Journal

Hong Kong J Radiol 2026 Mar;29(1):e4-14 | Epub 9 March 2026

 

 

Department of Clinical Oncology, Tuen Mun Hospital, Hong Kong SAR, China

 

  Full Article in PDF

Nasopharyngeal carcinoma (NPC) is an epithelial carcinoma originating from the nasopharyngeal mucosa. This malignancy is most prevalent in Asia, accounting for over 80% of global incident cases in 2022.[1] In endemic regions, NPC incidence peaks in the 45-59 years age-group and declines thereafter.[2] Data from the Hong Kong Cancer Registry indicate that in 2023, approximately 4.9% of new NPC cases occurred in patients aged 80 years or above.[3]

 

Standard treatment for NPC involves high-dose radical radiotherapy (RT) of 66 to 70 Gy, often combined with concurrent, induction, and/or adjuvant chemotherapy for locally advanced disease.[4] [5] However, these treatment guidelines are largely based on clinical studies that have underrepresented or excluded older adult populations. For instance, in a meta-analysis of chemotherapy in NPC, only 13% of the cohort was aged 60 years or above.[6]

 

Older adults with NPC have worse survival outcomes compared to their younger counterparts.[5] Previous studies have reported 5-year overall survival (OS) rates ranging from 44% to 60% among patients aged 70 years or above with NPC,[7] [8] [9] whereas those aged 80 years or above exhibit a considerably lower survival rate of approximately 30%.[10] Treating older adults with NPC presents particular challenges due to increased co-morbidities, nutritional issues, organ dysfunction, and greater susceptibility to treatment-related toxicities.[11] Despite these clinical challenges, studies specifically addressing treatment outcomes and strategies in older adults with NPC remain limited. Furthermore, inconsistencies exist regarding the definition of ‘older adults’ or ‘elderly’ across published studies, with age thresholds typically ranging from 65 to 70 years.[7] [8] [9] [10] [12] [13] Notably, outcomes for the oldest patients with NPC, specifically those aged 80 years or above, are rarely reported. These much older patients may represent a distinct subgroup, even within the broader geriatric population. Huang et al[10] reported that patients aged 80 years or above with NPC had significantly worse survival than those aged 65 to 69 years. This study aimed to investigate treatment patterns and survival outcomes in older adults aged 80 years or above with NPC in Hong Kong.

 

We conducted a retrospective review of the medical records of patients with NPC who received care at Tuen Mun Hospital between 1 January 2009 and 31 December 2023. Patients aged 80 years or above at diagnosis with histologically confirmed NPC were included. Those with distant metastasis at initial diagnosis were excluded. Data on demographics, disease status, co-morbidities, and treatment outcomes were retrieved from electronic patient records and analysed. Patients were categorised into those who received radical RT to the nasopharynx (Cohort A) and those who did not (Cohort B).

 

Patients underwent clinical evaluation, including history taking and physical examination. Local and regional staging was performed using magnetic resonance imaging of the nasopharynx and neck and/or computed tomography. Between 2009 and 2017, positron emission tomography–computed tomography (PET-CT) was selectively performed in patients with symptoms, laboratory abnormalities, or chest radiograph findings suggestive of distant metastasis. From 2018 onwards, PET-CT has been routinely performed for all patients with tumour (T) stage T4, nodal (N) stage N3, or T3N2 disease, as well as those with clinical suspicion of metastatic disease, in accordance with Hospital Authority (HA) standard indications.

 

NPC staging was performed according to the 8th edition of the American Joint Committee on Cancer (AJCC) staging manual.[14] Patients diagnosed prior to the introduction of the AJCC 8th edition were retrospectively re-staged. Patient performance status was assessed using the Karnofsky Performance Status (KPS) Scale.[15] Co-morbidities and overall health status were retrospectively evaluated using the Adult Comorbidity Evaluation–27 (ACE-27),[16] the Charlson Comorbidity Index (CCI),[17] and the modified Frailty Index–11 (mFI-11).[18]

 

All patients who received radical RT underwent intensity-modulated radiotherapy (IMRT). Patients were immobilised in the supine position using a thermoplastic cast applied to the head and shoulders. A non-contrast simulation computed tomography scan was acquired and fused with the diagnostic magnetic resonance imaging scan. Target volumes were contoured according to international guidelines.[19] [20] The gross tumour volume encompassed the primary tumour and enlarged lymph nodes. Clinical target volumes (CTVs) were defined as high-risk, intermediate-risk, and low-risk CTVs. The high-risk CTV included the gross tumour volume plus a 5-mm margin and the whole nasopharynx. The intermediate-risk CTV included the high-risk CTV plus a 5-mm margin and was expanded to cover sites at risk of microscopic extension, as well as the involved nodal levels. The low-risk CTV included uninvolved but potentially at-risk nodal levels. Prescribed doses to the high-, intermediate-, and low-risk CTVs were 70 Gy, 60 Gy, and 54 Gy, respectively, delivered in 33 daily fractions using the simultaneous integrated boost technique. A 3-mm margin from CTV to planning target volume was added to account for setup uncertainty. The planning target volume was subsequently cropped 3 mm from the external body contour, and midline avoidance structures were created to minimise skin and mucosal toxicities.

 

Patients undergoing radical RT were monitored weekly during treatment. RT-related toxicities were prospectively recorded and graded according to the National Cancer Institute Common Terminology Criteria for Adverse Events, version 5.0.[21] Treatment response evaluations were conducted 8 to 12 weeks after completion of RT and included physical examination and nasopharyngoscopy. For patients treated after 2021, routine magnetic resonance imaging of the nasopharynx and neck was also performed in addition to physical examination and nasopharyngoscopy. Patients were subsequently followed up at regular 3- to 6-month intervals by oncologists and otolaryngologists. Each visit included a clinical examination and nasopharyngoscopy. Further investigations (e.g., imaging and blood tests) were performed when recurrence was suspected.

 

Causes of death were determined from electronic medical records and classified into four categories: (1) NPC-related death, defined as death resulting from the primary NPC or its metastases; (2) treatment-related death, defined as death due to complications arising from NPC treatment; (3) non-NPC death, defined as death from causes unrelated to the cancer or its treatment; and (4) unknown, defined as death for which a definitive cause could not be established based on the available clinical information. Classification as NPC-related death required the terminal event to be attributable to metastatic disease or to a documented complication of symptomatic or progressive local disease. When competing causes were present, the primary cause was determined based on clinical documentation, imaging findings, and its temporal relationship to treatment. For example, aspiration pneumonia occurring with documented dysphagia secondary to progressive local NPC was classified as an NPC-related death, whereas aspiration pneumonia in the absence of documented treatment-related dysphagia or residual tumour was classified as a non-NPC death.

 

OS was defined as the interval from the date of histological diagnosis to the date of death. Progression-free survival was defined as the interval from histological diagnosis to the date of disease progression (including local, regional, or distant progression) or death. Cancer-specific survival (CSS) was defined as the interval from histological diagnosis to the date of NPC-related death. Survival rates were estimated using the Kaplan–Meier method. Univariable and multivariable Cox proportional hazards regression models were used to identify factors associated with survival. Variables with p < 0.05 in univariable analysis and those deemed clinically relevant were considered for multivariable modelling. To reduce multicollinearity, closely related clinical variables were not included simultaneously in the multivariable model, such as individual TNM (tumour-nodal-metastatic) components and overall stage or measures of performance status and frailty. Hazard ratios (HRs) with 95% confidence intervals (95% CIs) were reported. The Mann–Whitney U test was used to compare distributions of ordinal variables between patient cohorts. For categorical variables, the Chi squared test or Fisher’s exact test was applied, as appropriate. All statistical tests were two-sided, with a significance threshold of p < 0.05. Statistical analyses were performed using SPSS (Windows version 26.0; IBM Corp, Armonk [NY], United States).

 

In total, 42 patients were included. Patient characteristics are summarised in Table 1. The median age was 83 years (range, 80-94) and 29 patients (69.0%) were men. Most patients presented with stage III disease (33.3%), followed by stage II (26.2%), stage IVa (19.0%), and stage I (11.9%). A higher proportion of patients in Cohort A underwent PET-CT for distant metastasis screening compared with Cohort B (29.6% vs. 6.7%). Staging information was unavailable for four patients (9.5%), all of whom were in Cohort B.

 

Table 1. Baseline patient and disease characteristics.

 

Overall, 27 patients (64.3%) received radical RT to the nasopharynx (Cohort A), while 15 patients (35.7%) did not (Cohort B) [Table 1]. Reasons for not undergoing radical RT included patient refusal (n = 9), concomitant malignancy (n = 1), and medical unfitness for radical treatment (n = 5). Of the 15 patients in Cohort B, two (13.3%) received palliative RT. Chemotherapy was not administered to any patients in either cohort.

 

Cohort A had significantly more patients with a KPS score ≥70% compared with Cohort B. No significant differences were observed in ACE-27 scores or CCI scores. Although a higher proportion of patients in Cohort B had a mFI-11 score ≥0.27 (categorised as frail) compared with Cohort A, this difference was not statistically significant (Table 1).

 

At the time of analysis, eight patients (19.0%) were alive. The median follow-up duration was 20.3 months (range, 1.5-138) for the entire cohort, and 28.2 months for those who were alive. The median OS was 22.8 months (95% CI = 14.6-30.9).

 

Among patients who received radical RT (Cohort A), the median OS was 41.3 months (95% CI = 27.7-55.0). The median CSS was not reached. The median progression-free survival was 39.6 months (95% CI = 22.4-56.7). The 5-year OS and CSS rates were 38.1% and 74.2%, respectively (Figure 1).

 

Figure 1. (a) Overall survival and (b) cancer-specific survival in Cohort A.

 

Among patients who did not receive radical RT (Cohort B), the median OS was 12.8 months (95% CI = 10.9-14.7) and the median CSS was 14.4 months (95% CI = 10.9-17.9). No patient in Cohort B survived to 5 years (Figure 2).

 

Figure 2. (a) Overall survival and (b) cancer-specific survival in Cohort B.

 

Univariable analysis identified several factors significantly associated with worse OS, including absence of radical RT (no vs. yes; HR = 5.03, p < 0.001), male sex (male vs. female; HR = 2.55, p = 0.031), advanced nodal stage (N2-3 vs. N0-N1; HR = 2.70, p = 0.017), advanced overall AJCC stage (stage III-IV vs. stage I-II; HR = 2.99, p = 0.005), poor KPS score (<70% vs. ≥70%; HR = 3.29, p = 0.003), and frailty based on the mFI-11 (mFI-11 score ≥0.27 vs. <0.27; HR = 4.22, p = 0.010). On multivariable analysis, no receipt of radical RT (HR = 13.33; p = 0.006) and male sex (HR = 3.22; p = 0.033) were independently associated with worse OS (Table 2).

 

Table 2. Univariable and multivariable analyses of prognostic factors for overall survival.

 

Among the 34 patients who died, the most common cause of death was NPC-related death (n = 15, 44.1%), followed by non-NPC death (n = 13, 38.2%). Treatment-related mortality occurred in one patient (2.9% of deaths), and the cause of death was unknown in five patients (14.7%). The causes of death among patients who underwent radical RT (Cohort A) and those who did not (Cohort B) are summarised in Table 3. The two cohorts demonstrated distinct cause-of-death profiles. In Cohort A, the most common cause of death was non-NPC death (n = 11, 57.9%), followed by NPC-related death (n = 5, 26.3%), unknown causes (n = 2, 10.5%), and treatment-related death (n = 1, 5.3%). Among patients in Cohort A who died of non-NPC causes, the median interval from the last day of RT to death was 36.9 months (interquartile range, 16.1-71.0). In Cohort B, the majority of patients died of NPC-related causes (n = 10, 66.7%); two patients (13.3%) died of non-NPC causes and three patients (20%) died of unknown causes. Detailed descriptions of the circumstances of death for individual cases are provided in the online supplementary Table.

 

Table 3. Causes of death by treatment cohort.

 

Among the 27 patients in Cohort A who underwent radical RT, the majority (96.3%) completed the planned course of treatment. Local treatment response to RT was documented in 22 patients; of these, 95.5% achieved a complete response. One patient had persistent disease in the nasopharynx and achieved successful salvage with brachytherapy. No local or regional relapse was observed. Five patients (18.5%) developed distant recurrence, with a median time to onset of distant metastasis of 17.6 months (range, 8.3-34.0). None of these patients received further systemic anticancer therapy for metastatic disease.

 

Table 4 summarises the acute toxicities observed in Cohort A. Grade ≥3 acute RT toxicities, defined as those occurring during RT or within 3 months after RT, were observed in six of 27 patients (22.2%). The most frequently reported acute toxicities were mucositis (all grades, 96.3%; grade ≥3, 14.8%) and radiation dermatitis (all grades, 77.8%; grade ≥3, 3.7%). Seven patients (25.9%) required unplanned hospital admission during treatment: four for grade 3 mucositis, one for grade 3 dermatitis, one for feeding tube insertion to support nutrition in the absence of clinically significant mucositis, and one for a chest infection during the sixth RT fraction (this patient subsequently died). The fatal chest infection resulted in a treatment-related mortality rate of 3.7%. Two patients (7.4%) died within 90 days of completing RT.

 

Table 4. Acute treatment-related toxicities in Cohort A (n = 27).

 

Grade ≥3 late RT toxicities (defined as those occurring more than 3 months after RT) were observed in 14.8% of patients, the majority of which involved severe hearing loss. One patient (3.7%) required long-term feeding tube support due to dysphagia.

 

In this retrospective study of patients aged 80 years or above with NPC, radical RT using IMRT resulted in a median OS of 41.3 months and a 5-year OS rate of 38.1%, with manageable toxicity. To our knowledge, this is the first study to specifically evaluate treatment outcomes and toxicities in this group of patients, thereby addressing a critical knowledge gap.

 

The treatment of NPC in older adults is challenging and frequently overlooked, as this population is often excluded from or underrepresented in clinical trials. Older adults represent a heterogeneous group characterised by a wide range of co-morbidities and varying degrees of frailty. Management of NPC in this group is often complex, and survival outcomes are generally worse compared with those of their younger counterparts.

 

Yang et al[8] reported outcomes in patients aged 70 years or above with NPC, most of whom received RT combined with chemotherapy, achieving a 5-year OS rate of 59.5%. Notably, only 65.3% of patients in that cohort received IMRT, and most were younger than 75 years.[8] Jin et al[7] examined a similar cohort of patients aged 70 years or above with NPC who were treated exclusively with IMRT and reported a 5-year OS rate of 54%; however, chemotherapy was administered to 42.8% of patients, and the maximum age in that cohort was 73 years. Patients aged 80 years or above represent an especially challenging subgroup, even within the broader geriatric population. In a National Cancer Database analysis by Huang et al,[10] patients aged 80 years or above with NPC who received radical RT had a 5-year OS rate of 31.3%. Toxicity outcomes were not reported in that study.

 

Due to prevalent co-morbidities and reduced bone marrow reserve, older patients with NPC often have limited tolerance for chemotherapy, whether administered as induction therapy or concurrently with RT. The benefit of chemotherapy in this population remains a subject of debate. While some retrospective studies have reported improved outcomes with the addition of chemotherapy to RT in older adults,[12] [22] [23] others have shown no clear survival advantage.[7] [24] [25] In clinical practice, chemotherapy is seldom administered to patients aged 80 years or above.[8] Indeed, in our cohort, no patient in this age-group received chemotherapy.

 

High-dose RT to the head and neck region can be potentially morbid, and treatment tolerance is a significant concern, particularly among older adults. A study by Sze et al[9] reported significantly higher rates of acute grade 3 toxicities, RT incompletion, and 90-day mortality in patients aged 70 years or above with NPC compared with younger patients. As a result, clinicians may be hesitant to offer radical RT to patients aged 80 years or above.

 

Our findings demonstrated that radical RT is associated with meaningful survival outcomes in patients aged 80 years or above. Among those who received radical RT, a median OS exceeding 3 years and a 5-year OS rate of 38.1% are encouraging, suggesting that radical RT can provide reasonable survival even for octogenarians.

 

Our study also showed that patients who did not receive radical RT had poorer outcomes, with a median OS of only 12.8 months. However, direct survival comparisons between these two cohorts should be interpreted with caution due to important baseline differences. Patients in Cohort B had significantly worse performance status, with a greater proportion exhibiting a KPS score below 70 compared with Cohort A. Although no significant differences were observed between cohorts in terms of co-morbidity indices, inherent disparities undoubtedly existed. These differences may introduce confounding bias, whereby the observed survival advantage of radical RT may be partially attributable to baseline patient characteristics. Despite these limitations, the considerable difference in outcomes suggests a potential benefit of radical RT in appropriately selected older adults.

 

Perhaps more importantly, the cause-of-death analysis offers additional insight into the potential benefit of radical RT. Among patients who received radical RT, most deaths were due to medical conditions unrelated to NPC or its treatment, whereas in the non-radical RT group, the majority of deaths were attributable to NPC progression.

 

These findings may assist clinicians in discussions with patients and caregivers, facilitating personalised management strategies. It is important for clinicians to recognise the potential benefits of radical RT in appropriately selected patients, ensuring that advanced age alone does not preclude access to potentially curative treatment.

 

IMRT has become the standard of care for NPC, providing optimal tumour coverage while sparing critical organs at risk.[26] It is associated with improved tumour control and a reduction in both acute and late toxicities.[27] [28] In our study, however, grade ≥3 acute toxicities remained common (22.2%) among patients undergoing radical RT with IMRT. It is important to recognise that older adults are at increased risk of developing severe treatment-related toxicities; all toxicities should be identified promptly and managed proactively. In particular, RT-induced mucositis and dysphagia can lead to life-threatening infectious complications, as demonstrated by the single grade 5 toxicity observed in our cohort. Intensive clinical monitoring throughout treatment—combined with appropriate supportive medications and multidisciplinary collaboration involving nurses, dietitians, and speech therapists—is essential. Vigilance in nutritional management is particularly important, as older adults may already be at high risk of sarcopenia and have limited physiological reserves.[29] Clinicians should maintain a low threshold for feeding tube insertion during RT, and a prophylactic approach to nutritional support may be considered.

 

Although the incidence of grade ≥3 acute toxicities was relatively high, it was not prohibitive. In our study, the rates of grade ≥3 dermatitis and mucositis were 3.7% and 14.8%, respectively, both of which appear lower than previously reported figures of 21.6% to 22.3% for grade ≥3 dermatitis and 18.9% to 68% for grade ≥3 mucositis.[9] [25] This difference is likely attributable to our institutional protocol, which routinely includes a 3-mm skin clip and the creation of midline structure avoidance volumes. In the present study, the treatment-related mortality rate was 3.7% and the 90-day mortality rate was 7.4%, a figure comparable to the 7.8% reported by Sze et al[9] in patients aged above 70 years.

 

Late grade ≥3 RT toxicities were also infrequent in our study; only one patient remained dependent on a feeding tube. This observation may be partly explained by the relatively short follow-up period and limited survival duration, which may have precluded the full manifestation of late toxicities. Another contributing factor is that all patients received IMRT, which delivers a more conformal dose distribution to the target volume while better sparing adjacent normal tissues.[30]

 

Although this study focuses on patients aged 80 years or above, it is essential for clinicians to recognise that chronological age alone should not serve as the sole criterion for risk stratification. Co-morbidity and frailty assessments provide critical information to guide the management of older patients with NPC. Comprehensive geriatric assessment, considered the gold standard for evaluating older adults, is recommended by both the International Society of Geriatric Oncology[31] and the American Society of Clinical Oncology[32] to support treatment decision making. However, comprehensive geriatric assessment is not widely implemented due to its time-consuming nature. Several tools are available for co-morbidity assessment, including the CCI,[17] the ACE-27,[16] and the mFI-11.[18] Notably, both ACE-27 and CCI have been associated with survival outcomes. For example, Huang et al[10] identified CCI score ≥2 was an independent prognostic factor for mortality, while higher ACE-27 scores have been associated with poorer survival outcomes.[7] [9] [33] In our study, there was a trend towards worse survival outcomes in patients with higher CCI, ACE-27, and mFI-11 scores; however, none of these associations reached statistical significance in multivariable analysis, likely due to the small sample size.

 

Several questions remain unanswered. Although radical RT of 70 Gy remains the current standard of care,[4] it is unclear whether this ‘one-size-fits-all’ approach is appropriate for older adults with NPC. A logical consideration is RT dose de-escalation, aiming to balance optimal tumour control with minimised toxicity. Wang et al[34] demonstrated comparable outcomes between standard-dose RT (70 Gy) and reduced-dose RT (53-67 Gy) in patients with T1 to T3 NPC. However, there is currently no robust evidence supporting RT dose de-escalation specifically in older adults with NPC. Future studies are warranted to explore the optimal dose and fractionation schedules for this population.

 

This study has several strengths. To our knowledge, it is the first to specifically report treatment outcomes and toxicities in patients aged 80 years or above with NPC. All treatments were delivered using modern IMRT techniques, and acute and late treatment-related adverse events were prospective documented.

 

This study has several important limitations. First, inherent selection bias exists in this retrospective cohort comparison, as patients who received radical RT were likely to have been healthier overall, despite similar co-morbidity scores, and treatment decisions were influenced by unmeasured factors, including clinician judgement and patient preference. Second, comprehensive screening for distant metastases was not performed in some patients, particularly those who did not receive radical RT. It is therefore possible that a higher proportion of patients in Cohort B had undiagnosed stage IVb disease at presentation, which may have contributed to poorer outcomes. Third, the relatively small sample size limits the statistical power of the analysis and precludes the application of more sophisticated statistical methods, such as causal inference approaches (e.g., propensity score matching). Fourth, the follow-up duration was relatively short and some late toxicities may not yet have emerged. Fifth, formal geriatric assessments (such as comprehensive geriatric assessment) and quality-of-life evaluations were not conducted. Prospective multicentre studies with larger sample sizes, standardised geriatric assessments, and quality-of-life measurements are warranted to validate these findings and better inform clinical practice.

 

In appropriately selected patients aged 80 years or above with NPC, radical RT using modern IMRT techniques represents a viable treatment option, offering reasonable survival outcomes with an acceptable toxicity profile. Chronological age alone should not be regarded as a barrier to radical treatment in NPC.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hong Kong J Radiol 2026 Mar;29(1):e23-9 | Epub 27 February 2026

 

 

¹ Department of Orthopaedic Surgery, Singapore General Hospital, Singapore

² Department of Diagnostic Radiology, Singapore General Hospital, Singapore

 

  Full Article in PDF

Atypical femoral fractures (AFFs) were first recognised as a distinct clinical entity following multiple clinical reports, yet their pathophysiology and clinical characteristics remain incompletely understood.[1] [2] Over time, our understanding of AFFs has evolved, as reflected in ongoing efforts by a task force of the American Society for Bone and Mineral Research (ASBMR) to refine diagnostic criteria.[1] [2] Major features used to define AFFs were first established in 2010 and included fractures following low-energy or no trauma, transverse fractures originating from the lateral cortex which may become oblique medially, complete fractures with a medial spike, and incomplete fractures involving only the lateral cortex, with minimal or no comminution and localised periosteal or endosteal thickening of the lateral cortex.[1] Minor features associated but not required for diagnosis include generalised femoral diaphyseal cortical thickening, unilateral or bilateral prodromal pain in the groin or thigh, incomplete or complete fractures of both femoral diaphyses, and delayed fracture healing.[1] In 2014, new epidemiological studies and clinical data prompted the ASBMR to revise the definition of AFFs, emphasising their diaphyseal location and requiring at least four of the five major features for diagnosis.[2] This refined definition provides a more precise framework for identifying AFFs and distinguishing them from typical osteoporotic femoral fractures.2 This reflects the dynamic and evolving nature of our understanding of AFFs and highlights that much remains unknown, including the identification of potential novel clinical and radiological features and their implications for patient management.

 

Radiological studies have also expanded our understanding of AFFs, particularly when Mohan et al[3] described multifocal endosteal thickening along the femoral diaphysis in bisphosphonate-related AFFs, highlighting its association with a periosteal beak and/or a ‘dreaded black line’, also referred to as a radiolucent fracture line (RFL). These features were associated with an increased risk of progression to fracture.[3] A subsequent study by Png et al[4] demonstrated that when an RFL is present, the lesion is likely to persist, either remaining static or progressing to a displaced fracture. The significance of RFLs was also emphasised in the 2015 position statement by the Korean Society for Bone and Mineral Research, which recommended prophylactic femoral nailing in the presence of an RFL, especially when located in the subtrochanteric region.[5]

 

Despite these insights, gaps remain in the literature, as not all RFLs progress to complete fractures and there are no clear discerning features to guide when prophylactic fixation is indicated. During our review of patients with AFFs, we observed a previously undescribed radiological feature: perilesional sclerosis—an area of sclerosis closely associated with the presence of an RFL seen in an incomplete AFF. This finding, distinct from previously reported radiological features of AFFs, may have implications for understanding bone stability, fracture progression, and management strategies, as sclerosis has previously been suggested to be associated with fatigue fractures and delayed fracture healing.[6]

 

Although with established diagnostic criteria and the recognition of RFLs as high-risk markers, it remains unclear why not all RFLs progress to complete fractures or ultimately require intervention. To date, no study has described the presence or significance of perilesional sclerosis in relation to RFLs in AFFs. Our study aimed to address this gap by identifying and characterising this radiological feature in association with RFLs in incomplete AFFs, and by exploring its potential clinical implications.

 

We retrospectively reviewed plain radiographs of cases of incomplete AFFs in patients presenting to our institution, Singapore General Hospital in Singapore, while receiving bisphosphonate therapy between 2004 and 2020. These cases were retrieved from our institutional AFF registry, which includes patients exhibiting features of AFF that have not yet progressed to a complete fracture.

 

We reviewed all available plain radiographs of the AFFs, as well as those of the contralateral femur when available. Perilesional sclerosis was defined as a linear area of sclerosis observed on either side of an RFL. All anteroposterior and lateral views were obtained using standard radiographic techniques, and all analysed fractures met the ASBMR criteria for an AFF.[3] [4]

 

The study cohort of 100 AFFs was subsequently divided into three groups: Group 1A included AFFs with an RFL and perilesional sclerosis; Group 1B included AFFs with an RFL but without perilesional sclerosis; and Group 2 included AFFs without an RFL.

 

We also analysed age data and collected information on the type and duration of bisphosphonate therapy. Patients were followed up for sequelae, including progression to complete fracture or subsequent prophylactic fixation. Prophylactic fixation was performed in cases of persistent pain at the site of AFFs, while surgical fixation was performed for patients who progressed to complete fractures.

 

All radiographs were reviewed for the presence of RFLs with adjacent sclerosis using Vue Motion (Carestream Health, Rochester [NY], US), and independently assessed by two authors (SBK and TSH), each with over 20 years of clinical orthopaedic experience. RFLs were categorised into one of four patterns: (1) RFL without sclerosis (Figure 1); (2) RFL with linear sclerosis (Figure 2); (3) RFL with patchy continuous sclerosis (Figure 3); and (4) RFL with patchy non-continuous sclerosis (Figure 4).

 

Figure 1. (a) Illustration of radiolucent fracture line (RFL) without sclerosis. (b) Anteroposterior and (c) lateral radiographs of the left femur showing an RFL without sclerosis.

 

Figure 2. (a) Illustration of radiolucent fracture line (RFL) showing linear sclerosis. (b) Lateral radiograph of the left femur showing an RFL with linear sclerosis.

 

Figure 3. (a) Illustration of radiolucent fracture line (RFL) with patchy continuous sclerosis. (b) Anteroposterior and (c) lateral radiographs of the left femur showing an RFL with patchy continuous sclerosis.

 

Figure 4. (a) Illustration of radiolucent fracture line (RFL) with patchy non-continuous sclerosis. (b) Anteroposterior and (c) lateral radiographs of the left femur of the same patient in Figure 2 showing RFL with patchy non-continuous sclerosis.

 

We recorded the location of each lesion, along with the presence or absence of focal endosteal or periosteal thickening. Cases were followed up until fixation was required or a complete fracture occurred. Lesions were classified as being located in either the subtrochanteric or diaphyseal region, and further subdivided into proximal, middle, or distal thirds. Observations were collected independently by each of the same two authors and correlated. In the event of any discrepancies, a senior radiologist was consulted to provide a final decision.

 

Pearson’s Chi squared test was used to compare categorical data, while one-way analysis of variance was employed to analyse continuous variables. Statistical analyses were performed using SPSS (Windows version 23.0; IBM Corp, Armonk [NY], US). Statistical significance was defined as p < 0.05.

 

There were 100 radiographs of AFFs from 80 cases available for review. Demographic and clinical data of the study cohort are summarised in Table 1. There were 17 femurs in Group 1A, 35 femurs in Group 1B, and 48 femurs in Group 2. All 17 femurs with perilesional sclerosis were independently identified by the two authors previously described. There were no significant differences among the three groups in terms of patient demographics (age: p = 0.979); the patients were predominantly female and Asian. All had a history of bisphosphonate use, except for two AFF cases with a history of denosumab use only (one in Group 1A and one in Group 2). There were no significant differences in the duration of antiresorptive therapy (p = 0.418), progression to complete fracture (p = 0.078), or subsequent prophylactic fixation (p = 0.076) among the three groups. The radiographic finding of perilesional sclerosis was observed in 17 of the 100 femurs (17%), with bilateral involvement in three patients who were all female with a mean age of 66 years; two were Chinese (88.2%) and the remaining patient was of Indian descent. The mean (± standard deviation) duration of bisphosphonate use was 66 ± 31 months (range, 4-120). Only one femur in Group 1A was from a patient with a history of denosumab use without prior bisphosphonate therapy. Bisphosphonate treatment was discontinued upon diagnosis of AFF in all patients. Three femurs (17.6%) subsequently progressed to complete fractures, while six incomplete fractures required prophylactic fixation (35.3%). The mean time to surgical fixation or prophylactic fixation from the date of presentation with perilesional sclerosis was 9 ± 12 months.

 

Table 1. Clinical data of the study cohort (n = 100).

 

The radiographic features of perilesional sclerosis in Group 1A are summarised in Table 2. Each sclerotic lesion was observed in an incomplete AFF and only in the presence of an RFL. Perilesional sclerosis was identified on lateral views in 15 femurs (88.2%), while only nine femurs (52.9%) demonstrated sclerosis on anteroposterior views. The lesions were mainly located in the subtrochanteric region (n = 7, 41.2%), followed by the proximal diaphyseal region (n = 6, 35.3%) and the mid-diaphyseal region (n = 4, 23.5%). All lesions were associated with either adjacent endosteal thickening or periosteal thickening. Of the 17 lesions with perilesional sclerosis, 16 (94.1%) were RFLs with patchy sclerosis of varying widths along either side of the fracture line, and 10 (58.8%) demonstrated patchy non-continuous sclerosis.

 

Table 2. Radiographic features of perilesional sclerosis (n = 17).

 

Among the 17 femurs with sclerotic RFLs, three had earlier radiographs (mean, 56.4 months; range, 0.5-96.3) showing an RFL without adjacent sclerosis, indicating that perilesional sclerosis developed later. Once sclerosis appeared, it persisted in all subsequent follow-up radiographs. For the eight femurs that did not undergo surgery, the mean duration between the first presentation of RFL with perilesional sclerosis and the last available radiograph was 31 ± 22 months (range, 0-57.6). Follow-up was achieved for 100% of the 17 lesions, with a mean follow-up duration of 72 ± 45 months (range, 8-184). Regarding inter-observer variability, there was complete agreement between both readers on the presence and pattern of sclerosis.

 

Our study describes the presence of perilesional sclerosis adjacent to the RFL, previously described by Png et al,[4] and the radiological progression of AFFs in which the RFL is recognised as the penultimate radiological feature before progression to a complete fracture. We observed that an RFL can be associated with, or may later develop, perilesional sclerosis. This radiological feature has not previously been documented in AFFs, which are predominantly located between the subtrochanteric region and mid-diaphyseal regions of the femur, may be bilateral and are consistently associated with endosteal or periosteal thickening. In our study, perilesional sclerosis appears to occur in approximately one-third of AFFs with an RFL, and is usually seen on lateral radiographic views, and occasionally on anteroposterior radiographic views. While the variability in its appearance and its significance remain largely unstudied, and descriptions in the literature are scarce, our study presents observations that may enhance our understanding of this entity.

 

AFFs are considered to be ‘tensional’ stress fractures, typically initiating along the upper two-thirds of the lateral femoral shaft corresponding to regions subjected to greater tensional forces.[7] Accordingly, RFLs are better observed as linear structures across the femoral diaphysis on lateral views. As perilesional sclerosis appears to occur in association with RFLs, this may account for its notably high prevalence on lateral views of the femoral shaft.

 

In the majority of our cases, perilesional sclerosis was observed only on the lateral views. In two cases with both anterior and lateral cortical thickening, sclerosis was visible on both anteroposterior and lateral views. These findings suggest that, in most cases, perilesional sclerosis may be related to viewing cortical thickening at right angles to its long axis. However, in two cases, perilesional sclerosis was seen on the anteroposterior views despite cortical thickening being confined to the lateral cortex. This suggests that, in these cases, there was focal sclerosis at the intracortical fracture margins.

 

Radiologically, sclerosis at fracture sites has been described as a feature of fracture non-union.[8] Although sclerosis has been postulated to be associated with avascular necrosis or reduced metabolic bone activity,[9] it has also been linked to prolonged time to union.[10] Perilesional sclerosis has been mentioned in some cases of insufficiency fractures but is rarely described in AFFs. Only a single study by McKenna et al[11] described sclerosis in relation to AFFs, but only on computed tomography scans without specific reference to its relationship with RFLs. The fact that this feature is observed only in a subset of incomplete AFFs with variable continuity along the RFL, suggests that it may represent a phase in the pathophysiological progression of AFFs.

 

Perilesional sclerosis associated with cortical thickening may resemble that seen in stress or fatigue fractures. However, cases with intracortical perilesional sclerosis may represent an early phase of the process leading to non-union. Fracture non-union is usually associated with sclerosis at the fracture margins, and the two cases in our cohort where sclerosis was confined to the lateral cortex may represent non-union of the incomplete fracture, akin to hypertrophic non-union involving the lateral cortex. This may be the result of persistent tensile stresses that inhibit bony union.[12] These lesions also appeared to progress from an isolated RFL to an RFL with adjacent sclerosis, with this radiographic feature persisting for a mean duration of 31 ± 22 months. Perilesional sclerosis may take considerable time to develop but can persist long after initial presentation. We postulate that it could represent the development of a chronic non-union state in incomplete AFFs. Bisphosphonates such as alendronate are known to have prolonged effects on osteoclast function, and these may continue long after cessation of therapy.[13]

 

Although there were no significant differences in the proportion of cases that progressed to complete fracture or required prophylactic fixation between Group 1A and Group 1B, a higher rate of surgical fixation in Group 1B was noted (65.7% vs. 52.9%). A histological study by Schilcher et al[14] demonstrated signs of attempted healing at the site of AFFs; however, the current literature does not explain the pathological differences between AFFs that eventually heal and those that do not. Future histological studies could examine samples of perilesional sclerosis to explore the underlying pathology and provide insights into its clinical significance.

 

A strength of this study is the 100% follow-up rate over a mid-term duration for a previously undescribed radiological finding in AFFs. The main limitation is the limited sample size of patients with perilesional sclerosis, although this may reflect the low prevalence of AFFs among patients on antiresorptive therapy. Additionally, the predominance of female patients in the cohort limited our ability to assess potential gender-related differences. Another limitation is the irregular follow-up of patients due to variation in individual physicians’ clinical practices and the retrospective nature of the study. Longer-term, regularly scheduled follow-up with standardised radiographic imaging should be considered in future studies to better evaluate the relationship between these lesions and fracture outcomes.

 

We describe perilesional sclerosis as a previously unrecognised radiological feature along the RFL, in incomplete AFFs with distinctive characteristics. Its presence may suggest a state of non-union and was observed in approximately one-third of cases with an RFL. Further research involving larger cohorts could shed light on its pathophysiological and prognostic significance.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hong Kong J Radiol 2026 Mar;29(1):e30-3 | Epub 26 February 2026

 

 

Nuclear Medicine Unit, Department of Diagnostic and Interventional Radiology, Queen Elizabeth Hospital, Hong Kong SAR, China

 

  Full Article in PDF

A 79-year-old female with a past medical history of hypertension and impaired fasting glucose presented to our institution in April 2020 with a neck mass and fever. She was an ex-smoker with no known drug allergies. Following an ear, nose, and throat consultation, she was diagnosed with stage 4B diffuse large B-cell lymphoma (DLBCL). A biopsy of the left tonsil revealed high-grade B-cell lymphoma, consistent with DLBCL. Further evaluation including bilateral bone marrow aspiration and bilateral trephine biopsy showed no evidence of lymphoma involvement.

 

Staging fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) revealed hypermetabolic lymphadenopathy on both sides of the diaphragm, consistent with the biopsy-proven lymphoma, as well as hypermetabolic lesions in bilateral tonsils, confirming lymphomatous involvement (Figure 1).

 

Figure 1. Staging fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) in a patient with biopsy-proven diffuse large B-cell lymphoma. (a) Maximum intensity projection shows multiple hypermetabolic, enlarged lymph nodes on both sides of the diaphragm. Transaxial (b and c) plain CT and (d and e) fused PET/CT images show supradiaphragmatic involvement, and transaxial (f) plain CT and (g) fused PET/CT images show infradiaphragmatic involvement.

 

The patient commenced R-CHOP chemotherapy (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone), receiving six cycles over 5 months. The first cycle was administered at 50% dosage, with subsequent cycles adjusted for tolerance and side-effects. Following completion of the last cycle, an end-of-treatment ¹⁸F-FDG PET/CT scan demonstrated complete metabolic remission, with a Deauville score of 2 (Figure 2).

 

Figure 2. Fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) following treatment with R-CHOP chemotherapy (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone). (a) Maximum intensity projection shows significant metabolic improvement or resolution on both sides of the diaphragm. Transaxial (b) plain CT and (c) fused PET/CT images show resolved supradiaphragmatic lymph nodes, and transaxial (d) plain CT and (e) fused PET/CT images show resolved infradiaphragmatic lymph nodes.

 

Five months after completing R-CHOP chemotherapy, the patient developed a right neck mass and numbness over the right side of her neck and right lower limb, with muscle power graded at 2 out of 5. A CT scan revealed a large soft tissue mass on the right side of the oropharynx, and biopsy confirmed DLBCL with CD20 positivity. A subsequent ¹⁸F-FDG PET/CT scan for restaging revealed a new hypermetabolic soft tissue mass in the right side of the oropharynx, consistent with lymphomatous involvement, with a Deauville score of 5. Notably, the scan also revealed new, multiple hypermetabolic foci involving perineural and muscular involvements in the bilateral head and neck regions and the right proximal lower limb, raising suspicion for perineural lymphomatous infiltration (Figure 3).

 

Figure 3. Recurrence and suspected atypical lymphomatous involvement in neuromuscular regions. (a) Maximum intensity projection of the fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography (¹⁸F-FDG PET/CT) shows a hypermetabolic right oropharyngeal lesion (arrow). Transaxial (b) plain CT and (c) fused PET/CT images show the corresponding hypermetabolic lesion (arrows). Transaxial (d) plain CT and (e) fused PET/CT images show hypermetabolic left perineural involvement along the distribution of the left trigeminal branches (arrows). (f) Maximum intensity projection shows a hypermetabolic right neck perineural and muscular lesion (red arrow) and a right lower limb perineural and muscular lesion (purple arrow). Transaxial (g) plain CT and (h) fused PET/CT images show a hypermetabolic right neck neuromuscular lesion over the right trapezius muscle and accessory nerve region (arrows). Transaxial (i) plain CT and (j) fused PET/CT images show a hypermetabolic right lower limb neuromuscular lesion in the region of the right sciatic nerve (arrows).

 

The patient subsequently received six cycles of R-IMVP-16 (rituximab, ifosfamide, methotrexate, etoposide, and prednisone) over 5 months. End-of-treatment ¹⁸F-FDG PET/CT showed metabolic resolution of the right tonsillar/oropharyngeal mass and other infiltrative perineural lesions in the neck region and right lower limb, indicating a favourable treatment response (Figure 4). Clinically, her numbness subsided, with improved sensation in the previously affected regions and right lower limb power improved to 4 out of 5, consistent with the ¹⁸F-FDG PET/CT findings. Both clinical and imaging findings favoured a positive treatment response of the perineural and muscular lymphomatous involvement in this patient with recurrent lymphoma.

 

Figure 4. Maximum intensity projection of fluorine-18 fluorodeoxyglucose positron emission tomography/computed tomography with significant metabolic improvements or resolutions of the oropharynx, right neck and right lower limb hypermetabolic lesions after treated with R-IMVP-16 (rituximab, ifosfamide, methotrexate, etoposide, and prednisone) chemotherapy.

 

Perineural and muscular involvement in DLBCL is rare, with only a limited number of cases reported in the literature.[1] The underlying mechanisms are not fully understood, but it is believed that DLBCL may infiltrate muscle tissue either via a haematogenous route or through adjacent lymphatic structures.[2] Clinical manifestations can vary widely, with patients presenting with muscle weakness, myalgia, or neuropathic symptoms.[3] Differential diagnoses for FDG-avid perineural and muscular lesions include polyneuritis, compartment-related compression radiculopathy, and tuberculosis. In polyneuritis, the pattern of increased FDG uptake is usually symmetrical and occurs without associated soft tissue thickening.[4] [5] The significant soft tissue thickening in our case made compartment-related compression radiculopathy less likely. Active tuberculosis was excluded through microbiological investigations.

 

This case demonstrated that the patient’s neuropathic symptoms and imaging findings were indicative of perineural and muscular involvement. The identification of hypermetabolic activity in the muscles on ¹⁸F-FDG PET/CT was crucial in establishing the diagnosis due to the asymmetrical metabolic distribution and soft tissue thickening in the affected regions. These abnormalities resolved in parallel with the biopsy-proven recurrent right oropharynx DLBCL, both metabolically and morphologically. Such findings are often mistaken for primary myopathies or neuropathies.

 

In our case, ¹⁸F-FDG PET/CT not only confirmed the recurrence of DLBCL but also revealed the unusual sites of perineural and muscular involvement. This underscores the importance of considering extranodal manifestations of DLBCL, as it ultimately guided treatment decisions. Furthermore, the most recent ¹⁸F-FDG PET/CT showed both metabolic and morphological resolution of the hypermetabolic perineural and muscular lesions, supporting the diagnosis of atypical lymphomatous involvement and reflecting a significant treatment response.

 

Previous studies[6] [7] revealed that perineural and muscular involvement in DLBCL is largely underreported, with only a limited number of cases documented—primarily in patients with advanced-stage disease—and highlighted the importance of recognising ¹⁸F‑FDG PET/CT findings in atypical sites of lymphomatous involvement to avoid misdiagnosis and ensure appropriate management. Primary muscular lymphoma[6] and other atypical sites of DLBCL involvement[6] [7] have also been reported.

 

The utility of ¹⁸F-FDG PET/CT in the staging and treatment monitoring of DLBCL has been examined,[8] [9] which concluded that this imaging modality provides valuable insights into disease burden and can identify sites of active disease that may not be evident on conventional imaging. This aligns with our case, in which ¹⁸F-FDG PET/CT played a pivotal role in diagnosing perineural and muscular involvement in a one-stop-shop manner.

 

The management of DLBCL with perineural and muscular involvement is complex and often requires a multidisciplinary approach.[10] [11] Treatment options may include chemotherapy, radiotherapy, and targeted therapies, depending on the extent of disease and the patient’s overall health.

 

In our case, the patient was commenced on a salvage chemotherapy regimen following relapse of DLBCL. Given the aggressive nature of her disease, close monitoring with repeat ¹⁸F-FDG PET/CT was planned to assess treatment response. The prognosis for patients with perineural and muscular involvement in DLBCL varies, but early detection and timely intervention can significantly improve clinical outcomes.

 

This case highlights the importance of ¹⁸F-FDG PET/CT in detecting perineural and muscular involvement in patients with recurrent DLBCL. Early detection of the disease involvement using ¹⁸F-FDG PET/CT can guide biopsy targeting, inform appropriate treatment strategies and serve as a reference for assessing treatment response on end-of-treatment imaging, all of which are crucial for improving patient outcomes.

 

 

 

 

 

 

 

 

 

 

 

 

 

Hong Kong J Radiol 2026 Mar;29(1):e38-43 | Epub 10 March 2026

 

 

¹ Department of Radiology, Princess Margaret Hospital, Hong Kong SAR, China

² Vascular Surgery Department, Princess Margaret Hospital, Hong Kong SAR, China

 

  Full Article in PDF

A 59-year-old male patient with a history of smoking, metabolic syndrome, ischaemic heart disease, and long-standing peripheral arterial disease presented to our institution in October 2022 with recurrent claudication. He had previously undergone multiple lower limb angioplasties and stenting procedures at various institutions between 2018 and 2021 for recurrent in-stent restenosis. These included an EverFlex stent (Medtronic, Plymouth [MN], US) in the left external iliac artery (EIA), a Protégé stent (Medtronic, Plymouth [MN], US) in the left common iliac artery (CIA), an EverFlex stent in the right EIA, a Supera stent (Abbott, Santa Clara [CA], US) from the right common femoral artery to the proximal superficial femoral artery (CFA-pSFA), a Zilver stent (Cook Medical, Limerick, Ireland) in the right mid superficial femoral artery (mid-SFA), and a Supera stent from the right distal superficial femoral artery to the popliteal artery (dSFA-pop) [Figure 1].

 

Figure 1. A summary of stents previously placed in the patient between 2018 and 2021 at various institutions. Left common iliac artery (LCIA), left external iliac artery (LEIA), right external iliac artery (REIA), and mid superficial femoral artery (mid-SFA) stents were placed in 2018 for peripheral vascular disease. A distal superficial femoral artery–popliteal artery (dSFA-pop) stent was placed in early 2021. A common femoral artery to proximal superficial femoral artery (CFA-pSFA) stent was placed in November 2021 to bridge the REIA and mid-SFA stents. A thin white line depicts the course of the retrograde guidewire during the 2023 SAFARI (subintimal arterial flossing with antegrade-retrograde intervention) double-barrel stenting procedure. The intraluminal position within the dSFA-pop stent, subintimal position outside the lumen of the mid-SFA stent, and subsequent intraluminal re-entry into the CFA-pSFA stent were confirmed by fluoroscopy and intravascular ultrasound. The retrograde wire was subsequently advanced into a long sheath to establish a floss wire between the right ankle and left groin access.

 

The patient presented in 2022 with recurrent claudication following placement of a bridging CFA-pSFA stent between the right EIA and mid-SFA stents, with claudication distance reduced to 10 metres. In view of his recurrent symptoms, the authors were consulted for suspected stent occlusion of the previously placed multi-stent system. Computed tomography angiography revealed an in-stent occlusion due to misalignment of the CFA-pSFA and mid-SFA stents (Figure 2), likely resulting from subintimal placement of the CFA-pSFA stent.

 

Figure 2. Computed tomography angiogram in 2022 showing stent occlusion, likely resulting from malalignment of the common femoral artery to the Supera stent (blue arrows) of the proximal superficial femoral artery (CFA-pSFA) and the Zilver stent (black arrows) of the mid superficial femoral artery (mid-SFA). The distal margin of the CFA-pSFA stent is seen within the subintimal space, external to the mid-SFA stent. (a) Axial view. (b) Sagittal reconstruction.

 

Digital subtraction angiography images in the anteroposterior projection from the previous procedure in November 2021 revealed apparent alignment of the CFA-pSFA and mid-SFA stents, with improved runoff post-stenting (Figure 3). Lateral views were unavailable. In view of the recurrent claudication and the in-stent occlusion, repeat angioplasty was performed in October 2023. Left CFA access with crossover was performed. A 6-Fr Destination long sheath (Terumo, Somerset [NJ], US) was placed in the right CIA. A 0.035-inch guidewire (Terumo, Tokyo, Japan) was advanced through the lumen of the occluded right CFA-pSFA stent, encountering resistance (Figure 4a). Inability to negotiate the wire into the right mid-SFA stent led to a decision to obtain retrograde access via the right posterior tibial artery (PTA). With the aid of a 2.6-Fr CXI microcatheter (Cook Medical, Bloomington [IN], US), a 0.018-inch Advantage wire (Terumo, Tokyo, Japan) was advanced retrogradely through the PTA and the dSFA-pop Supera stent intraluminally. The wire was manipulated at the junction of the mid-SFA Zilver stent and dSFA-pop Supera stent, entering the subintimal space. After further subintimal manipulation, re-entry of the retrograde wire into the lumen of the occluded CFA-pSFA stent was achieved. The wire was then advanced into the right EIA/CIA stent lumen (Figures 1 and 4b). Wire position was confirmed by intravascular ultrasound (IVUS) [Visions PV 0.018-inch catheter (Phillips, Rancho Cordova [CA], US)] and angiography. The retrograde wire was externalised through the 6-Fr crossover sheath and retrieved via the left groin access to establish through-and-through access.

 

Figure 3. Retrospective review of prior common femoral artery to proximal superficial femoral artery bridging stent placement in 2021 showing (a) apparent stent alignment (black arrow) and (b) acceptable runoff on completion angiography.

 

 

Figure 4. Digital subtraction angiography images of angioplasty and double-barrel stenting performed in November 2023. (a) Crossover wire from left femoral access, with the tip positioned within the right common femoral artery to the proximal superficial femoral artery (CFA-pSFA) stent, encountering resistance due to occlusion. The occlusion was eventually navigated; however, in view of failure to re-enter the mid superficial femoral artery (mid-SFA) stent lumen, a retrograde approach was employed. (b) A 0.018-inch retrograde wire via right posterior tibial artery access was advanced intraluminally through the occluded distal superficial femoral artery popliteal artery stent, with subsequent entry into the subintimal space outside the Zilver SFA stent and re-entry into the intraluminal occluded CFA stent. The long sheath from the left groin access was cannulated by the retrograde wire and subsequently externalised, establishing a through-and-through floss wire. Position was confirmed by intravascular ultrasound (Figure 6). (c) After establishment of the floss wire, angioplasty of the intraluminal-subintimal-intraluminal wire tract was performed. (d, e) Angioplasty of the posterior tibial artery and tibioperoneal trunk was performed, followed by mid-SFA stenting with double-barrel exclusion of the Zilver stent. Completion angiography showed significant restoration of flow between the newly deployed mid-SFA stent and adjacent stents.

 

Angioplasty was performed along the wire path from the right popliteal artery stent to the left EIA stent with an Armada 6 × 200 mm² balloon (Abbott, Santa Clara [CA], US), expanding the subintimal space for subsequent stenting. Following IVUS sizing, double-barrel subintimal stenting was performed by deploying a Supera 5.5 × 80 mm² stent to bridge the CFA-pSFA and dSFA-pop stents (Figure 4c). Additional angioplasty of the newly deployed stent, as well as the PTA and tibioperoneal trunk, was performed with an Armada 2.5 × 200 mm² balloon. Completion angiography demonstrated re-establishment of flow through the double-barrel subintimal stent, with a patent intraluminal-subintimal-intraluminal channel and crush exclusion of the Zilver mid-SFA stent (Figure 4d and e). Postoperatively, the patient resumed apixaban 5 mg twice daily and aspirin 80 mg daily.

 

At 1-month follow-up, symptoms improved from claudication after walking 20 steps (Rutherford grade III) to no claudication (Rutherford grade 0). There was no evidence of tissue loss or vascular compromise. At 8 months, follow-up computed tomography showed successful crush exclusion of the mid-SFA Zilver stent (Figure 5). There was complete alignment of the mid-SFA Supera stent with adjacent proximal and distal stents, with preserved patency and no significant in-stent restenosis (Figure 5). However, mild-to-moderate instent restenosis was noted in the previously implanted popliteal and iliac stents. The patient remains under surveillance and is scheduled for repeat angioplasty to preserve the patency of the multi-stent system (online supplementary Figure).

 

Figure 5. Follow-up computed tomography angiogram at 8 months postprocedure showing successful double-barrel stenting with exclusion of the mid superficial femoral artery (mid-SFA) Zilver stent (black arrows in [a] and [c]) and a patent new Supera mid-SFA stent (blue arrows in [a] and [c]): (a) axial view; (c) oblique sagittal view. Oblique coronal reconstruction showing patency throughout the intraluminal-subintimal-intraluminal multi-stent system (b), including the distal overlapping stent segments (black arrowhead in [d]) and the proximal overlapping stent segments (blue arrowhead in [e]).

 

Our case highlights several important considerations for interventionists. In retrospect, inadvertent subintimal stent placement could have been avoided through several measures. First, routine biplanar imaging could prevent false assurance from a single anteroposterior projection and detect stent misalignment. Attention should also be paid to contrast pooling around the stent tips and the rate of contrast runoff; delayed clearance may alert the operator to possible distal outflow impairment or subintimal entry. Second, careful observation of the guidewire tip behaviour and mobility may alert interventionists to inadvertent subintimal entry. In cases of initial intimal dissection and subintimal entry, the tip load of the guidewire may be exceeded with the wire tip bending in the reverse direction and a ‘crushing’ sensation commonly reported.[1] Initial entry into the potential subintimal space may restrict free wire rotation. Nonetheless, where manipulation continues and the wire tracks further into an enlarging subintimal space, guidewire rotation may become freer, with loss of resistance. Prolonged manipulation should be avoided if early intraluminal re-entry fails, as this may enlarge the subintimal space and further complicate luminal re-entry. Third, in cases of equivocal wire position, familiarity with IVUS may assist operators in accurately stenting within the true lumen. The IVUS images from our salvage procedure are shown (Figure 6). Although resource-intensive and operator-dependent, IVUS enables more accurate visualisation of the vascular and subintimal spaces with applications not only in stent positioning but also in the accurate arterial stent sizing.[2] [3]

 

Figure 6. Intravascular ultrasound (IVUS) images confirming wire positioning from the 2023 SAFARI (subintimal arterial flossing with antegrade-retrograde intervention) double-barrel stenting procedure. (a) IVUS enables vessel sizing for selection of appropriate catheters and stents. (b) IVUS image showing the echogenic guidewire in the intraluminal space (arrow). (c) IVUS image showing echogenic guidewire in the subintimal space (arrow). (d) IVUS allows assessment of stent margins to prevent malalignment and inadvertent subintimal entry.

 

In cases of inadvertent subintimal entry or dissection, achieving luminal re-entry remains a major challenge, and familiarity with re-entry techniques is essential for interventionists. If spontaneous re-entry cannot be achieved with a standard wire, specific re-entry devices such as the Outback (Cordis, Miami Lakes [FL], US) may be utilised. Promising data demonstrated technical success and primary stent patency rates of up to 92.3% at 12 months with the Outback, as subintimal angioplasty gains increasing recognition in the treatment of long-segment TransAtlantic Inter-Society Consensus II class C/D lesions.[4] Where such devices are unavailable, several alternative approaches may be considered, including retrograde access via the distal artery with establishment of a through-and-through floss wire using the subintimal arterial flossing with antegrade-retrograde intervention (the SAFARI [subintimal arterial flossing with antegrade-retrograde intervention] technique[5] as in our case), the parallel wire technique,[6] the wire rendezvous technique with ballooning of subintimal space as seen in CART (controlled antegrade and retrograde subintimal tracking), reverse CART, or the double-balloon technique.[7] [8]

 

In our experience, SAFARI can be performed in several ways once luminal re-entry has been achieved. First, a nitinol snare system may be deployed via antegrade access to capture the retrograde wire intraluminally and establish a through-and-through access.[9] Alternatively, retrograde manipulation of the wire tip within the sheath or catheter via groin access may be performed (as in our current case).

 

To the best of our knowledge, cases of double-barrel subintimal stenting are sparsely reported in the literature and have not been reported locally. Several case reports describe double-barrel stenting (DBS) for exclusion of occluded stents in salvage procedures for lower limb and coronary arterial occlusions,[10] [11] [12] although this remains an uncommonly utilised technique. One case series of three patients with peripheral arterial disease following DBS reported varying degrees of success, with the longest assisted secondary patency of up to 85 months,[13] supporting its feasibility and long-term patency.

 

We report a case of prior inadvertent subintimal stenting of a bridging CFA-pSFA stent, followed by successful salvage with subintimal DBS using the SAFARI technique within a multi-stent system. Methods to reduce the risk of inadvertent subintimal stenting are discussed. Subintimal manipulation and re-entry techniques with antegrade-retrograde approaches are also discussed as important tools for interventionists. Although not commonly employed, DBS has been described in several case reports and small case series. Our case affirms the feasibility of this technique where salvage of inadvertent subintimal stenting is necessary.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Hong Kong J Radiol 2026 Mar;29(1):e44-57 | Epub 6 March 2026

 

 

¹ Department of Radiodiagnosis, Basavatarakam Indo American Cancer Hospital and Research Institute, Hyderabad, India

² Department of Surgical Oncology, Basavatarakam Indo American Cancer Hospital and Research Institute, Hyderabad, India

 

  Full Article in PDF

Inferior vena cava (IVC) is the largest vein in the body, draining blood from the lower extremities, pelvis, and abdomen into the right atrium. Accurate anatomical assessment is crucial when planning vascular interventions, resections, anastomoses, and reconstructions that form an integral part of the surgical management of abdominopelvic malignancies. Anomalies and variants can complicate access to the IVC and its tributaries during interventional procedures and filter placement. Given that abdominopelvic oncological surgeries require extensive dissections, unawareness of vascular involvement and congenital anomalies can lead to inadvertent injuries with catastrophic outcomes. Contrast-enhanced computed tomography with reconstruction is the gold-standard non-invasive investigation for presurgical mapping; ultrasound with colour Doppler, magnetic resonance imaging, and positron emission tomography/computed tomography often play complementary roles in evaluating the IVC and its draining veins. This pictorial essay presents several illustrative cases from our experience at a tertiary care cancer centre in India.

 

The embryogenesis and development of the IVC is a complex process, and multiple congenital variations can arise from abnormal persistence or regression of embryological veins (Table 1).[1] [2] [3] [4] [5] [6] [7] [8] These congenital anomalies are collectively present in 4% of the population.[2] [3] [4] [5] [6] [7] [8] The most common clinically significant variations include duplication of the IVC and absence or agenesis (interruption) of the IVC with prominent hemiazygos-azygos pathways[2] (Figures 1, 2, 3, 4, and 5). Because visceral thoracic and abdominal organs demonstrate left-right anatomical asymmetry, awareness of discrepancies in laterality and venous drainage into the IVC—such as in situs inversus and heterotaxy syndromes—is critical before undertaking biliary, hepatic, and gastric surgeries (Figure 6). Variations in renal vein anatomy are often asymptomatic and overlooked but are crucial during renal or adrenal surgeries and retroperitoneal dissections. Anomalous veins and collateral vessels may be misdiagnosed as lymphadenopathy; hence, contrast imaging is essential in all cases of malignancy (Figures 7, 8, and 9).

 

Table 1. Common congenital anomalies involving the inferior vena cava and their incidence.[1] [2] [3] [4] [5] [6] [7] [8]

 

Figure 1. Volume-rendered contrast-enhanced computed tomography images. (a) Interrupted inferior vena cava (IVC) with absence of the infrahepatic IVC. The suprahepatic IVC (white arrow) drains into the right atrium (RA) of the heart. The normal portal vein (PV) and the aorta (A) are visible. (b) Rare case of complete duplication of the superior vena cava in the thorax (orange arrows) and the IVC in the abdomen (white arrows), with multiple bridging veins between duplicated segments (blue arrows). (c) The IVC (white arrow) lies to the left of the aorta (A), with dextrocardia in a patient with situs inversus totalis.

 

Figure 2. (a) Anterior and (b) posterior volume-rendered contrast-enhanced computed tomography images show a left inferior vena cava with hemiazygos continuation, crossing the midline posterior to the aorta (A) and draining into the azygos-superior vena cava pathway (white arrows). Hepatic veins are visible draining separately into the right atrium (green arrows).

 

Figure 3. Coronal maximum intensity projection contrast-enhanced computed tomography images of the abdomen. (a) Normal inferior vena cava (IVC) [white arrow] to the right of the aorta (A), formed by the confluence of the bilateral common iliac veins (red arrows), draining into the right atrium (RA). (b) Duplication of the IVC (white arrows) in a patient with gastric malignancy (star). The left infrarenal IVC crosses the midline anterior to the aorta (A) and joins the right IVC. (c) Duplication of the IVC (white arrows) on both sides of the aorta (A) in a patient with endometrial malignancy (star) and a complex right ovarian cyst (yellow arrow).

 

Figure 4. Coronal (a) and axial (b, c) contrast-enhanced computed tomography images in a patient with gastric malignancy (star in [a]) show polysplenia (S) and a right inferior vena cava with azygos continuation (white arrows) draining into the superior vena cava (curved arrow in [a]). Hepatic veins drain separately into the right atrium (black arrow in [b]), and para-aortic lymphadenopathy is also noted (red arrows in [c]).

 

Figure 5. (a) Coronal contrast-enhanced computed tomography and (b) posterior maximum intensity projection images of a patient with cervical carcinoma and pyometra (star in [a]) show a left inferior vena cava crossing the midline posterior to the aorta (A), continuing as the azygos-superior vena cava pathway (yellow arrows). Hepatic veins (black arrows) drain separately into the right atrium (RA).

 

Figure 6. Two cases of situs inversus totalis. (a) A 48-year-old woman with bilateral ovarian malignancy (stars). The right ovarian vein (red arrows) crosses the midline posterior to the aorta (A) and drains into the inferior vena cava (IVC) [yellow arrow], while a prominent left ovarian vein drains directly into the left-sided IVC (blue elbow arrow). (b) A 55-year-old man with hepatocellular carcinoma (star). The IVC (yellow arrow) lies to the left of the aorta (A), with an intraluminal thrombus present in the hepatic and suprahepatic segments (black arrow).

 

Figure 7. (a) Axial and (b, c) coronal contrast-enhanced computed tomography images of a patient with left renal cell carcinoma (stars) show a duplicated inferior vena cava (IVC) on either side of the aorta (A), with azygos continuation of the right IVC (red arrows). The left IVC (white arrows) crosses the midline and drains into the right IVC-azygos-superior vena cava pathway (green curved arrow in [b]). Tumour thrombus is present in the left renal vein and the left IVC, extending across the midline into the azygos continuation of the right IVC (black arrows). Hepatic veins drain separately into the right atrium (yellow elbow arrow in [c]).

 

Figure 8. Contiguous axial contrast-enhanced computed tomography images showing anomalous renal veins. (a) Retroaortic left renal vein (red arrow) posterior to the aorta, draining into the normal right inferior vena cava (IVC) [white arrow] in a patient with abdominal liposarcoma (star). (b) Retroaortic right renal vein (red arrow) draining into the left IVC (white arrow). (c) Circumaortic left renal veins (red arrows) passing anterior and posterior to the aorta, draining into the right-sided IVC (white arrows).

 

Figure 9. (a) Tortuous left renal vein draining into the left common iliac vein (white arrow) instead of the inferior vena cava (IVC) [black arrow]. (b) Anterior and (c) posterior views show ‘horseshoe’ kidneys (K) with vertically oriented renal veins (white arrows in [b]) and gonadal veins (red arrows in [c]) draining into the IVC (black arrow in [b]) in a patient with endometrial malignancy (stars).

 

The major acquired venous pathologies in abdominopelvic malignancies include external compression or infiltration of the IVC and its draining veins by neoplasms (Figures 10 and 11), metastatic lymph nodes (Figure 12), and/or intraluminal thrombosis.

 

Figure 10. Right adrenocortical malignancy. (a) Axial and (b) coronal contrast-enhanced computed tomography images of the abdomen show a large, heterogeneously enhancing hypoattenuating lesion (stars) invading the inferior vena cava (arrows).

 

Figure 11. Retroperitoneal liposarcoma. (a) Axial and (b) coronal contrast-enhanced computed tomography images of the abdomen show large fatty component (stars) and small soft-tissue component (red arrows) encasing the inferior vena cava (white arrows).

 

Figure 12. A 24-year-old man with lymphoma. (a) Axial and (b) coronal contrast-enhanced computed tomography images of the abdomen show splenomegaly (stars) and conglomerated nodal mass (red arrows) encasing and causing narrowing of the inferior vena cava (IVC) [white arrows], aorta, and their branches. (c, d) Corresponding post-chemotherapy images show a significant decrease in the size of the nodal mass (red arrows) and spleen (star in [d]), with expansion and visualisation of the IVC (white arrows).

 

Malignancies most commonly involving the IVC include those of the liver (4.0%-5.9%), kidney (4%-10%), and adrenal glands (9%-19%).[4] [8] Although the portal veins are more frequently involved, abnormalities of the hepatic artery, hepatic veins, and IVC may occur in hepatocellular carcinomas; accordingly, triphasic computed tomography should be performed in the evaluation of liver malignancies (Figure 13).

 

Figure 13. Hepatocellular carcinoma (HCC) in a 56-year-old man. (a) Axial and (b) coronal contrast-enhanced computed tomography (CECT) images in the early arterial phase show contrast opacification of the aorta (A), hepatic artery, and portal vein due to an arterioportal fistula in the right lobe of the liver (black arrows). (c) Axial CECT image shows HCC (star) with thrombus in the inferior vena cava (IVC) [black arrow]. (d, e) Coronal CECT images in the venous phase show HCC (stars), along with thrombus in the portal vein (red arrow in [d]) and hepatic veins (black arrows) extending into the IVC (white arrows).

 

Cancer-associated thrombosis is recognised as the most common complication of cancer and is attributed to several factors (Table 2).[7] [8] [9] [10] [11] [12] Compared with the general population, patients with cancer have a 12-fold increased risk of developing venous thrombosis, as well as a significantly worse prognosis[9] [10] (Figure 14). The IVC and its tributaries, especially the renal and gonadal veins, should be assessed in all abdominal malignancies to exclude thrombosis (Figure 15). Postsurgical venous thromboembolism is the leading cause of postoperative death in cancer patients, and IVC thrombosis is associated with substantial morbidity and mortality.[11] [12]

 

Table 2. Risk factors associated with increased incidence of thrombosis in patients with cancer. [7] [8] [9] [10] [11] [12]

 

Figure 14. (a) Coronal contrast-enhanced computed tomography (CECT) image of a 56-year-old man with adenocarcinoma of the stomach shows antropyloric gastric malignancy (stars), metastatic lymph nodes (white arrows), and multiple intraluminal tumour thrombi (black arrows) in the inferior vena cava (IVC) [red arrow], with extraluminal infiltration of the IVC by right iliac lymph nodes (yellow arrow). (b) Coronal and (c, d) axial CECT images of a 42-year-old woman with mucinous adenocarcinoma of the stomach show diffuse thickening of the gastric wall (stars in [b] and [c]) and widespread metastatic lymph nodes with multiple tiny calcifications (blue arrows in [b] and [c]). A focal intraluminal thrombus in the IVC (black arrows in [b] and [c]) and a long-segment thrombus in a dilated, non-enhancing right ovarian vein (green curved arrows in [b] and [d]) are evident. The left ovarian vein (yellow arrows in [b] and [d]) is compressed by retroperitoneal lymphadenopathy.

 

Figure 15. Two cases of ovarian cancer. (a) Coronal contrast-enhanced computed tomography (CECT) image of the abdomen of a 62-year-old patient shows right ovarian malignancy (star) with an intraluminal thrombus in the inferior vena cava (IVC) [yellow arrow] and extraluminal compression (green arrows) by enlarged lymph nodes (red arrows). (b, c) Coronal CECT images of the abdomen of a 53-year-old patient show a complex cystic lesion in the pelvis and left adnexa (stars), with bland thrombi (black arrows) in the infrahepatic IVC (white arrows) and the right renal vein (red arrows).

 

Tumour thrombus results either from direct extension of the malignancy or embolisation of neoplastic cells into the abdominal veins and/or the IVC. Differentiation between bland and tumour thrombi is crucial for management: anticoagulation or catheter-directed thrombolysis is the mainstay of treatment for bland thrombus, whereas tumour thrombus may require surgical resection (Table 3).[8] [13] [14] [15] In addition to tumour thrombectomy, adherent tumour thrombus invading the IVC wall necessitates en bloc excision, segmental resection, and vascular reconstruction.[15] Magnetic resonance imaging is superior to computed tomography in detecting and characterising tumour thrombus, as well as in identifying vessel wall invasion[8] (Figure 16). The extent of tumour thrombus within the IVC and the right atrium, along with vessel wall invasion, determines staging and resectability. These two factors are also independent predictors of adverse prognosis and poor survival rates in abdominal malignancies.[7] [8]

 

Table 3. Differentiating imaging features between tumour thrombus and bland thrombus.[8] [13] [14] [15]

 

Figure 16. Renal cell carcinoma (RCC) involving the inferior vena cava (IVC) in three patients. (a) Patient 1. Axial contrast-enhanced computed tomography (CECT) image shows left RCC (star) with an enhancing tumour thrombus in the dilated left renal vein (white arrow) and a bland thrombus in the IVC (black arrow). (b) Patient 2. Coronal CECT image shows intense heterogeneous enhancement of right upper-pole RCC (star), with an enhancing tumour thrombus in the right renal vein and IVC (black arrows), extending up to the right atrium. (c) Patient 3. Axial T2-weighted magnetic resonance image shows left RCC (star) with tumour thrombus in the IVC (white arrow) focally invading the IVC wall (red arrows).

 

Anatomical variants in the hepatic vasculature and the IVC should be identified before segmental resection in hepatoblastoma (Figure 17). Retroperitoneal malignancies in children may involve the abdominal vasculature, including the IVC (Figure 18). Thrombosis and vascular displacement are more common in Wilms tumours than vessel encasement, whereas vascular invasion occurs more frequently in neuroblastomas[4] (Figure 19).

 

Figure 17. A 3-year-old child with biopsy-proven hepatoblastoma. (a) Greyscale and (b) colour Doppler ultrasound images show a mixed echogenic lesion (star in [a]) in the liver with narrowing of the intrahepatic inferior vena cava (IVC) [white arrows in (b)]. (c) Coronal contrast-enhanced computed tomography image of the abdomen shows a heterogeneously enhancing hypoattenuating lesion (star) in the right lobe of the liver with pronounced luminal narrowing of the intrahepatic IVC (white arrow). Hepatomegaly with patchy heterogeneous parenchymal enhancement and architectural distortion is also noted.

 

Figure 18. Retroperitoneal rhabdomyosarcoma in a 4-year-old child. (a) Axial, (b) coronal, and (c) sagittal contrast-enhanced computed tomography images show a large heterogeneously enhancing soft tissue–attenuation mass (stars) partially encasing, compressing, and narrowing the inferior vena cava (white arrows), with infiltration of the vessel wall (red arrows in [a] and [b]).

 

Figure 19. (a) Axial contrast-enhanced computed tomography (CECT) and corresponding (b) positron emission tomography/computed tomography fusion images of a 5-year-old child with neuroblastoma of the right adrenal gland show a hypermetabolic enhancing lesion with non-enhancing necrotic areas in the right suprarenal region (stars), encasing and infiltrating the inferior vena cava (IVC) [white arrows]. (c) Axial and (d) coronal CECT images of the abdomen of a 7-year-old child with Wilms tumour of the right kidney show a large heterogeneously enhancing tumour in the right kidney (stars), with tumour thrombus in the right renal vein (black arrows). The lesion also compresses the right lobe of the liver and causes extraluminal compression of the IVC (white arrows).

 

Compression and narrowing of the IVC may occur as immediate or delayed complications in patients undergoing extensive retroperitoneal surgeries and abdominal lymph node dissections (Figure 20).

 

Figure 20. Immediate postsurgical complications on day 8 after para-aortic nodal dissection. (a) Axial and (b) sagittal contrast-enhanced computed tomography (CECT) images of the abdomen show a large hypodense collection (white arrows) causing extraluminal compression of the infrarenal inferior vena cava (IVC) [red arrow in (b)]. Delayed postsurgical complications in a case of carcinoma of the testis, status post-orchidectomy, and para-aortic lymphadenectomy demonstrate chronic IVC thrombosis with collateral formation. (c) Sagittal CECT image of the abdomen shows significant luminal narrowing of the IVC (white arrow). (d) Coronal maximum intensity projection image shows multiple dilated, tortuous collateral vessels in the pelvis and abdomen extending into the thoracic walls (white arrows), along with a contracted right kidney (black arrow).

 

Comprehensive evaluation of the IVC and its tributaries is a critical component of pre-surgical imaging. Cancer-associated thrombosis of the IVC and abdominal veins remains underrecognised and requires a high index of clinical suspicion due to non-specific symptoms. Identifying abnormal drainage patterns and congenital variations, along with recognising intrinsic or extrinsic involvement of the IVC by abdominopelvic malignancies, is vital before undertaking major oncological surgery.