Tissue characterization is vital to planning subsequent therapy after diagnosis of an adrenal mass. Over the last decade, imaging algorithms have been developed that have substantially reduced the need for tissue diagnosis by percutaneous biopsy. Biopsy should be avoided, not so much because it is an invasive technique with a small risk of pneumothorax, ³ but because it is time-consuming, expensive, and perhaps most importantly, has a reported diagnostic accuracy of only 80 to 90%. ⁸

Several morphological features of the adrenal gland may permit lesion characterization, including lesion size, shape, and heterogeneity. ⁹ Lesions greater than 5 cm are almost always malignant (figure 1) (except for the rare myelolipoma). ² Such large lesions are usually adrenal carcinomas. Lesion irregularity also may indicate a malignant lesion, but a smooth lesion can represent either benign or malignant disease. Most incidentally-detected adrenal masses are in fact small and smooth in shape and, therefore, lesion shape often is not very helpful.

Occasionally, larger lesions may demonstrate inhomogeneity, particularly if intravenous contrast has been administered. Benign lesions do not demonstrate areas of necrosis. However, as most incidentally-detected lesions are small and homogeneous (whether benign or malignant), this is not a reliable test (figure 2). The easiest and least costly test is using a prior study for comparison. If an older CT or magnetic resonance scan (MRI) is available and the lesion has not changed in size in 6 months, then this most likely represents a benign lesion. Any increase in size over this time is highly suspicious for malignancy. If there is a marked increase in size and no features of hemorrhage, then the lesion is definitely malignant and biopsy is not required.

As the above morphologic features are often nonspecific, a more accurate test is required to accurately differentiate benign from malignant disease. Fortunately, both CT and MRI techniques have been developed that enable lesion characterization without the need for percutaneous biopsy in the majority of cases. This is because the adrenal cortex and many benign adrenocortical tumors contain intracytoplasmic fat (mainly of cholesterol, fatty acids, and neutral fat) while malignant lesions do not. ¹⁰ Any technique that can detect intracytoplasmic fat will enable differentiation of benign from malignant lesions. However, for this test to be clinically useful, its specificity for the detection of an adenoma needs to approach 100%. As the test specificity decreases, more malignant lesions may be called benign. This unfortunate situation might mean that the referring physician would attempt unnecessary curative surgery for the primary lesion, in the assumption that the detected adrenal mass is a benign lesion.

The converse is not as important, however. If a potential adrenal adenoma is found to be indeterminate, most institutions will require tissue diagnosis before palliative surgery is initiated. Put another way, the referring physician will require a tissue diagnosis before denying the patient potential curative surgery.

Lipid-sensitive imaging techniques are not 100% specific and sensitive because only 60 to 70% of benign adenomas contain enough intracytoplasmic fat to permit separation of benign from malignant disease. Therefore, some adrenal adenomas might potentially be characterized as metastases. Rather than characterizing all lipid-poor lesions as malignant, lipid-poor adenomas and malignant lesions are deemed indeterminate. These indeterminate lesions will, however, require further characterization, either by an alternative imaging method or by percutaneous biopsy. Therefore, lipid-sensitive tests that can detect benign disease can be made close to 100% specific for the detection of an adenoma, with some reduction in sensitivity.

However, in order for a lipid-sensi-tive test to be clinically useful, it should not be too stringent. The test has relatively little use if in order to achieve close to 100% specificity, sensitivity is reduced to levels below 50%. In this scenario, too many lesions remain indeterminate and additional imaging or biopsy is required. A trade-off is, therefore, required between sensitivity and specificity. In order for sensitivity to be relatively high (>70%) the specificity needs to be lowered slightly. Certainly specificity needs to be close to 100%, and an acceptable level is probably greater than 95%. Although this results in a clinically practical test, there will inevitably be the occasional malignant lesion that is "down-staged" to an adenoma. However, this unfortunate situation can often be avoided, as there may be other clues that permit a lesion to be characterized (recent increase in size, irregularity, or inhomogeneity). Clinical experience has demonstrated over the last decade that some slight lowering of specificity is still clinically acceptable.

Most incidentally detected adrenal lesions are identified on CT. Ideally, then, lesion characterization should be performed by CT in order to avoid the need for further imaging or biopsy. Lipid sensitive imaging is well suited to CT. Not only can macroscopic fat be evaluated by CT (i.e. myelolipoma) but sufficient concentrations of intracytoplasmic fat also can be detected. As CT density values are related to tissue attenuation coefficients, fat-containing structures will lower the Hounsfield unit (HU) measurements more than would other soft-tissue structures, including water-containing structures, if fat is present in a sufficient concentration. By placing a region of interest (ROI) on the adrenal lesion, density measurements can determine if the adrenal lesion contains enough intracytoplasmic fat to be diagnosed as an adenoma.

Korobkin et al were able to demonstrate an inverse linear relationship between the intracytoplasmic fat content of an adrenal adenoma and the CT density value. ¹⁰ Adenomatous lesions usually contain significant intracytoplasmic fat and CT density values are low (usually -10 to +10 HU) (figure 3). Pure fat-containing lesions (i.e. myelolipomatous lesions) will have density values of ­80 to ­100 HU (figure 4). Conversely, non-adenomatous lesions typically have higher CT density values, as their cytoplasm is relatively lipid-poor (figure 5).

Using this principle, chemical shift MR techniques also have revealed similar accuracies to unenhanced CT for differentiating benign from malignant adrenal masses. ^11-14 However, CT continues to remain the investigational modality of choice for evaluating adrenal masses due to its widespread availabil-ity, its high speed, and its low cost.

Many studies have confirmed the ability of unenhanced CT to consistently characterize the incidentally detected adrenal mass. ^12,15-24 Most of these have attempted to determine a threshold density value which would permit separation of benign from malignant lesions. ^18-24 However, a wide range of sensitivities has been reported, and no study has yet reported that any threshold will result in both 100% sensitivity and specificity. A recent study ²⁵ pooled the density values from 10 previous reports using noncontrast CT in order to optimize the correct density threshold value by which benign from malignant adrenal lesions can be characterized. On the basis of this analysis, it was suggested that if the goal is to avoid false positive diagnoses at all costs, one should use a threshold value of 2 HU, as this results in 100% specificity for the test. Any lesions with CT attenuation coefficients below this level can safely be considered benign. However, if one is willing to accept a small false negative rate in order to achieve improvements in sensitivity, then perhaps a threshold of up to 10 HU could be used. The precise cut-off used must ultimately depend on the prevalence of malignant disease in the patient population, and the goals of the imaging procedure. Conversely, lesions with attenuation values above 25 HU are rarely benign, and these lesions should be assumed to be malignant. Of course, further confirmation will be required, as the occasional adenoma will demonstrate HU values above this.

Although noncontrast CT is a highly effective method for differentiating benign from malignant adrenal disease, most incidental adrenal lesions are initially detected by contrast-enhanced CT, often in patients being investigated for an extra-adrenal malignancy. Rather than performing an additional noncontrast CT at a later stage, several authors have attempted to characterize adrenal masses using contrast-enhanced CT, again by using adrenal density values to detect intracytoplasmic lipids. As expected, adrenal glands enhance, complicating the lipid sensitive tests used to characterize their lesions. It has been shown that the initial dynamic contrast-enhanced CT is not useful for lesion characterization because both benign and malignant lesions enhance to similar density values. ^23,24,26-28 A threshold value cannot, therefore, be chosen by which benign fat-containing lesions can be separated.

However, several authors have attempted to characterize these lesions using a delayed contrast-enhanced CT. ^23,24,26 These researchers suggest that by waiting a period of time after dynamic contrast-enhanced CT, there may be enough "washout" of contrast from the adrenal gland after which density values can again be used to characterize the incidentally detected adrenal mass. Followed to its extreme, a delay of more than 24 hours would equate to a noncontrast CT. However, this is neither practical nor desirable, and as a result, these authors have attempted much shorter time delays.

It has been shown that characterization can be performed as early as 10 to 15 minutes after the dynamic scan. Researchers in the above studies have demonstrated that benign, fat-containing lesions have significantly lower density values than non­fat-containing lesions on these delayed scans (figure 6). The precise cut-off threshold has yet to be fully established, though it is probably approximately 30 HU. ^23,24,26 Lesions below this value can be confidently assumed to be benign, and lesions above this value remain indeterminate, requiring further imaging with MRI, or biopsy.

More recent reports have attempted
to characterize adrenal lesions using a different paradigm than simple density values at a single time point after contrast-enhanced CT. As it has been demonstrated that benign adrenal lesions "washout" significantly faster than malignant adrenal lesions after injection of IV contrast, ^28,29 more accurate characterization might be performed by comparing density values on the delayed CT scan to the initial dynamic contrast study. A ratio can be derived, from which a threshold can be used in a similar fashion to the threshold values used to separate benign from malignant lesions on noncontrast and simple delayed contrast-enhanced CT.

In one study, Korobkin et al demonstrated that the density value of adenomas at 10 minutes after contrast administration was only 28% of its peak enhancement on the initial dynamic scan. ²⁷ The density values of malignant lesions, on the other hand, had only decreased to 93% of their initial value on the dynamic scan. Using this data, Korobkin et al, ²⁷ and Szolar et al, ²⁸ from a similar study, were able to derive relative percent washout threshold ratios (relative because the unenhanced value is unknown) to separate benign from malignant lesions. The formula for deriving the relative threshold ratio is: delayed enhanced value/initial enhanced value * 100%. ²⁷ Although the values from the two studies varied slightly, it has been determined that if the density value on the 10-minute delayed study drops by more than 50% compared to the initial dynamic scan, then the test is close to 100% specific and sensitive. Interestingly, Korobkin and coworkers went a step further and demonstrated that lesions found to be indeterminate by noncontrast CT (adenomas with density values greater than 10 HU) could indeed be characterized using the relative delayed washout percent value without the need for further imaging or biopsy. If further studies can corroborate this finding, then MRI and/or percutaneous biopsy might be avoided altogether.

MRI techniques are also able to detect intracytoplasmic fat within adrenal adenomas and are useful for differentiating benign from malignant lesions. However, there is a decreasing requirement for MRI, as CT can characterize an increasing number of adrenal lesions. The technique by which MRI can differentiate between lipid and non-lipid containing structures, chemical shift MRI, still has an important role to play, as some institutions prefer MRI to delayed contrast-enhanced CT techniques.

Normally, fat protons precess faster (and therefore dephase faster) than water protons. Due to these differing precessional frequencies, the two will be out-of-phase after a variable time period (2.3 msec at 1.5T). Using normal spin-echo T1-weighted imaging, the emitted signals from dephasing fat and water protons are summated by using a refocusing pulse, and are therefore imaged as structures with intermediate signal. Gradient-echo (GRE) pulse sequences can now be performed such that fat and water protons can be separated within an imaged voxel. In GRE imaging, proton and fat refocusing pulses do not occur, so water and fat protons will continue to cycle in- and out-of-phase with respect to one another. Therefore, at 1.5T, fat and water protons are out-of-phase after 2.3 msec and in-phase again at 4.6 msec, out-of-phase at 6.9 msec, and so on. By choosing echo times at 2.3 and 4.6 msec, in-phase and out-of-phase images can be obtained. On the in-phase images, fat and water signals will summate and produce adrenals of intermediate signal, but on out-of-phase images, the signals from fat and water will cancel each other out and appear dark (figure 7).

Various modifications of the chemical shift have been suggested and have been shown to be highly effective for the characterization of adrenal masses (figure 8). ^11-14 Some authors have suggested measuring the signal intensities on in-phase and out-of-phase images and calculating adrenal/liver of adrenal/spleen ratios in order to differentiate adrenal lesions. However, it has been demonstrated that qualitative evaluation of the adrenal mass signal in-phase and out-of-phase is just as accurate as quantitative measurements. ²⁹ Therefore, any visualized signal loss on out-of-phase images indicates a benign lesion. Using these techniques, some indeterminate adrenal lesions on noncontrast CT can be characterized, although some adrenal adenomas contain insufficient concentrations of fat to be characterized in this manner. Percutaneous biopsy is, therefore, still required in some patients.

Several researchers have now suggested an imaging algorithm by which adrenal masses can be characterized. ^9,20,30,31 Because CT detects most lesions, any algorithm should, ideally, be CT-based. The algorithm depends on whether or not the patient has a known extra-adrenal malignancy. If no extra-adrenal malignancy is present, lesions with HU values below 10 are benign. However, even lesions with HU values above this usually can safely be assumed to be benign, unless demonstrating unusual morphologic features (such as large size, irregularity, inhomogeneity). If some uncertainty persists, a follow-up CT in 6 months is recommended to evaluate for any change in the size of the lesion.

If the patient has a known primary malignancy, the lesion must be characterized. Initially the study must be compared to any prior CT images. Any increase in lesion size strongly suggests a metastasis. If a noncontrast CT has been performed, then an adenoma can be diagnosed for lesions with HU values of 10 HU or less. For values above 10 HU, a chemical shift MRI should be performed to evaluate for signal drop-off on the out-of-phase images. If the lesion is still indeterminate, a percutaneous biopsy should be performed (although recent studies have suggested that these lesions can be characterized by washout characteristics when using 10-minute delayed contrast-enhanced CT).

If a contrast-enhanced CT has been performed in a patient without a history of malignancy, delayed CT measurements should be made. If on the 10- to 15-minute scan the HU value is less than 30 HU, then the lesion is benign. However, even if above 30 HU, the lesion can similarly be assumed to be benign, as discussed above. For patients with a known malignancy, demonstration of a lesion with density values of 30 HU or less confirms a benign lesion on 10- to 15-minute delayed scans. If the lesion is still indeterminate (having a HU of greater than 30), then a relative percent washout value should be performed. If the lesion has "washed out" by more than 50%, then this lesion is benign. Remaining indeterminate lesions should be imaged by MRI and if still indeterminate, a biopsy should be performed. By using such an imaging algorithm, very few lesions will require percutaneous biopsy. ²⁰ AR