Changing paradigms in evaluating adrenal incidentalomas: Bayesian evaluation of [18F]Fluorodeoxyglucose positron emission tomography use, honed on adrenocortical carcinoma

Abstract

We present the diagnostic performance of [¹⁸F]Fluorodeoxyglucose positron emission tomography (FDG PET) for adrenal incidentalomas based on lesion size and unenhanced computed tomography (CT) density in Hounsfield units (HU), following current literature and guidelines. A 20 HU cutoff can be applied to differentiate potentially benign from malignant lesions, particularly in ruling in or out adrenocortical carcinoma. While FDG PET provides valuable metabolic information, its likelihood ratios for a positive (LR+) or negative (LR-) result do not exceed the robust diagnostic thresholds of > 10.0 or < 0.1, respectively. This suggests that positron emission tomography alone is insufficient for definitive characterization and should be integrated with CT or magnetic resonance imaging to leverage their complementary anatomical and functional imaging strengths for optimal diagnostic accuracy.

Key Words: Adrenal gland neoplasms; Diagnosis; Epidemiology; Positron emission tomography; Bayes theorem

Core Tip: [¹⁸F]Fluorodeoxyglucose positron emission tomography (FDG PET) diagnostic performance for adrenal incidentalomas (AdIn) was evaluated based on their size as measured on computed tomography (CT) and their unenhanced density in Hounsfield units (HU), using a 20 HU cutoff to assess adrenocortical carcinoma (ACC) risk. The likelihood ratios for positive/negative tests do not meet the usual thresholds, respectively, indicating strong but not definitive diagnostic accuracy. FDG PET should be combined with CT or magnetic resonance imaging for optimal assessment of AdIns.

INTRODUCTION

Adrenal incidentalomas (AdIns) are adrenal masses measuring ≥ 1 cm in diameter, discovered incidentally during imaging studies conducted for non-adrenal indications. The prevalence of AdIns ranges from 3% to 10% in cross-sectional imaging studies, with detection rates increasing with age[1,2].

While the majority of these incidentalomas are benign, non-secreting adenomas, distinguishing them from malignant lesions, particularly adrenocortical carcinoma (ACC), is crucial due to ACC's aggressive nature and poor prognosis.

ANATOMICAL IMAGING EVALUATION

Computed tomography (CT) remains the cornerstone of AdIns evaluation. On unenhanced CT, adenomas are typically lipid-rich and demonstrate low attenuation values [Hounsfield units (HU) ≤ 10]. This threshold has a specificity of approximately 98% for benign adenomas[3]; however, lipid-poor adenomas may remain indeterminate. For indeterminate lesions, contrast-enhanced CT with washout analysis can provide additional diagnostic information. Benign adenomas exhibit rapid contrast washout, with an absolute percentage washout > 60% or a relative washout > 40% being highly suggestive of a benign lesion. Size is another important consideration; lesions > 4 cm, especially those with irregular borders, necrosis, or calcifications, are more likely to be malignant and warrant further evaluation. A seminal relevant article from the Cleveland Clinic appeared in 2022: In this work, 2219 patients with AdIn were either followed at least for a year or underwent surgery[4]. This impressive series changed the paradigm of dealing with AdIns, since its authors concluded that the optimal cutoff in adrenal lesion density, which was associated with a higher probability of ACC, is at 20 HU, in lieu of - the until now - accepted value of 10 HU in unenhanced CT. Additionally, the authors found that the increasing probability of ACC vis-à-vis larger AdIn size is lower than previously thought[5]. The cutoff of 20 HU has been integrated in the most recent European Society of Endocrinology AdIns guidelines[6].

Thus, the 20 HU threshold on unenhanced CT is the currently recognized criterion for distinguishing benign lipid-rich adenomas from other adrenal lesions, including lipid-poor adenomas, metastases, and primary adrenal malignancies; nevertheless, the 10 HU threshold is also applied, defining a “grey” zone between 10-20 HU[6]. Additionally, its clinical applicability is influenced by variations in patient demographics, imaging technology, and scanning protocols, necessitating a nuanced approach to its interpretation. The diagnostic performance of the 20 HU threshold is contingent on the prevalence of different adrenal lesion types within specific populations and is specific for lipid-rich adenomas[3]. However, among oncological patients, where adrenal metastases are more common, specificity declines[7]. Notably, in malignancies such as lung cancer or melanoma, up to 30% of adrenal metastases may exhibit HU values below 20, leading to false-negative results[8,9]. Additionally, ethnic and geographic variations may affect the lipid content of adenomas, with studies suggesting higher HU values in certain populations due to genetic or environmental factors[10,11]. Rare adrenal tumors further complicate assessment; for example, myelolipomas, though benign, can contain regions exceeding 20 HU, while atypical pheochromocytomas may present with low attenuation, leading to diagnostic uncertainty. Technical differences across CT scanners and manufacturers can introduce variability in HU measurements. Variations in reconstruction algorithms, slice thickness, and iterative reconstruction techniques may alter HU values, leading to discrepancies in lesion characterization[12,13]. Partial volume averaging in thicker slices (> 5 mm) can artificially lower HU values in small lesions, while beam-hardening artifacts from metallic implants or increased scatter in obese patients further contribute to inconsistencies. The advent of dual-energy CT, offering improved lipid quantification, has the potential to refine the use of the 20 HU threshold by providing more accurate tissue characterization[14,15]. Differences in CT scanning protocols significantly affect the reliability of the 20 HU threshold. Importantly, this threshold is validated only for non-contrast CT; in contrast-enhanced phases, adrenal adenomas may demonstrate attenuation values well above 20 HU, necessitating additional imaging techniques such as delayed contrast washout studies[13]. On venous-phase imaging, adenomas can reach 40-60 HU. Emerging techniques such as spectral CT and virtual non-contrast imaging can help improve lesion differentiation without requiring traditional HU-based cutoffs[15].

Given these limitations, clinicians should incorporate additional imaging modalities when interpreting adrenal lesions. Lesions with HU < 10 are almost always benign lipid-rich adenomas and typically require no further workup unless malignancy risk factors are present. Lesions between 10-20 HU fall within an indeterminate zone where lipid-poor adenomas and malignancies overlap, warranting further characterization. While the 20 HU threshold remains a useful initial tool in adrenal lesion evaluation, its limitations necessitate a more comprehensive approach that integrates patient demographics, imaging technology, and additional imaging modalities. Future advancements in dual-energy CT and artificial intelligence-driven texture analysis hold promise for improving lesion characterization beyond traditional HU thresholds, potentially leading to more precise and individualized diagnostic strategies.

Magnetic resonance imaging (MRI) is another valuable tool, particularly chemical shift MRI, which is useful for differentiating lipid-rich adenomas from malignant lesions. A significant signal drop on out-of-phase imaging is characteristic of adenomas due to intracellular lipid content. However, lipid-poor adenomas remain indistinguishable from malignancies using MRI alone, and its superiority over contrast-enhanced CT remains uncertain.

FUNCTIONAL IMAGING: FLUORODEOXYGLUCOSE POSITRON EMISSION TOMOGRAPHY IN ADRENAL LESION CHARACTERIZATION

[¹⁸F]Fluorodeoxyglucose positron emission tomography ([¹⁸F]FDG-PET) has emerged as a critical tool in differentiating ACC from benign adrenal lesions due to its ability to assess metabolic activity. ACC typically exhibits intense fluorodeoxyglucose (FDG) uptake due to its high metabolic activity, with standardized uptake values (SUVmax) often exceeding 5. A commonly used semiquantitative parameter is the adrenal-to-liver SUV ratio, with a threshold of > 1.80 strongly suggesting malignancy[16] (other thresholds, ranging from 1.45 to 3.10 have been suggested by various research groups[16,17]). The sensitivity and specificity for FDG-PET/CT in identifying ACC have been reported at approximately 97% and 91%, respectively[18]. In the most recent European Society of Endocrinology AdIns guidelines[6] experts presented updated meta-analyses for the diagnostic characteristics of unenhanced CT and FDG PET (which is usually combined with CT) in evaluating AdIns and ruling in or out the presence of ACC. However, FDG-PET is not without limitations. False-positive findings can occur in pheochromocytomas and metabolically active adenomas, which may demonstrate increased FDG uptake, leading to potential diagnostic pitfalls. Conversely, false-negative results may arise in large necrotic ACCs or lesions with hemorrhagic degeneration, which may exhibit reduced or heterogeneous FDG uptake. While FDG-PET is the preferred functional imaging tool for malignancy detection, other radiotracers, such as ¹⁸F-F-DOPA and 68Ga-DOTATATE, may be considered in specific scenarios, such as for pheochromocytomas and paragangliomas. Of note, the guidelines on AdIns of the Canadian Urological Association (endorsed by the American Urological Association) leave out assessment with nuclear medicine methods[19], whereas the analogous guidelines of the French Association of Urology suggest the use of CT, MRI and functional/nuclear medicine imaging, primarily with FDG PET[20].

DIAGNOSTIC PERFORMANCE OF FDG-PET ACCORDING TO CT CHARACTERISTICS

In light of the above we believe that the time has come to re-evaluate FDG PET with Bayesian terms in the diagnostic workup of AdIns, as we have already done regarding ACC, AdIns and pheochromocytomas[21,22].

The Bayesian approach can be a valuable tool in the assessment of FDG-PET scans for AdIns, offering a probabilistic and evidence-based estimation of malignancy risk. This methodology integrates pretest probability with imaging-derived data, thereby refining risk stratification and guiding clinical decision-making.

Bayesian inference vis-à-vis AdIns operates by combining prior knowledge-derived from epidemiological data, clinical risk factors, and imaging characteristics, mainly from CT (or MRI)-with observed data from FDG-PET metrics to determine the posterior probability of malignancy. The pretest probability considers patient history, where known primary malignancy significantly increases the risk of ACC, as well as lesion size, where lesions greater than 4 cm raise suspicion. Additional imaging characteristics, such as the HU measurements on non-contrast CT (> 10 HU or > 20 HU suggesting lipid-poor adenoma or malignancy), further refine the baseline probability. The key FDG-PET parameters contributing to the likelihood ratio include the SUVmax, where higher values are more suggestive of malignancy.

Using Bayes’ theorem, the posterior/posttest probability of malignancy (for a positive test) is determined through the equation: Posttest probability = (likelihood ratio × pretest probability)/evidence.

The pretest probability represents the initial belief about the hypothesis before observing the evidence, the likelihood ratio quantifies how well the new evidence supports the hypothesis and the evidence normalizes the result to ensure probabilities sum to 1. Incorporating the pretest probability, sensitivity and specificity the equation is transformed to: Posttest probability = (pretest probability × sensitivity)/[pretest probability × sensitivity + (1 - pretest probability) × (1 - specificity)].

This approach allows for dynamic risk estimation, thereby improving diagnostic accuracy. In Table 1 we present the diagnostic characteristics of FDG PET according to AdIns CT size and unenhanced CT density in HU, per the recent literature; recent meta-analyses indicate lower sensitivity and specificity for FDG PET compared to older publications[6,21].

Table 1 Diagnostic characteristics of [¹⁸F]Fluorodeoxyglucose positron emission tomography according to adrenal incidentaloma size in computed tomography scans size and unenhanced computed tomography density in Hounsfield units.

CT		[18F]FDG PET
Pre-test probability of ACC (%)		Sensitivity (%)	Specificity (%)	LR+ (95%CI)	LR- (95%CI)	Post-test probability of ACC (%)
Incidentaloma diameter		90	73	3.33 (3.26-3.40)	0.14 (0.11-0.16)	Test (+)	Test (-)
< 4 cm	0.1					0.1	0.0
4-6 cm	2.4					8.0	0.0
> 6 cm	19.5					45.0	3.0
Incidentaloma unenhanced density						Test (+)	Test (-)
< 10 HU	0.0					0.0	0.0
10-20 HU	0.5					2.0	0.0
> 20 HU	6.3					18.0	1.0

[¹⁸F]FDG PET: [¹⁸F]Fluorodeoxyglucose positron emission tomography; CT: Computed tomography; HU: Hounsfield units; ACC: Adrenocortical carcinoma; LR+: Likelihood ratio of a positive test; LR-: Likelihood ratio of a negative test.

The likelihood ratios of positive tests (LR+) or negative tests (LR-) do not pass the thresholds of > 10.0 or < 0.1, respectively[23]. This indicates that FDG PET is very good, but not excellent, and apparently, has to be combined with CT or MRI to make the best of their characteristics. In the worst-case scenario, a subject with an AdIn larger than 6 cm has almost a one-in-five pre-test probability of harboring ACC, which, post-test, more than doubles to almost a one-in-two probability, with a positive FDG PET scan. In an analogous scenario, a subject with an AdIn having a CT density higher than 20 HU has a 6% pre-test probability of harboring ACC, which, post-test, trebles to 18%, with a positive FDG PET scan.

CLINICAL IMPLICATIONS

The integration of FDG-PET into the diagnostic algorithm of AdIns is particularly useful in several scenarios. First, it aids in the evaluation of indeterminate lesions on CT or MRI when washout characteristics or lipid content analysis are inconclusive. Second, it is valuable in cases with high clinical suspicion for ACC, such as in patients with large adrenal masses (> 4 cm) and irregular imaging features. Finally, FDG-PET/CT is instrumental in staging and metastatic workup, particularly in ACC, where metastases may be present at the initial diagnosis.

To enhance clinical decision-making, a structured diagnostic algorithm, integrating CT attenuation values with additional imaging modalities when needed, could be followed:

HU < 10 on unenhanced CT → Consistent with lipid-rich adenoma → No further workup unless high clinical suspicion for malignancy.

HU 10-20 on unenhanced CT → Indeterminate → Consider contrast washout assessment, FDG-PET or chemical-shift MRI.

HU > 20 on unenhanced CT → Potentially lipid-poor adenoma or malignant lesion → Proceed with:

Absolute washout < 60% or relative washout < 40% on delayed contrast CT → Possible malignancy → Consider FDG-PET, biopsy, or surgical evaluation.

Absolute washout ≥ 60% or relative washout ≥ 40% → Likely benign → Clinical follow-up.

Oncological patients/patients with indeterminate washout/MRI findings → FDG-PET.

Persistent diagnostic uncertainty → Multidisciplinary discussion or biopsy for metastases if clinically appropriate.

This structured approach aims to minimize unnecessary interventions while ensuring timely identification of malignant adrenal lesions.

CONCLUSION

While our study is based on a synthesis of existing literature and retrospective analyses, it provides a comprehensive evaluation of the 10 or 20 HU thresholds and FDG-PET positivity/negativity across diverse adrenal findings. By integrating findings from multiple studies, we aim to offer a critical appraisal of current evidence and highlight the need for future prospective research to further refine adrenal lesion characterization. The comprehensive evaluation of AdIns requires a combination of anatomical and functional imaging. While CT and MRI remain the first-line modalities, FDG-PET has demonstrated substantial utility in differentiating ACC from benign adrenal lesions, especially in cases where traditional imaging is inconclusive. Further refinement of PET-based diagnostic thresholds and prospective validation studies will enhance its clinical applicability in adrenal oncology. By integrating pretest probability with imaging-derived data, the Bayesian approach significantly enhances the diagnostic utility of FDG-PET in evaluating AdIns. This refined risk stratification minimizes unnecessary interventions while improving the identification of high-risk lesions. Future research should focus on refining Bayesian models through machine learning algorithms to further enhance individualized malignancy prediction, optimizing clinical workflows and patient outcomes. With the growing worldwide number of PET/CT scanners[24], we believe that further experience on AdIns will be gathered, particularly when combining data from different centers.