Paraneoplastic syndromes and cancer of unknown primary

• A paraneoplastic syndrome is an effect of a cancer on a distant organ that the tumour does not invade and that is not explained by metastasis, infection, or side effects of treatment.
• The mechanisms are usually one of two things: the tumour secretes a hormone, peptide, or cytokine, or the immune response to the tumour cross-reacts with normal tissue.
• Syndromes are grouped by the system they disturb: endocrine (salt and water, calcium, cortisol, glucose), neurological (brain, cerebellum, spinal cord, nerves, muscle), dermatological (skin and mucous membranes), and haematological (blood cells and clotting).
• They affect up to 8% of cancer patients ^[1] and are sometimes the first clue to an otherwise silent cancer.
• Cancer of unknown primary (CUP) is a biopsy-proven carcinoma or undifferentiated tumour for which a standard work-up cannot find the primary site ^[5]. About 20% of patients have a "favourable" subset with a recognisable, site-specific treatment plan ^[6].

By the end of this page you should be able to:

1. Define a paraneoplastic syndrome and distinguish its two main mechanisms.
2. Match the common endocrine, neurological, dermatological, and haematological syndromes to their usual tumours and, where relevant, to their antibodies.
3. Recognise the neurological "high-risk phenotypes" and explain how tumour screening is planned and repeated.
4. Outline the standard work-up of a metastasis with an unknown primary and name its favourable subsets.
5. Apply the ESMO CUP guideline of 2023, the NICE NG12 referral framework, and the EFNS/EAN screening guidance for paraneoplastic neurological syndromes, stating the edition used.
6. Plan low-resource and special-group management.

The defining feature is distance from the tumour. A brain metastasis from lung cancer is not paraneoplastic; Lambert-Eaton myasthenic syndrome in the same patient is. This distinction matters because it changes the search for cause and the treatment plan. Every syndrome on this page is "paraneoplastic" in the sense that its cause is the cancer, not the cancer's local effects.

Paraneoplastic effects are grouped by the system they disturb. This page covers four groups: endocrine, neurological, dermatological, and haematological. There are overlapping pictures such as hypertrophic osteoarthropathy (digital clubbing) that sit beside the dermatological syndromes ^[1].

Paraneoplastic syndromes are uncommon overall but not rare in oncology practice. They affect up to 8% of cancer patients ^[1]. Paraneoplastic neurological syndromes (PNS) develop in roughly 1 in 300 people with cancer, with population-based incidence between 1.6 and 8.9 per million person-years, and are probably underdiagnosed ^[3].

The cancers most often associated overall are small cell lung cancer, breast cancer, gynaecological tumours, and haematological malignancies ^[1]. Single syndromes have their own tumour affinities, set out below. The most useful bedside habit is to ask, when a paraneoplastic picture is new, "which tumour does this point to, and is it already present?" ^[4]

Two mechanisms explain almost all syndromes.

Secretion. Endocrine and some haematological syndromes occur because the tumour behaves like an endocrine or blood-cell factory. A small cell lung cancer makes antidiuretic hormone; a squamous lung cancer makes parathyroid hormone-related protein (PTHrP). These substances reach the bloodstream and act on distant targets exactly as the normal hormone would. ^[6]

Immune cross-reactivity. Neurological and several dermatological syndromes occur because the cancer expresses an antigen that resembles an antigen in normal tissue. The patient raises antibodies and T-cells against the tumour; these also attack the nervous system or skin. This is why PNS often improve poorly after cancer treatment: the immune damage can be irreversible ^[1].

Mixed and unknown mechanisms exist. For example, hypertrophic osteoarthropathy (digital clubbing) is linked to vascular endothelial growth factor, platelet-derived growth factor, and prostaglandin E2 ^[1].

Endocrine syndromes are usually detected after the cancer is known. They can, however, be the first manifestation, and they often resolve when the tumour responds to treatment. ^[2]

SIADH is characterised by hypo-osmotic, euvolaemic hyponatraemia. Tumour cells make antidiuretic hormone (arginine vasopressin) and atrial natriuretic peptide, which increase free-water reabsorption. In a euvolaemic patient, a urinary sodium level above 40 mmol/L or a urine osmolality above 100 mOsm/kg supports the diagnosis ^[1]. In contrast, a volume-depleted patient with high urine sodium has appropriately, not inappropriately, raised hormone.

Severity tracks the speed of fall. Serum sodium below 125 mmol/L (125 mEq/L), particularly if it develops within 48 hours, can produce seizures, coma, and respiratory collapse ^[1]. Treatment must respect that speed: symptomatic hyponatraemia developing within 48 hours can be raised by 1-2 mmol/L per hour and usually no more than 8-10 mmol/L in the first 24 hours, whereas chronic hyponatraemia is corrected at only 0.5-1.0 mmol/L per hour to avoid osmotic demyelination ^[1]. Definitive therapy is treatment of the cancer. In the short term use fluid restriction (usually under 1000 mL per day), stop offending medicines, and correct sodium with hypertonic saline in severe cases; demeclocycline and vasopressin receptor antagonists (conivaptan, tolvaptan) are pharmacological options ^[1].

Malignancy-associated hypercalcaemia complicates up to 10% of advanced cancers and carries a 30-day mortality of about 50% ^[1]. Four mechanisms exist, and knowing them helps the work-up:

1. Humoral hypercalcaemia of malignancy, from tumour PTHrP: about 80% of cases, most common with squamous cell tumours. ^[1]
2. Local osteolysis at skeletal metastases: about 20%, commonly breast cancer, multiple myeloma, and lymphoma.
3. Tumour calcitriol (vitamin D) production, described in some lymphomas.
4. Ectopic parathyroid hormone production, rare.

Calcium above 3.5 mmol/L (about 14 mg/dL) counts as severe, but symptoms also depend on speed of onset. Laboratory clues: high ionised calcium, low or normal parathyroid hormone (normal range 10-55 pg/mL), and often high PTHrP (normal under 2.0 pmol/L) ^[1]. Corrected calcium (mg/dL) = measured calcium (mg/dL) plus [0.8 x (4.0 - albumin in g/dL)] ^[1].

Management is to treat the cancer and rehydrate with normal saline, which raises glomerular filtration rate and inhibits renal calcium reabsorption; loop diuretics follow only after rehydration. Intravenous bisphosphonates (pamidronate, zoledronate) are first-line pharmacotherapy for persistent hypercalcaemia, with the calcium fall beginning in 2-4 days and lasting up to 3 weeks; calcitonin, mithramycin, gallium nitrate, and dialysis are alternatives ^[1].

About 5-10% of all Cushing syndrome is paraneoplastic, and about 50-60% of those cases come from neuroendocrine lung tumours: small cell lung cancer and bronchial carcinoids ^[1]. Tumour adrenocorticotropic hormone (ACTH) or corticotropin-releasing factor drives cortisol production in the adrenal glands, so the picture is hypertension, hypokalaemia, muscle weakness, and oedema rather than the central obesity and striae that dominate non-paraneoplastic disease ^[1]. Laboratory guides: baseline serum cortisol above 29 ug/dL, urinary free cortisol above 47 ug per 24 hours, midnight ACTH above 100 ng/L ^[1]. Failure of high-dose dexamethasone suppression (2 mg by mouth every 6 hours for 72 hours, with a 17-hydroxycorticosteroid fall of at least 50%) points to an ectopic source ^[1]. Imaging with CT, MRI, or somatostatin receptor scintigraphy then hunts the primary; inferior petrosal sinus sampling is generally unnecessary ^[1]. Drugs that block steroid production (ketoconazole, mitotane, metyrapone, aminoglutethimide, octreotide, etomidate, mifepristone) and adrenalectomy if they fail sit alongside cancer therapy ^[1].

This is rare, usually in older patients with advanced tumours. Glucose can fall to 1.1 mmol/L (20 mg/dL). Insulin and C-peptide are low (insulin often under 1.44-3.60 uIU/mL; C-peptide often under 0.3 ng/mL), growth hormone and IGF-1 are low, IGF-2 is normal or raised, and the IGF-2 to IGF-1 ratio is raised ^[1]. Treat the tumour if possible; otherwise maintain glucose with oral and parenteral dextrose, and use corticosteroids, growth hormone, diazoxide, octreotide (after a short-acting test dose), or glucagon ^[1].

These are immune-mediated. There is often no cancer at the time the neurological symptoms begin: PNS are detected before the cancer is diagnosed in 80% of cases ^[1]. Small cell lung cancer and lymphoma or myeloma are overrepresented, because they produce neuroendocrine proteins, involve immunoregulatory organs, or affect antibody production ^[1].

The 2021 PNS-Care criteria replaced the older term "classical PNS" with "high-risk phenotypes" and added "intermediate-risk phenotypes" and a PNS-Care score ^[3]. The high-risk phenotypes are: encephalomyelitis, limbic encephalitis, rapidly progressive cerebellar syndrome, opsoclonus-myoclonus syndrome, sensory neuronopathy, gastrointestinal pseudo-obstruction (enteric neuropathy), and Lambert-Eaton myasthenic syndrome ^[3]. These carry a strong association with cancer in most patients and should lead to a systematic tumour search ^[3].

It is worth noting that some of these syndromes also occur outside cancer. More than 70% of limbic encephalitis and subacute sensory neuropathy, and about 40% of LEMS, are not paraneoplastic, and about 50% of subacute cerebellar ataxia is not ^[1]. The cause is never assumed from the name alone.

The table of antibody-tumour pairs is the starting point for the cancer search. Antibodies are grouped by how often they accompany cancer: high-risk (more than 70% of cases), intermediate-risk (30-70%), and lower-risk (under 30%) ^[3]. The exact antigen depends on the laboratory method, so results should be read with the clinical picture.

Some clinically useful antibody hints from the 2021 criteria: more than 85% of patients with Hu antibodies have small cell lung cancer; more than 50% with AMPAR and at least 50% with GABABR antibodies have cancer; and NMDAR antibodies in women aged 18-35 correspond to an ovarian teratoma in 35-50% of cases ^[3].

Screening is directed by the antibody and the syndrome, not by guesswork. The EFNS Task Force recommends (good practice points):

• Thoracic cancers: CT of the chest, and if negative, FDG-PET or integrated PET-CT.
• Breast cancer: mammography, then breast MRI, then FDG-PET/CT if both are negative.
• Pelvic cancers (teratoma, carcinoma): transvaginal ultrasound (ultrasound of the testes in men), followed by CT or MRI; CT-thorax may be needed if pelvic imaging is negative.
• Dermatomyositis: CT of chest and abdomen, plus (in women) pelvic ultrasound and mammography, (in men under 50) testicular ultrasound, and (in everyone over 50) colonoscopy ^[4].

Sensitivity numbers guide humility: CT of the chest detected only about 80-85% of small cell lung cancers in paraneoplastic series, so a negative first scan does not exclude cancer ^[4].

If cancer is not found at first, do not reassure the patient too soon. In high-risk phenotypes with high- or intermediate-risk antibodies, repeat screening every 4-6 months for 2 years ^[3]. Older EFNS guidance is slightly different and can be followed where it is the local standard: repeat screening after 3-6 months, then every 6 months for up to 4 years in patients with paraneoplastic antibodies, but only 2 years in LEMS ^[4]. Both sources agree on the principle: repeated, targeted searching for the first two years.

Dermatological syndromes are often visible long before the cancer. They are divided into "obligate" (always linked to cancer) and "facultative" (often linked, but also seen without cancer) ^[2]. Useful clinical pointers follow.

Points that prevent mistakes:

• Most acanthosis nigricans is benign and linked to insulin resistance or obesity. Malignant change is sudden, severe, often generalised, and may involve mucosal surfaces; gastric adenocarcinoma is the most commonly associated cancer ^[1][2]. Up to 90% of "tripe palms" are cancer-associated ^[1].
• Dermatomyositis linked to cancer is treated with corticosteroids and steroid-sparing agents, but up to one third of patients keep significant muscle weakness ^[1]. When new lesions appear or an existing lesion returns, think recurrence before calling the treatment a failure ^[2].
• Paraneoplastic pemphigus (PAMS) is now called paraneoplastic autoimmune multiorgan syndrome. It is an obligate paraneoplastic disease: non-Hodgkin lymphoma is the most common trigger, followed by chronic lymphocytic leukaemia and Castleman disease, with adenocarcinoma or squamous carcinoma in about 8% and sarcoma in about 6% of cases. Presentation of PAMS precedes the cancer in up to 50% of cases. Bronchopulmonary disease (bronchiolitis obliterans) affects up to 90% and is a leading cause of death ^[2].
• Sweet syndrome and pyoderma gangrenosum are neutrophilic dermatoses. About 20% of Sweet syndrome is paraneoplastic, most often with acute myeloid leukaemia; first-line treatment is the same as for non-paraneoplastic disease, with systemic corticosteroids ^[2].

These are usually silent, found after the cancer diagnosis and in advanced disease, rarely need a treatment of their own, and may improve when the tumour responds ^[1].

Granulocytosis rarely needs treatment of its own: mature, deformable neutrophils do not cause leukostasis below about 250 x10^9/L, so leukapheresis is not required for the paraneoplastic form ^[1]. Leukocytosis in cancer patients is more often due to growth factor use (69%) or infection (15%) than to a paraneoplastic mechanism (10%), so exclude the common causes first ^[1].

A practical point: Trousseau syndrome is the most clinically important haematological paraneoplastic picture because it announces an occult adenocarcinoma, especially of the pancreas. Migratory superficial thrombophlebitis that is resistant to anticoagulation should prompt a cancer search ^[1].

The differential is wide and must be cleared before calling a syndrome paraneoplastic, particularly for PNS, where the more mundane diagnoses are more common.

• For hyponatraemia: hypovolaemic causes (gut losses, diuresis, adrenal insufficiency, salt wasting, cerebral salt wasting) rather than SIADH ^[1].
• For neurological presentations: brain metastases, leptomeningeal spread, cord compression, treatment toxicity (platinum, taxane, vinca alkaloid neurotoxicity, radiation), infection, metabolic disturbance, and the non-paraneoplastic autoimmune diseases that mimic them (for example anti-LGI1 encephalitis usually has no cancer) ^[1][3].
• For rashes: insulin-resistance acanthosis, viral exanthems, seborrhoeic dermatitis, drug eruptions, contact dermatitis, psoriasis, and (for PAMS) lichenoid drug eruptions and toxic epidermal necrolysis ^[2].
• For haematological findings: infection, myeloproliferative disease, clotting disorders, and (for Trousseau syndrome) inherited thrombophilia.

The approach is consistent across systems:

1. Characterise the syndrome (clinical pattern, severity).
2. Screen for the tumour guided by the syndrome and the antibody (see above).
3. Check the body the syndrome affects (electrolytes for endocrine, brain/cord imaging and cerebrospinal fluid for PNS, skin biopsy for dermatological, full blood count and clotting studies for haematological).
4. Look for the mechanism (serum sodium, urine osmolality and sodium for SIADH; calcium and PTH/PTHrP for hypercalcaemia; ACTH/cortisol for Cushing; autoantibody panels for PNS; histology and direct immunofluorescence for skin).

For PNS specifically: neuronal antibody testing should use the gold standard methods - tissue immunohistochemistry or immunofluorescence plus confirmatory immunoblot for intracellular antigens and cell-based assay for surface antigens. Indiscriminate testing lowers the value of a positive result, and about 30% of people with a presumed PNS have no detectable antibody ^[1]. The 2021 criteria ask that testing be restricted to high- or intermediate-risk phenotypes. If both serum and CSF are available, test both, because sensitivity differs by antibody (for example LGI1 is more reliably detected in serum) ^[3].

When antibody re-screening or repeated tumour search is negative but the clinical picture is high-risk, keep surveilling rather than dropping the hypothesis.

CUP is a carcinoma or undifferentiated neoplasm for which a standardised diagnostic work-up fails to identify the primary tumour responsible for the metastatic seeding ^[5]. It is not a single disease; it is a label for a heterogeneous group. CUP accounts for approximately 2-5% of cancer diagnoses globally and remains a leading cause of cancer death ^[6]. It occurs most often after age 60, slightly more in men ^[6]. Approximately 20% of patients have a more favourable prognosis (termed favourable CUP) ^[6], and between 10-15% have a favourable CUP subtype based on clinical and histopathological similarity to known primaries ^[6]. The remainder, roughly 80-85%, fall into the "unfavourable" group with poor response to empiric chemotherapy and overall survival around 6-12 months (median about 9 months on platinum-based regimens) ^[6]. ^[2]

The work-up aims to find the subgroups that benefit from targeted or site-specific therapy without trapping patients in endless and expensive testing.

Serum tumour markers, with a few exceptions, are not diagnostic or prognostic and should not be measured routinely: they include PSA in men with bone metastases, AFP and beta-HCG in men with midline disease, CA-125 in women with peritoneal disease, and chromogranin A when a neuroendocrine tumour is suspected ^[6].

Between 10-15% (and roughly 20% overall with a more favourable prognosis) of patients with CUP have a favourable subset, because their disease behaves like a known primary and should be treated as such ^[6]. The subsets, as aligned across ESMO 2023 and the SEOM 2025 update, are:

Clinically this is the reason exhaustive testing is worth doing: a CK7/CK20 pattern plus markers can move a patient from "poor prognosis CUP" into a treatable disease bracket. IHC on its own is only suggestive and should be combined with clinical and molecular information ^[6].

Two points on what is no longer called favourable CUP in ESMO 2023: poorly differentiated midline carcinoma in young men (many were extragonadal germ cell tumours) and neuroendocrine carcinomas of unknown primary (these are now classified and treated as neuroendocrine tumours) ^[6].

For the majority (80-85%) without a favourable or actionable profile, empiric platinum-based doublet chemotherapy (cisplatin or carboplatin with taxane, gemcitabine, or irinotecan) is standard, with objective response rates of 15-20% and median overall survival of about 9 months ^[6]. Fitness matters: patients with ECOG performance status 0-2 are candidates for chemotherapy and molecularly guided therapy, while those with performance status above 2 are usually managed with best supportive care ^[6].

Molecular profiling has changed the outlook. The CUPISCO trial (ESMO Congress 2023) randomised over 600 patients with unfavourable non-squamous CUP: molecularly guided therapy achieved a median progression-free survival of 6.1 months versus 4.4 months with platinum-based chemotherapy, hazard ratio 0.72 ^[5]. TMB-high (above 10 mutations per megabase), MSI-high, NTRK/RET/BRAF/FGFR/ALK alterations and other targets open the way to tumour-agnostic therapy ^[6]. NGS is therefore increasingly performed routinely in CUP ^[6]. Genomic profiling should not replace clinical reasoning, and the response to a drug targeting a mutation can depend on the tumour type in which it is found.

• ESMO Clinical Practice Guideline: cancers of unknown primary site, Krämer et al., Annals of Oncology 2023;34:228-246 ^[5], key edition for diagnosis, risk assessment, classification, treatment, and follow-up.
• Losa et al., SEOM-GECOD clinical guideline for CUP (update 2025), Clinical and Translational Oncology 2026;28:1610-1623 ^[6], used for the work-up table, favourable-subset list, morphology percentages, and molecular profiling numbers; this update aligns its favourable-subset list with ESMO ^[6].
• NICE guideline NG12 "Suspected cancer: recognition and referral" ^[7], published 23 June 2015 and last updated 15 April 2026.
• EFNS Task Force guideline on screening for tumours in paraneoplastic syndromes, Titulaer et al., European Journal of Neurology 2011 ^[4], successor to the 2006 EFNS management guideline, now used by the European Academy of Neurology (EAN).
• PNS diagnostic criteria: Graus et al. 2021 PNS-Care ^[3]; paraneoplastic syndrome overview: Pelosof and Gerber 2010 ^[1]; dermatological review: Didona et al. 2025 ^[2].

• Children. Neuroblastoma is a classic cause of opsoclonus-myoclonus; anti-NMDAR encephalitis is usually non-cancer in children. Dermatomyositis in children warrants examination for splenomegaly and lymphadenopathy ^[4].
• Pregnancy. Dermatomyositis and PNS are rare; manage with obstetric, oncology, and (for PNS) neurology teams. For CUP, staging and treatment depend on gestation and maternal condition.
• Older adults. CUP incidence rises after 60. PNS can be missed because neurological change is attributed to age or treatment toxicity.
• Immunocompromised. Immune checkpoint inhibitors can itself produce paraneoplastic-like neurological immune-related adverse events in 1-3% of patients, most often graded severe ^[3]. Patients with a known PNS are at higher risk of neurological worsening on checkpoint inhibitors ^[3].

The single most useful habit in a low-resource setting is to consider paraneoplastic syndromes at all. Many first encounters happen where the definitive test or tumour-directed treatment is far away. Practical steps: document the syndromes in detail and the cancer family history; treat the immediate danger (hypertonic saline for severe hyponatraemia, rehydration and bisphosphonate for hypercalcaemia, corticosteroids for dermatomyositis) before waiting for staging; sample and send blood for a paraneoplastic antibody panel if the transport chain exists; request a core biopsy and CK7/CK20 immunohistochemistry for suspected CUP, because this costs far less than advanced imaging and is often enough to place the patient in a favourable or unfavourable group. Where PET is unavailable, CT of thorax, abdomen, and pelvis carries most of the diagnostic yield; where CT is unavailable, careful clinical staging and referral are the safer course.

• New, progressive, or multifocal neurological disease without a clear cause, in anyone with cancer or in a smoker.
• Unexplained hyponatraemia, hypercalcaemia, or recurrent hypoglycaemia.
• Rapidly evolving skin disease in an older adult, especially sudden seborrhoeic keratoses, dermatomyositis, or tripe palms.
• Migratory superficial thrombophlebitis resistant to anticoagulation.
• A metastasis with no obvious primary, or with an IHC pattern that suggests a treatable breast, prostate, colon, or renal primary.

• Small cell lung cancer and lung cancer recognition
• Breast cancer
• Ovarian cancer
• Dermatomyositis
• SIADH and hyponatraemia
• Hypercalcaemia of malignancy
• Lambert-Eaton myasthenic syndrome
• Carcinoma of unknown primary work-up