducted for 85 s and the following four parameters were extracted from the fitted curve: PI, k value, wash-in slope (PI × k), and TTP. DCE-U: Dynamic contrast-enhanced ultrasonography, PI: Peak intensity, TTP: Time to peak.

Export to PPT

In the first step, we evaluated the Pearson’s correlation coefficient between the relative change in tumor size over 10 days and that of each parameter in the control group for 3 days, to identify the most correlative parameter with tumor progression. In the second step, we classified each tumor as either a responder (defined as an increase in tumor size of <20%) or a non-responder (defined as an increase in tumor size of ≥20%) on day 10 compared with the size before treatment. We then compared the responders and the nonresponders in the treatment group in terms of relative change in the most correlated parameter as determined in the first step.

RESULTS

The mean maximum tumor diameter on day 0 was 12.3 mm (95% confidence interval [CI]: 10.4–14.2 mm) in the control group and 14.2 mm (95% CI: 12.6–15.9 mm) in the treatment group. There was no significant difference in mean maximum tumor diameter before treatment between the responder (14.6 mm, 95% CI: 12.6–16.9 mm) and non-responder (13.9 mm, 95% CI: 11.2–16.7 mm) groups (P = 0.702).

Assessment of tumor response

In the second step, there were six responders and eight nonresponders in the treatment group on day 10 [Figure 3]. Table 3 lists the mean tumor sizes and the mean relative changes for these two groups. Table 4 and Figure 4 show the mean parameters and the mean relative changes in the two groups. Mean growth ratios on days 3 and 10 were +2.81% and +5.62% in the responder group and +5.47% and +36.8% in the non-responder group, respectively. There was a significant difference between them in relative change of size on day 10 (P = 0.0004). On the other hand, there was no significant difference between the two groups in terms of relative change of size on day 3 (P = 0.602).

Close

Figure 3:

Representative responder and non-responder fitted curves. Time-intensity curves on days 0 and 3 and fitted curves after modelization at baseline for a responder (a, b, and e, respectively) and a non-responder (c, d, and f, respectively). In the responder, TTP shortened 1.2 s and PI decreased 1.2 dB. In the non-responder, TTP shortened 5.2 s and PI increased 3.9 dB. The relative change of TTP in the non-responder was shorter than in the responder. TTP: Time to peak, PI: Peak intensity.

Export to PPT

Table 3: Relative change of tumor size in the treatment group(n=14).

	Mean tumor size(mm)			Relative change(%)
	Day 0	Day 3	Day 10	Day 0–3	Day 0–10
Total(n=14)	14.2±0.80	14.8±0.83	17.4±1.09	4.33±0.71	23.4±5.25
Responder(n=6)	14.6±0.99	15.0±0.93	15.4±1.10	2.81±1.32	5.62±3.70
Non-responder(n=8)	13.9±1.25	14.7±1.33	18.9±1.56	5.47±0.52	36.8±4.80

Data are means±standard deviations

Table 4: Mean values of TIC parameters in the treatment group.

	Responder(n=6)		Non-responder(n=8)
	Day 0	Day 3	Day 0	Day 3
TTP(second)	28.3±2.81	28.9±8.60	32.67±1.80	23.3±2.75
k value	0.157±0.017	0.160±0.015	0.133±0.013	0.226±0.022
PI(dB)	12.7±2.89	16.3±2.56	11.7±1.43	16.4±1.69
Slope(dB/second)	2.20±0.64	2.79±0.75	1.63±0.33	3.69±0.52

Data are means±standard deviations. TTC: Time-intensity curve, TTP: Time to peak, PI: Peak intensity

Close

Figure 4:

Relative change in the four DCE-US parameters between days 0 and 3 (second step). The k value was significantly lower in the responder group (4.96%) than in the non-responder group (72.5%, P=0.003). Significant difference was also found for TTP between the responders (2.45%) and non-responders (−27.4%) groups (P = 0.044). DCE-US: Dynamic contrast-enhanced ultrasonography, TTP: Time to peak, PI: Peak intensity. The error bars represent the standard deviation of each data. *: P<0.05, **: P<0.01.

Export to PPT

In contrast, the mean relative change in k value in the responder group on day 3 was significantly lower than that in the non-responder group (4.96 vs. 72.5, P = 0.003, Figure 4). In addition, the mean relative change in TTP on day 3 was significantly longer in the responder group than in the non-responder group (2.45 vs. −27.4, P = 0.044). There was no significant difference in PI or PI × k (53.9 vs. 50.8, P = 0.92; 69.2 vs. 168, P = 0.12, respectively). In summary, there were significant differences in k value and TTP between the responder and non-responder groups on day 3 even though there was no significant difference in size.

The animals showed gradually reducing body weight at each evaluation day. Reduction in body weight did not show significant differences between the control and treatment groups (mean 7.7% and 6.3%, P = 0.40, respectively).

DISCUSSION

Our study demonstrated that k value and TTP correlated strongly with tumor progression and could assess early tumor response to chemotherapy in a rat colorectal tumor model. The results suggest that time-related parameters (k value and TTP) are more suitable than the intensity-related parameter (PI) for assessing early tumor response on day 3 after treatment.

The previous studies have shown that early changes in TTP may predict tumor growth^{[12,13,22,23]} and that TTP was elongated in a treatment group compared with a control group.^[13,14] It has been suggested that TTP elongation is caused by a decrease of MVD within the tumor.^[24] Wang et al. reported that chemotherapy using cytotoxic agents decreases tumor MVD.^[25] The element flow velocity in a capillary is proportional to the fourth power of the radius and inversely proportional to length.^[26] Changes in MVD can influence capillary flow velocity.^[27]

Wei et al. reported k value as the flow velocity and TTP as inversely proportion to k value.^[20] TTP (second) and PI (dB) are independent of each other in the fitted curve. The flow is also proportional to the slope (PI × k), derived by the following exponential function:

Likewise, the present results demonstrated that k value and TTP showed opposite correlations with tumor progression.

Due to baseline fluctuation, analysis error occasionally occurs in determining the peak of a fitted logistic curve (as used in our study); consequently, the TTP obtained from a fitted curve can be a subjective parameter. In addition, in the DCE-US technique, the flow rate of the contrast agent cannot be controlled if the bolus injection is performed manually. To improve objectivity, we used an automatic injector, which greatly improved the accuracy of the raw data and the fitted curve. Therefore, we consider that k value is the most accurate of the four parameters studied.

Some studies have reported reduced PI in treatment groups.^[12,13] However, others have shown that PI was not a significantly effective parameter for predicting tumor growth.^[10,28] In the present study, there was no significant difference in PI between the responder and non-responder groups. Since slope is proportional to PI, there was also no significant difference in slope.

The lack of significant correlation between PI and tumor growth in the present study could be caused by a change in the depth of ROIs due to tumor growth and by saturation of signal intensity due to the high concentration of Sonazoid. In our study, the factors affecting PI were the movement of the rat liver lobes in the abdominal cavity and inconsistent ROI depth across the evaluation times.

The pathophysiologic mechanism for changes in tumor perfusion following chemotherapy is probably associated with the requirements associated with microvessel changes.^[25] In our study, the necrotic rate was significantly higher in the responder group than in the non-responder group. It is known that the response to chemotherapy causes cytotoxic tumor cell death resulting in reduced concentrations of tissue endothelial growth factor and therefore apoptosis of immature endothelial cells, with secondary vascular shutdown.^[29] Most cytotoxic agents impair the endothelial cells as well as the tumor cells.^[30] This mechanism acts to decrease tumor perfusion, thus leading to changes in the DCE-US parameters.

Assessment of MVD to predict tumor response has been reported in several studies, using perfusion CT,^[22] dual-energy CT,^[31] and DCE-MRI.^[32,33] A previous study has shown a correlation of DCE-US parameters with those of DCE-MRI and with immunohistochemistry.^[32] Therefore, DCE-US is a worthwhile modality for evaluating changes in tumor perfusion. Further comparative studies of DCE-US and these other modalities are needed to confirm the relative advantages of each technique.

There are some limitations to our study. First, as a potential limitation of the DCE-US technique in animal experimentation, there were fluctuations in intensity due to respiratory variation. Since those factors can introduce variability and fail to obtain the generalizability of results, operators are required to be adequately trained so that subsequent scans could be performed consistently in DCEUS qualification.^[34] When applying our method in a clinical examination, the evaluation of CEUS is generally affected by the patient’s body shape and background liver and tumor location. Second, in our study, only two-dimensional US images were obtained in the maximum plane; therefore, the tumor may not have been evaluated as a whole. Third, only single-tumor cases were evaluated in this study, but it could be difficult to assess multiple or fused lesions with US in clinical examination. Finally, pathological MVD in each group was not evaluated on day 3 because it is impossible to confirm both the changes in MVD and tumor size in the same individual. However, it has been demonstrated that cytotoxic chemotherapy induces changes in MVD at an early point.^[25] Further prospective and clinical investigations are mandatory to confirm the validity of k value in DCEUS for the early assessment of tumor response to systemic chemotherapy.

CONCLUSION

DCE-US could be a helpful method for evaluating the efficacy of the use of cytotoxic agents for the treatment of liver tumors. Among perfusion parameters, k value obtained from DCE-US appears to be particularly effective as a parameter for detecting early tumor response to chemotherapy in a rat colorectal liver metastasis model.