Intrathoracic lymphadenopathy is a common diagnostic problem, and the tissue required to resolve it has conventionally been obtained from the mediastinum or lung itself. Transthoracic core needle biopsy under computed tomographic guidance is effective, achieving a definite histopathological diagnosis in around 90% of mediastinal masses in reported series.1 It is not, however, a trivial procedure. It requires traversing aerated lung or lying close to the great vessels, and a meta-analysis of CT-guided lung biopsy reported pneumothorax in 25.3% of core biopsies, with 5.6% requiring intervention, and pulmonary haemorrhage in 18.0%.2 It also exposes both patient and operator to ionising radiation, with mean patient surface doses reported between 18.7 and 34.9 mSv depending on whether CT fluoroscopy is used, and it commonly requires sedation.3 In centres without on-site thoracic surgical support, these considerations are not merely theoretical.

Many of these patients, however, do not have isolated intrathoracic disease. Lymphadenopathy is frequently generalised, and a cervical, supraclavicular or axillary node is often palpable or sonographically evident at the time of referral. Where such a node exists, the operator has a choice: sample the intrathoracic lesion that prompted the referral, or sample an accessible node that is presumed to reflect the same disease process. We refer to the latter as a surrogate-site strategy. The approach is not new — sonographically guided supraclavicular node biopsy was proposed two decades ago as a simple alternative to lung biopsy4 and has since been shown to yield tumour subtype and molecular results in patients

supraclavicular nodes in suspected lung cancer.6 What has not been established is how it performs when the comparison is made within a single operator's consecutive practice, and when the endpoint includes whether the tissue was sufficient for the ancillary work-up the diagnosis demanded.

The strategy is intuitively attractive. A superficial node can be biopsied under ultrasound at the bedside, without radiation, without sedation, without crossing pleura, and with direct compression available if bleeding occurs. It is also faster and cheaper. The objection to it is equally intuitive: the surrogate node may not represent the intrathoracic process, or may yield less tissue than a mass lesion, leaving the pathologist unable to complete the immunohistochemical work-up on which a modern lymphoma or metastatic diagnosis depends.7 That concern is the same one that underlies the continuing debate over whether image-guided core biopsy can replace excisional node biopsy in lymphoproliferative disease.8

That objection is testable, and it has two distinct parts. The first is representativeness: does the surrogate site harbour the same disease? The second is sufficiency: does it yield enough tissue to answer the question completely? A biopsy that returns a definite diagnostic category but leaves too little material for an immunopanel has not served the patient, even though conventional diagnostic-yield figures would count it as a success.

These considerations carry particular weight where histopathology is not available on site. In such settings the specimen leaves the hospital for processing, the operator has effectively one attempt, and an inadequate sample cannot be quietly supplemented. We therefore analysed a consecutive series of percutaneous biopsies performed for intrathoracic disease at a single tertiary care hospital, classified by where the tissue was actually taken, and compared representativeness and tissue sufficiency between surrogate and direct sampling.

Study design and population This was a retrospective observational analysis of consecutive image-guided biopsies performed in the interventional radiology suite of Lady Reading Hospital, Peshawar, a tertiary care teaching hospital, between December 2021 and October 2024. The source dataset comprised 272 specimen reports from 260 patients. All procedures were performed at a single institution, and 250 of the 272 specimens were obtained by a single interventional radiologist, so both the institutional pathway and operator technique are essentially constant across the series; within the analysed cohort, 84 of the 89 biopsies were performed by that operator. The study hospital does not maintain an on-site histopathology service equipped for the immunohistochemical work-up these specimens require. Specimens were therefore fixed in formalin at the point of collection and transported to one of two external referral laboratories — Shaukat Khanum Memorial Cancer Hospital, Lahore, and Shifa International Hospital, Islamabad — where grossing, tissue processing, reporting and all immunohistochemistry were performed. Allocation to a laboratory followed prevailing referral and funding arrangements rather than any clinical or radiological characteristic of the case. Biopsies were eligible for the present analysis if the clinical history accompanying the request documented intrathoracic disease — mediastinal, hilar, pulmonary, pleural or pericardial — and if tissue was obtained percutaneously. Ninety-two specimen records met the clinical criterion; three were excluded because the source report did not state the procedure performed and it could not be determined from the record, two of these being immunohistochemistry addenda. Eighty-nine biopsies remained. The unit of analysis is the biopsy rather than the patient. The 89 biopsies were performed in 86 patients: three patients each contributed two biopsies, in two instances an inconclusive first attempt followed by a successful repeat at the same target, and in one instance an initial diagnosis subsequently refined. Both members of each pair fell within the same sampling-strategy group, so no patient contributes to more than one arm. A patient-level sensitivity analysis is reported alongside the primary results. Specimen handling and processing Because all histopathology was performed off site, every specimen in this series underwent formalin fixation at the point of collection followed by road transport to a laboratory in another city. Turnaround from request to authorised report, which therefore incorporates transit as well as processing time, had a median of 7.0 days for specimens sent to Islamabad and 8.1 days for those sent to Lahore. Grossing protocols differed between the two laboratories: one recorded the number of cores received and their individual lengths in millimetres, together with whether the container was marked for anatomical site, while the other recorded neither. Procedural tissue metrics are therefore available for a subset of specimens, and this is stated wherever such a denominator is used.   Sampling-strategy groups Each biopsy was assigned to one of three groups on the basis of the anatomical site from which tissue was actually taken: Direct intrathoracic biopsy — mediastinum, lung, pleura, chest wall, pericardium, or a named pulmonary lobe. Reaching these targets requires crossing the thorax. Superficial surrogate — cervical, supraclavicular, retroclavicular, suprathyroid, axillary or inguinal. These are compressible, ultrasound-visible and accessible without crossing pleura. Deep non-pulmonary surrogate — liver, abdominopelvic (including retroperitoneal, mesenteric, para-aortic, adrenal and pelvic) or bone. These avoid the thorax but remain deep targets requiring cross-sectional guidance. The two surrogate categories were analysed separately and not pooled, on the prior reasoning that they differ materially in both accessibility and biological relationship to the intrathoracic process. Endpoints Conclusive diagnosis was defined as the issue of a definite diagnostic category. Reports returning an indeterminate atypical infiltrate, a non-diagnostic or inadequate specimen, an immunoprofile without a morphological diagnosis, or an explicit deferral pending further immunohistochemistry were counted as not conclusive. Tissue sufficiency was quantified as the number of distinct immunohistochemical markers reported for the specimen, extracted from the free-text immunohistochemistry field against a vocabulary of approximately 100 markers. Thresholds of three and five markers were pre-specified as indicating, respectively, a usable and a substantial panel, reflecting published minimum panels for lymphoma subtyping, which require five to seven markers for classic Hodgkin lymphoma and comparable numbers for the major non-Hodgkin categories.7 Biopsy fully served the patient — the composite primary endpoint — required a conclusive diagnosis and, for diagnoses that depend on immunophenotyping (lymphoma, metastatic and primary carcinoma, sarcoma, plasma cell and other malignant neoplasms), at least three markers. This composite is proposed because conventional diagnostic yield overstates procedural success: it credits a biopsy that names a category but leaves the oncologist without the subtyping needed to treat. The number of cores obtained and individual core lengths were extracted from the gross description where the reporting laboratory recorded them. Statistical analysis Because specimens were divided between two external laboratories, the processing laboratory was tested as a potential confounder of both the sampling-strategy distribution and the immunohistochemistry endpoint before the primary analysis was interpreted. Categorical outcomes were compared using Fisher's exact test with odds ratios; continuous and count variables using the Mann–Whitney U test; three-group comparisons using the Kruskal–Wallis test. Binomial proportions are reported with Wilson 95% confidence intervals. The pre-specified primary comparison was superficial surrogate versus direct intrathoracic biopsy; the deep non-pulmonary group is reported descriptively. Two-sided p values below 0.05 were considered statistically significant. No adjustment was made for multiple comparisons, and all analyses should be regarded as exploratory.

Cohort and sampling pattern

Of 89 percutaneous biopsies performed in 86 patients with documented intrathoracic disease, 57 (64.0%) sampled the thorax directly and 32 (36.0%) sampled elsewhere. Of the 32 surrogate biopsies, 20 (62.5%) were taken from a superficial ultrasound-accessible site and 12 (37.5%) from a deep non-pulmonary site. The cervical and supraclavicular region accounted for 19 of the 20 superficial procedures. The distribution of surrogate sites is given in Table 1.

Table 1. Extrathoracic sites sampled in patients whose clinical history documented intrathoracic disease.

The three groups were well matched for age (median 39, 36 and 37 years for superficial surrogate, deep surrogate and direct biopsy respectively; Kruskal–Wallis p = 0.85) and for sex (male 70.0%, 66.7% and 68.4%; p = 0.98). Baseline characteristics are shown in Table 2.

Table 2. Baseline characteristics by sampling strategy. Age Kruskal–Wallis p = 0.85; sex chi-square p = 0.98.

Effect of the processing laboratory

Fifty-four of the 89 biopsies were processed in Islamabad and 35 in Lahore. The processing laboratory was not associated with sampling strategy (chi-square p = 0.43), so surrogate and direct biopsies were distributed across the two laboratories in similar proportions. Nor did the laboratory affect the immunohistochemistry endpoint: the median number of markers reported was 4.0 at both, and the distributions did not differ (Mann–Whitney p = 0.83). Differences in laboratory practice are therefore unlikely to account for the comparisons that follow.

Representativeness of the surrogate site

The superficial surrogate site harboured malignant disease in 15 of 20 patients (75.0%), against 43 of 57 (75.4%) sampled directly — a difference of 0.4 percentage points (odds ratio 0.98, p = 1.000). Lymphoma specifically accounted for 40.0% of superficial surrogate diagnoses and 43.9% of direct diagnoses. On the two measures that determine whether a surrogate node reflects the intrathoracic process, the superficial surrogate was indistinguishable from the thorax itself.

The deep non-pulmonary surrogates behaved differently. Malignancy was found in only 7 of 12 (58.3%) and lymphoma in 3 of 12 (25.0%). Inspection of these cases shows why: liver, adrenal and retroperitoneal targets frequently represent a separate organ process rather than the same nodal disease, and three of the twelve returned a reactive or benign diagnosis. Full diagnostic distributions appear in Table 3.

Table 3. Final diagnosis by sampling strategy. Superficial surrogate versus direct biopsy for any

malignant diagnosis: odds ratio 0.98, p = 1.000.

Diagnostic yield

A conclusive diagnostic category was issued in 18 of 20 superficial surrogate biopsies (90.0%, 95% CI 69.9–97.2), 10 of 12 deep surrogate biopsies (83.3%, 95% CI 55.2–95.3) and 54 of 57 direct biopsies (94.7%, 95% CI 85.6–98.2). The difference between the superficial surrogate and direct sampling was not statistically significant (odds ratio 0.50, p = 0.60), and the confidence intervals overlap substantially.

The composite endpoint — a conclusive diagnosis with adequate tissue for the required work-up — was met in 85.0% of superficial surrogate, 83.3% of deep surrogate and 94.7% of direct biopsies (superficial versus direct, odds ratio 0.31, p = 0.18).

Two superficial surrogate biopsies were not conclusive. One returned an atypical lymphoid infiltrate with immunohistochemistry requested but not completed; the other described findings consistent with metastatic carcinoma but deferred final classification pending an additional panel. Three direct biopsies were not conclusive, two of them likewise atypical lymphoid infiltrates. In both arms, the failure mode was the lymphoid lesion rather than the anatomical route.

Tissue sufficiency

The superficial surrogate produced more material for ancillary testing than direct thoracic sampling, not less. The median number of immunohistochemical markers reported was 5.5 (IQR 2.8–7.5) for superficial surrogate specimens against 4.0 (IQR 2.0–5.0) for direct specimens (Mann–Whitney p = 0.096). At the pre-specified five-marker threshold the difference was statistically significant: 13 of 20 superficial surrogate specimens (65.0%) supported a panel of five or more markers, compared with 21 of 57 direct specimens (36.8%) — odds ratio 3.18, p = 0.038.

The mechanism is apparent in the procedural data. A median of 3.0 cores was obtained from superficial targets against 2.0 from intrathoracic targets (p = 0.013). Where the target is compressible, sonographically visible in real time, and carries no risk of pneumothorax, the operator can afford additional passes; that additional tissue is what supports the larger panel. Deep non-pulmonary surrogates, which share the safety disadvantage of a deep target, showed neither the additional cores (median 2.0) nor the larger panels (25.0% reaching five markers).

Table 4. Diagnostic and tissue-sufficiency endpoints by sampling strategy.

Table 5. Primary comparisons. Odds ratios from Fisher's exact test; count comparisons by Mann–Whitney U test.

Patient-level sensitivity analysis

Because three patients each contributed two biopsies, the analysis was repeated with one row per patient (n = 86), retaining the conclusive record where a patient had both. Group sizes became 19 superficial surrogate, 12 deep non-pulmonary surrogate and 55 direct. Every conclusion was unchanged. Malignancy remained equivalent between the superficial surrogate and direct sampling (73.7% versus 76.4%; odds ratio 0.87, p = 1.000). The immunohistochemical advantage persisted and strengthened slightly: 63.2% of superficial surrogate specimens reached five or more markers against 34.5% of direct specimens (odds ratio 3.25, p = 0.035). The core-count difference was identical (median 3.0 versus 2.0, p = 0.013).

The one measure that moved was diagnostic yield, and it moved in favour of the surrogate. Conclusive diagnosis rose to 94.7% (95% CI 75.4–99.1) for the superficial surrogate against 96.4% (95% CI 87.7–99.0) for direct sampling (odds ratio 0.68, p = 1.00), narrowing the gap from 4.7 to 1.7 percentage points. This is expected: at biopsy level a failed first attempt and its successful repeat are counted separately, whereas at patient level they resolve to a single patient in whom the question was answered.

Lymphoma subtyping

Eight lymphomas were diagnosed from superficial surrogate specimens and 25 from direct specimens. A named World Health Organization subtype⁹ was reached in 5 of 8 (62.5%) surrogate cases and 20 of 25 (80.0%) direct cases. Subtypes obtained from surrogate nodes included nodular sclerosis and lymphocyte-rich classic Hodgkin lymphoma, T-lymphoblastic lymphoma and mantle cell lymphoma. The numerically lower subtyping rate in the surrogate arm rests on eight cases and three events, and should not be over-interpreted; it is nonetheless the one endpoint on which the surrogate arm did not equal or exceed direct sampling, and it warrants specific attention in any larger series.

The central question for an interventional radiologist offered a patient with mediastinal lymphadenopathy and a palpable neck node is not whether transthoracic biopsy works — it plainly does — but whether anything is lost by taking the easier target. In this series, on the measures that matter, nothing was. Representativeness was effectively identical. The proportion of malignant diagnoses differed by 0.4 percentage points between the superficial surrogate and the thorax, and the lymphoma fraction by less than four points. This is the finding on which the entire strategy depends, and it is reassuring precisely because it is unremarkable: in a population where intrathoracic lymphadenopathy is usually accompanied by generalised nodal disease, an accessible node carries the same process. It is also consistent with earlier supraclavicular series, in which malignancy was confirmed in the great majority of sampled nodes in patients referred with suspected thoracic malignancy.4,5 Diagnostic yield was numerically lower from the surrogate site — 90.0% against 94.7% — and this difference was neither statistically significant nor, in our view, clinically meaningful at the observed magnitude. It corresponds to one additional inconclusive result per twenty procedures, set against the avoidance of twenty transthoracic punctures. Our 90.0% figure is close to the 90% overall accuracy reported for supraclavicular tissue core biopsy in suspected thoracic malignancy5 and sits within the range described in a recent meta-analysis of ultrasound-guided core needle biopsy for lymphoma.10 It must be stressed that a non-significant difference in a sample of this size is not evidence of equivalence, and we make no claim of non-inferiority. The unexpected finding is that the surrogate site yielded more tissue for ancillary testing, not less. Nearly two-thirds of superficial surrogate specimens supported a panel of five or more immunostains, against just over a third of direct specimens, and the odds ratio of 3.18 was statistically significant. The mechanism appears straightforward and is supported by the procedural data: operators took a median of three cores from superficial targets and two from intrathoracic ones. A target that is compressible, continuously visible under ultrasound, and carries no pneumothorax risk permits an additional pass at negligible marginal cost. This accords with work showing that the diagnostic accuracy of core needle biopsy in lymphoproliferative disease is driven by the amount of tissue retrieved, and that six or more passes are feasible specifically at superficial, easily accessible targets.11 In an era where the diagnosis of lymphoma and of carcinoma of unknown primary depends less on morphology than on the completeness of the immunopanel,7 this is not a minor advantage. Two features of the setting strengthen these observations. All procedures were performed at one hospital by predominantly one operator, so variation in technique, equipment and patient pathway is minimal. And because histopathology was outsourced to two independent external laboratories that were not associated with either the choice of sampling site or the immunohistochemical yield, the tissue-sufficiency finding is unlikely to be an artefact of a single laboratory's reporting habits. The separation between the two surrogate tiers is instructive and, we would argue, is the practical lesson of this analysis. Deep non-pulmonary surrogates — liver, adrenal, retroperitoneum — shared none of the superficial group's advantages. They were malignant in only 58.3% of cases and supported a five-marker panel in only 25%. Choosing a deep abdominal target merely to avoid the chest sacrifices representativeness without gaining accessibility. The strategy that works is specifically the superficial one, and these tiers should not be pooled in future analyses. The diagnostic mix we observed also reflects the setting. Granulomatous disease accounted for one in eight of the biopsies in this cohort, consistent with the high burden of tuberculous lymphadenitis in Pakistan, where tuberculosis accounts for roughly half of superficial lymphadenopathy referred for sampling.12 In such a population an accessible node is as likely to answer the question as the intrathoracic lesion that prompted the referral, which strengthens rather than weakens the case for sampling it. Limitations The principal limitation is selection bias, and it is not remediable here: the surrogate site was chosen precisely when an accessible, sonographically abnormal node was present, so the groups are not exchangeable and the comparison describes a real-world selection process rather than a causal effect. Second, there is no reference standard. Patients whose surrogate biopsy was diagnostic did not undergo thoracic sampling, so discordance could not be measured; we report diagnostic yield, not accuracy. Third, the unit of analysis is the biopsy rather than the patient, and three patients contributed two biopsies each; the patient-level sensitivity analysis shows this does not alter the findings. Most important for the clinical argument, no procedural safety data were available. The rationale rests on avoided pneumothorax, radiation and sedation, none of which is recorded in a pathology dataset. The study shows the strategy does not cost diagnostic performance; it cannot quantify what it saves. Twenty superficial procedures give wide confidence intervals and cannot support a non-inferiority claim, which at a 15-percentage-point margin would require approximately 150 patients per arm. Histopathology was performed off site, adding pre-analytical variables — transit, temperature and handling — that were neither captured nor controlled. Processing laboratory was associated with neither sampling strategy nor immunohistochemical yield, but only one laboratory recorded core counts and labelling, so tissue metrics rest on a subset. Finally, endpoints were derived by structured extraction from free-text reports: marker counts reflect what was transcribed rather than what was performed, and reporting pathologists were not blinded to sampling site.

In patients presenting with intrathoracic lymphadenopathy who also have an accessible superficial node, biopsy of that surrogate site was not a compromise. It detected malignancy at the same rate as direct intrathoracic sampling, to within half a percentage point, and returned the same spectrum of diagnoses. It produced a conclusive diagnostic category in nine of ten patients, a rate not statistically distinguishable from direct biopsy. And it supported a significantly larger immunohistochemical panel, with nearly twice the proportion of specimens reaching five or more markers — the material on which contemporary lymphoma and carcinoma classification actually depends.

That last point deserves emphasis, because it inverts the usual objection. The concern about surrogate sampling has always been that a small node yields too little tissue. In practice the opposite occurred: because a superficial node can be approached safely and repeatedly under real-time ultrasound, operators took more cores, and more cores translated into more complete immunophenotyping.

The safety of the target and the sufficiency of the specimen are not in tension; the first enables the second.

We therefore consider the surrogate-site strategy defensible as a default rather than as an expedient. Where a patient with intrathoracic lymphadenopathy has an accessible cervical, supraclavicular or axillary node, we would sample that node first, take three or more cores while the safety of the target permits it, and reserve transthoracic biopsy for patients in whom no such surrogate exists or in whom the superficial specimen proves insufficient. This sequencing avoids a transpleural puncture and its attendant radiation exposure in a substantial minority of patients — 36% of our intrathoracic referrals were sampled off the chest — without any demonstrated cost in diagnostic performance.

The setting sharpens rather than weakens this argument. Where histopathology is outsourced and specimens travel to another city before they are processed, the operator gets one attempt and cannot easily supplement an inadequate sample. Under those conditions the case for taking the target that permits an extra core, rather than the one that discourages it, is stronger still.

Two qualifications bound this recommendation. It applies to superficial targets only: deep non-pulmonary surrogates performed materially worse on both representativeness and tissue sufficiency, and substituting an abdominal target for a thoracic one cannot be justified on this evidence. And it is a recommendation grounded in a modest, non-randomised series subject to selection bias; it establishes that the strategy is reasonable, not that it is superior. What would settle the question is a prospective series capturing complications and radiation dose alongside diagnostic outcome — at which point the argument for the easier target would rest not only on equivalent tissue, but on the harm it avoids.

Declarations

Funding: none. Conflicts of interest: none declared. Ethical approval: Institutional Review Board, Lady Reading Hospital, Peshawar . Informed consent: waived for this retrospective analysis of anonymised records.