Revisiting anti-Hu paraneoplastic autoimmunity: phenotypic characterization and cancer diagnosis

Abstract Anti-Hu are the most frequent antibodies in paraneoplastic neurological syndromes, mainly associated with an often limited stage small cell lung cancer. The clinical presentation is pleomorphic, frequently multifocal. Although the predominant phenotypes are well characterized, how different neurological syndromes associate is unclear. Likewise, no specific study assessed the performance of new-generation CT and PET scanners for cancer screening in these patients. Herein, we aimed to describe the clinical pattern and cancer screening in a retrospective cohort of 466 patients with anti-Hu autoimmunity from the French Reference Centre on Paraneoplastic Neurological Syndromes registry. Clinical presentation, cancer screening and diagnosis were analysed. Among the 466 patients, 220 (54%) had multifocal neurological involvement. A hierarchical cluster analysis grouped the patients into (i) mainly limbic encephalitis, (ii) predominantly peripheral neuropathy and (iii) broad involvement of the nervous system (mixed group). Compared with limbic encephalitis and mixed groups, patients in the neuropathy group more frequently had a chronic onset of symptoms (29 versus 13 and 17%), elevated CSF proteins (83 versus 47 and 67%) and died from cancer progression (67 versus 15 and 28%; all P < 0.05). No significant difference in overall survival was observed between groups. Dysautonomia and brainstem signs were associated with a higher risk of death from the neurological cause; cancer diagnosis was the main predictor of all-cause death, especially when diagnosed within 2 years from clinical onset (all P < 0.05). Three hundred and forty-nine (75%) patients had cancer: in 295 (84%) neurological symptoms preceded tumour diagnosis, being lung cancer in 262 (89%), thereof small cell lung cancer in 227 (87%). First CT scan revealed lung cancer in 205/241 (85%), and PET scan shortened the interval to diagnosis when the initial CT scan was negative [7 months (1–66) in 27 patients versus 14 months (2–45) in 6; P < 0.001]. Although cancer diagnosis mostly occurred within 2 years from clinical onset, 13/295 (4%) patients exceeded that threshold. Conversely, 33 patients (7%) were ‘cancer-free’ after 2 years of follow-up. However, 13/33 (39%) had initial suspicious imaging findings that spontaneously regressed. In conclusion, although anti-Hu autoimmunity clinical presentation is mostly multifocal, we observed patients with a predominant limbic syndrome or isolated sensory neuropathy. Early implementation of PET scan shortens the interval to cancer diagnosis, which was the strongest predictor of death, especially if diagnosed ≤2 years from clinical onset. As cancer was diagnosed >2 years after clinical onset in few patients, screening should be extended up to 5 years. In addition, tumour regression was suspected in a substantial proportion of ‘cancer-free’ patients.


Introduction
Anti-Hu are the most frequent antibodies in PNS, 1,2 and are mainly associated with an underlying small cell lung cancer (SCLC). 3Although well known for more than 30 years, 4 the rarity and pleomorphic clinical presentation of anti-Hu PNS, ranging from peripheral nervous system (e.g.sensory neuropathy or gastroparesis) to CNS (e.g.limbic encephalitis or cerebellar ataxia) involvement, still pose a diagnostic challenge.Early studies found a common multifocal nervous system involvement; however, the clinical characterization was mostly focused on the predominant phenotypes, 3,[5][6][7] while a comprehensive picture of how different neurological syndromes associate is unclear.Moreover, some presentations such as brainstem encephalitis have unfavourable neurological prognosis, 8 but the outcome of other associated syndromes remains imprecise.A further difficulty in the setting of anti-Hu PNS is the diagnosis of the cancer, as the screening may be initially negative before the histological diagnosis of an often small, limited stage malignancy. 5,9It is of note that 20 years ago, it was reported that PET scan improves cancer diagnostic accuracy in the setting of PNS; 10,11 however, no specific study has assessed the performance of new-generation CT and PET scanners for cancer screening in patients with anti-Hu PNS.
The main objectives of the present study were therefore to comprehensively evaluate the combination of neurological phenotypes in a large retrospective cohort of anti-Hu PNS and to describe the diagnostic workflow to detect a cancer.Additionally, we investigated the outcome according to different clinical aspects and we took a closer look at patients without a detectable malignancy after a long follow-up.

Study design and population
The medical charts of all patients who tested positive for anti-Hu antibodies at the French Reference Centre on Paraneoplastic Neurological Syndromes and Autoimmune Encephalitis (1990-2022) were retrospectively reviewed.Antibody positivity was established when at least two independent tests (in-house immunohistofluorescence on rodent brain sections, cell-based assay, western blot and/or commercial line blot) gave a concordant result in serum and/or CSF. 12 When needed, physicians in charge were contacted by e-mail or telephone to retrieve clinical information; patients with insufficient data for clinical presentation and/or cancer screening or confirmed alternative diagnosis were excluded, as were those who developed the syndrome after immune checkpoint inhibitor treatment (and have been published elsewhere). 13The onset of the disease was considered acute if symptoms developed to substantially impair the patients' activities of daily living within ≤7 days, subacute if within ≤3 months and chronic if within >3 months.The type of neurological involvement was first classified as CNS and/ or peripheral nervous system according to signs, symptoms and/or diagnostic findings.5][16] If the patients' clinical picture was uncertain (e.g.encephalopathy in the absence of supportive features of limbic involvement) or not classifiable in the previous categories (e.g.isolated generalized chorea), we only used the broader category 'central and/or peripheral nervous system involvement'.Disability was measured by the modified Rankin Score (mRS). 17Cancer stage was extracted from the available oncological reports; for SCLC, the term 'limited-stage' was used when the tumour was confined to one hemithorax and regional lymph nodes. 18As 2 years is the recommendation to extend cancer screening in patients with high-risk phenotypes and antibodies for PNS, 16 we used this threshold to define 'cancer-free' patients when a tumour was not detected, or 'delayed-cancer' when the tumour was found beyond this time-point.All included patients or their legal representatives provided written informed consent.The study is part of the project NCT05745792 and was approved by the institutional review board of the Hospices Civils de Lyon (IRB00012204).

Statistical analysis
All analyses were performed using RStudio (version 4.2.1, 23 June 2022).Data are presented as the number (percentage) or median (range).The correlation (r) between two variables was analysed using Spearman's coefficient.Hierarchical clustering was performed with the ComplexHeatmap v2.12.1 Rpackage using the Manhattan distance and the Ward method ('ward.D') to evaluate the combination of neurological phenotypes.To this end, the characteristics of the clinical picture of the patient were coded as binary variables (e.g.sensory neuropathy: presence/absence).The three identified clusters were further considered for comparison of other clinical and diagnostic variables.Groups were compared using χ 2 or Fisher's exact test for categorical variables and by the Kruskal-Wallis or Mann-Whitney U-test for continuous variables.Post hoc analysis with correction for multiple tests was used when significant results were obtained between the three groups.All statistical tests were two-sided and a P-value of <0.05 was considered significant.
We used multivariate logistic regression to evaluate the adjusted effect of potential predictive variables to explain: (i) mRS score strictly less than four at the last visit (this cutoff was chosen in accordance with previous studies in the anti-Hu PNS population); 3,7 (ii) a 'delayed-cancer' diagnosis and (iii) 'cancer-free' patients.When the inclusion of covariates was limited due to a low number of observations, we selected those with the greatest clinical value to explain the dependent variable.The Kaplan-Meier method was used to estimate the survival probabilities of patients according to: (i) the identified clinical clusters and (ii) the presence of a cancer and its time of diagnosis; they were then compared by the Log-rank test.We used a Cox regression analysis to (i) assess the association of different specific nervous system involvement with dying for a neurological cause and (ii) reproduce the survival analysis to identify predictors of mortality performed by Graus et al. 3 in patients presenting with PNS symptoms.For the former, the dependent variable was 'death due to neurological cause', dichotomizing patients as to whether they died from PNS or not (including both alive patients at the end of follow-up and those who died of cancer progression following PNS stabilization).For both analyses, the assumption of hazard proportionality was tested.If not satisfied, a stratified model was constructed, and the hypothesis of no interaction verified.

Results
Of the 624 patients tested positive for anti-Hu antibodies in the French reference centre between 1990 and 2022, 158 were excluded (133 had insufficient clinical data, 12 had a diagnosis that was not PNS and 13 developed the syndrome after the initiation of immune checkpoint inhibitor treatment); a total of 466 patients with anti-Hu PNS were included (Supplementary Fig. 1).The median (range) age of the study population was 65 years (0-92), 265 (57%) patients were male and 317/333 (95%) patients were current or former smokers.The median (range) interval from neurological onset to the detection of anti-Hu antibodies was 3 months (0-104), and this was negatively correlated to the mRS score at antibody detection (r = −0.24,P < 0.001).Of the 466 patients, 77 (16%) harboured in their serum or CSF neural autoantibodies other than anti-Hu (Supplementary Fig. 2).The median (range) follow-up was 14 months (0-237).

Cancer status
Overall, a histological diagnosis of cancer was reported in 349 (75%) patients (Supplementary Table 1).As the objectives of the present study included to describe the workflow to detect a cancer and the characterization of 'cancerfree' patients, we focused on patients with anti-Hu PNS symptoms predating cancer diagnosis (295/349, 85%).However, a thorough comparison with patients developing PNS symptoms after a known cancer can be found in Supplementary Table 2.

Screening workflow in patients with cancer
Of the 295 patients with anti-Hu PNS symptoms predating cancer diagnosis, 274 (93%) had information regarding cancer screening techniques, 241/274 (88%) with lung and 33/ 274 (12%) with other cancer.Except for two patients in whom the cancer was identified by chest X-ray and a palpable cervical lymphadenopathy, patients with lung cancer underwent thoracic CT scan imaging that first revealed lung cancer in 205/241 (85%), being SCLC in 175/205 (85%).First thoracic CT scan was negative in 33 (14%) patients; among whom, 25/33 (76%) had a subsequent first positive thoracic PET scan, 2 (6%) had a first negative and second positive PET scan and 6 (18%) without a PET scan were diagnosed either in a following CT scan (n = 5) or bronchoscopy (n = 1; Fig. 3).Performing a PET scan after a first negative CT scan significantly shortened the median (range) interval to histological diagnosis [7 months (1-66) versus 14 months (2-45); P < 0.001], which corresponded to SCLC in 31/33 (94%) of these patients.The screening of patients with tumours other than lung is summarized in Fig. 4. Three of these patients (3/33, 9%) had enlarged mediastinal lymph nodes upon cancer screening that regressed before the diagnosis of other cancer different from lung (n = 1 Hodgkin lymphoma, n = 2 urothelial bladder cancer).

Patients with a delayed cancer diagnosis
A total of 13/295 (4%) patients had a cancer diagnosis more than 2 years after the onset of the neurological syndrome [median (range) 43 months (25-106); Supplementary   in thorax imaging prior to cancer definite diagnosis (2/13, 15% versus 4/280, 1%; P = 0.024).Conversely, patients with a delayed cancer diagnosis less often harboured tumours with neuroendocrine histology (8/13, 62% versus 241/278, 87%; P = 0.026).There was no significant difference in the median time from PNS onset to first cancer screening test (Table 2).In a multivariate logistic regression including the chronic onset of the neurological symptoms and the clinical category, only a chronic onset remained independently associated with a cancer diagnosis delayed by more than 2 years [OR 9.85; 95% CI (2.50; 49.26); P = 0.002].

Causes for the absence of histological diagnosis
A cancer diagnosis was not formally established in 117/466 (25%) of the patients, although 78 (66%) of them had some evidence suggestive of an underlying malignancy in imaging screening (Supplementary Fig. 3).The follow-up stopped just after the screening findings in 9/78 (12%) of the patients.
Among the 69 remaining patients with a longer follow-up, the causes for the lack of histological diagnosis, alone or in combination, were inconclusiveness of biopsy in 44/69 (64%) patients, clinical deterioration or death in 33/69 (48%) and/ or decision to not further investigate in 9/69 (13%).Furthermore, regression of previous findings in a subsequent screening was observed in 21/69 (30%) patients.Patients with evidence suggestive of a regressed tumour mainly had lung findings (19/21, 90%) in CT or PET scan, particularly, enlarged and/or hypermetabolic mediastinal lymph nodes and/or lung nodules (15/19, 79%; Supplementary Table 5 and Fig. 5).For these patients, the median (range) follow-up was 36 months (1-237).

Patients without cancer diagnosis after 2 years follow-up
Overall, 33 patients did not have a cancer diagnosis during the entire follow-up, being it of more than 2 years ('cancerfree' patients).Of these, 13 (39%) had initial evidence

Discussion
Herein, we revisited anti-Hu PNS through the analysis of 466 patients that allowed us to provide new clinical and cancer screening insights.First, despite a broad clinical spectrum, we consistently observed three main phenotypic categories: a predominant limbic syndrome, an almost exclusive peripheral involvement and a combination of both central extra-limbic and/or peripheral dysfunction.Second, the results support the role of early screening with PET scan in order to shorten the delay to lung cancer diagnosis when conventional imaging is negative, as well as the extension of close oncological screening to 5 years when a cancer is not found sooner.Additionally, we identified clinical variables associated with outcome and further described a subset of patients without detectable cancer after a long follow-up and arguments suggesting initial cancer regression.Consistent with previous reports, 3,[5][6][7] the clinical picture was often multifocal, and the most frequently involved areas were sensory nerves, cerebellum and limbic system.By collectively assessing each individual combination of neurological involvement, we obtained an overview of the anti-Hu PNS clinical spectrum allowing to deconstruct the heterogenous term 'encephalomyelitis' to define three major phenotypic categories.The most notable differences are possibly those defining the group of patients with an almost exclusive, mainly sensory, neuropathy.As opposed to patients in the LE(+) or mixed group who mostly presented an acute or subacute onset, patients in the neuropathy group had an insidious installation of symptoms in up to a third of cases.Additionally, we less frequently found oligoclonal bands in their CSF, but, conversely, we more frequently found elevated protein levels when compared with patients in the other two groups.As there were no differences in the interval to CSF extraction between the three groups, these findings cannot be explained by different CSF inflammation kinetics.Moreover, both observations have been formerly reported in the setting of anti-Hu and other PNS patients, 19,20 supporting that the different CSF profiles may be related to the clinical presentation.Although we observed a more frequent histological diagnosis of cancer in patients with a peripheral syndrome compared with both categories of patients with CNS involvement [i.e.LE(+) and mixed groups], the proportion of 'cancer-free' patients after a long follow-up was similarly distributed.This difference may therefore be explained by the higher risk of dying in the acute phase of the disease without a complete oncological work-up in patients with CNS involvement, as we and others observed that the presence of dysautonomia or brainstem signs, commonly found in these groups, portend a greater risk of death. 8,21onsistent with this, but also a previous study, 22 patients in the groups with prominent CNS dysfunction more frequently died from neurological cause, while patients with a mainly peripheral syndrome more commonly died from cancer progression.When considering all-cause death, we and others 3,22,23 observed that age above 60 years, and, chiefly, the diagnosis or evidence of cancer had the greatest risk of death.However, compared with the largest previously reported cohort of anti-Hu PNS patients, 3 we found improved 1-, 3-and 5-year survival rates.As the interval to anti-Hu PNS and cancer diagnosis was shorter by nearly half in the present study compared with the report of Graus et al., 3 an early diagnosis may partially explain the observed survival differences.Additionally, the modest improvement in survival trends reported in the last decades for SCLC might have also influenced these results. 24nsurprisingly, a cancer diagnosis was established in the majority of patients, the most frequent being lung cancer, specifically limited stage SCLC confined to mediastinal lymph nodes.In contrast, only a third of SCLC cases present a limited stage of the disease diagnosis in the general population, 24 suggesting that a particular tumour background is characteristic in anti-Hu PNS. 25 When another type of cancer is found in anti-Hu, PNS clinicians may suspect that they have missed SCLC, as diagnosing a small-size mediastinal cancer may be difficult and that this has been reported in a few patients by Lucchinetti et al.; 5 however, this concern could be overcome by studying the Hu expression in the found tumour. 3,26This was not done in the present study reduced by half the time to diagnosis when a first CT scan was negative, supporting an undeniable role for its early implementation in anti-Hu PNS patients.
As anti-Hu PNS symptoms usually antedates the diagnosis of cancer, a key question faced by physicians is until when should they extend screening.Current recommendations in the field of PNS have set this threshold at 2 years for patients with high-risk antibodies and high-risk phenotypes. 16eassuringly, the vast majority of the patients herein had cancer diagnosis within that period.Notwithstanding, the 4% of cases developing a cancer beyond 2 years from onset had some peculiarities, notably a chronic onset of the symptoms and a broad, yet slightly less disabling, involvement of the nervous system.Although these traits often led to a delayed detection of anti-Hu antibodies, the late cancer diagnosis probably depended on additional factors other than the clinical recognition of PNS.This hypothesis is supported by the observation of similar intervals to the first screening test among patients with cancer diagnosis before or after 2 years from onset and a higher proportion of initially negative tests among the latter.Despite the cancer being diagnosed before 5 years from onset in the majority of these patients, four of them exceeded that time-point, the latest developing a lung adenocarcinoma almost 9 years after the initial neurological presentation.These results parallel prior observations of few anti-Hu PNS cases with delayed cancer diagnosis. 3,7,27,28On the contrary, 7% of the present cohort did not have a cancer diagnosis, although they were followed for more than 2 years.This matches previous reports setting the occurrence of anti-Hu PNS without detectable cancer systematically below 10%. 7,27Interestingly, this subset of patients shared some distinctive features, including a younger age, a less frequent smoking habit and, clinically, a less frequent involvement of peripheral nerves compared with more common dysfunction of limbic areas and myenteric plexus than patients with identified cancer.Moreover, they less frequently had pleocytosis, elevated CSF protein level, presented with less disabling clinical pictures and had substantially better median overall survival.Although it may be tempting to interpret these differences as signs of a distinct anti-Hu non-paraneoplastic entity, it is of note that around a third of these patients initially had some suspicious imaging findings (mainly enlarged mediastinal lymph nodes) that spontaneously regressed upon a subsequent screening.As anti-Hu antibodies have been recurrently associated with partial control of tumour growth, 9,23,29 and tumour regression has been hypothesized in a few patients, 3,27,30,31 it remains possible for these cases, as observed in other PNS (e.g.testicular burned-out seminoma with anti-Kelch-like protein 11 antibodies). 32Therefore, given the rarity of the disease, the strong cancer association and the pivotal role of cancer in determining survival, we deem reasonable to extend the screening with CT and PET scan from the current 2-5 years and to remain vigilant until 10 years after PNS onset.However, treating physicians should be aware that while a cancer diagnosis within 2 years from clinical onset carries poor survival rates, once their patient overpasses that threshold, their 5-and 10-year survival rates improve independently of being subsequently diagnosed with a cancer.
A puzzling result from the present analysis concerns the role of elevated CSF proteins level in anti-Hu PNS, as this was less frequently found in 'cancer-free' patients, more frequently reported in patients with a predominant peripheral syndrome, and was also as an independent predictor of severe disability at last visit in the total cohort.While the relationship between the three observations is consistent ('cancer-free' patients less frequently had a predominant peripheral involvement and they also had lower disability burden), the biological significance of it remains unclear.However, evaluating the value of elevated CSF protein as a predictive biomarker of outcome is an accessible and easy-to-implement task in a prospective study.
The main limitations of the study result from its retrospective design, which, does however suit the rarity of the disease.First, we were not able to include a fifth of the patients tested positive for anti-Hu antibodies in the study centre due to the unavailability of clinical information, potentially introducing a selection bias.However, we find it unlikely that a particular subpopulation of anti-Hu PNS patients was overrepresented in the group of excluded patients, and we believe the analysed cohort (which includes patients on a nationwide level over a 30-year period) to be representative of the anti-Hu PNS population.Second, as the diagnostic work-up was not performed equally in all patients, the clinical classification sometimes relied on treating physician interpretation or that of the study investigator who collected the information, which may have been subject to a reporting bias.Third, the hierarchical cluster analysis grouped together patients with different clinical syndromes in the mixed group (e.g.patients with brainstem encephalitis and patients with LEMS), adding internal heterogeneity.However, as there were only a few patients with isolated clinical syndromes in this cluster, we consider this approach accurately represents the multifocality in anti-Hu PNS.Fourth, the sample size for sub-group analyses was small, which limits the strength of conclusions.However, as discussed above, certain results parallel previous findings in anti-Hu PNS studies, 3,[5][6][7]19,[20][21][22][23]25,27,28,30,31 supporting the validity of the present work. Future studies explorin the physiopathology underpinning anti-Hu PNS and causes leading to immune tolerance breakdown should interpret the analysis of biological samples and tumours with the clinical aspects reported herein.

Conclusion
Although anti-Hu PNS clinical presentation is mostly multifocal, we also observed patients presenting with a predominant limbic syndrome or isolated sensory neuropathy.Early implementation of PET scan in the screening shortens the interval to cancer diagnosis, which was the strongest predictor of death, especially if diagnosed within 2 years from clinical onset.As cancer was diagnosed more than 2 years after PNS onset in few patients, screening should be extended up to 5 years.In addition, tumour regression was suspected in a substantial proportion of patients without cancer after a long follow-up.

Figure 1
Figure 1 Clinical spectrum and survival in anti-Hu PNS patients.(A) Heatmap and hierarchical clustering of anti-Hu PNS patients taking into account the specific involvement of the nervous system.(B) Kaplan-Meier curves.Tick marks indicate censored patients, and comparison was made using the Log-rank test.Abs, antibodies; ALS, amyotrophic lateral sclerosis; LEMS, Lambert-Eaton Myasthenic syndrome.

Figure 2
Figure 2 Variables associated with outcome.(A) Forest plot of Cox regression analysis assessing the association of different specific nervous system involvement with dying from neurological cause.(B) Forest plot of multiple logistic regression analysis evaluating the adjusted effect of potential predictive variables to explain a mRS strictly less than four at the last visit.(C) Cox regression analysis using the same model reported by Graus et al.3 to identify predictors of all-cause mortality in patients presenting with PNS symptoms (n = 412/466, 88%).As the assumption of hazards proportionality was not satisfied by this model, we conducted a stratified model according to the putative variable (mRS at diagnosis), and the hypothesis of no interaction was verified.*Confirmed cancer or evidence of it upon screening.Abs, antibodies; LEMS, Lambert-Eaton myasthenic syndrome; mRS, modified Rankin score; N, number.

Figure 3
Figure 3 Survival according to the diagnosis of cancer and its timing.(A) Kaplan-Meier curves.Tick marks indicate censored patients, and comparison made using the Log-rank test.(B) Three-, 5-and 10-year survival rates table.*Of these, n = 13 patients were subsequently diagnosed with cancer and n = 33 remained 'cancer-free' during the entire follow-up.CI, confidence interval.

Figure 4
Figure 4 Cancer screening workflow.Screening techniques in patients with paraneoplastic symptoms predating the diagnosis of cancer.aIn both patients, the following PET scan was positive, 26 and 8 months after the first screening.b One patient was diagnosed directly by bronchoscopy, the other 5/6 had a subsequent positive CT scan.c These cancers included n = 4 of unknown origin (one undifferentiated adenocarcinoma, one small cell from pelvis adenopathy conglomerate, one neuroendocrine undifferentiated from liver metastasis and one neuroendocrine inguinal mass), n = 3 neuroblastoma, n = 2 prostate (one adenocarcinoma and one small cell), n = 2 breast carcinoma, n = 1 urothelial bladder cancer, n = 1 renal cell cancer, n = 1 thymoma, n = 1 small cell thymus, n = 1 small cell hypopharynx and n = 1 neuroendocrine tumour of the ileum.d This patient had a neuroendocrine undifferentiated cancer of the rectum.e These cancers included n = 3 diagnosed in subsequent CT scan (one breast cancer, one Hodgkin lymphoma, one ganglioneuroblastoma); n = 2 diagnosed on subsequent PET scan (one prostate adenocarcinoma, one chronic lymphocytic leukaemia), n = 1 bladder cancer by cystoscopy, n = 1 breast cancer on MRI and n = 1 prostate epithelioma found due to elevated prostate-specific antigen.f These cancers included n = 1 prostate adenocarcinoma found due to elevated prostate-specific antigen, n = 1 biopsy of lymph node found in physical examination revealing an undifferentiated adenocarcinoma suggestive of prostatic origin, n = 1 urothelial bladder cancer in cystoscopy, n = 1 neuroblastoma in thorax MRI and n = 1 poor differentiated adenocarcinoma in sigmoidoscopy.g These cancers included n = 1 breast cancer found on mammography and ultrasound, and n = 1 urothelial bladder cancer on cystoscopy (haematuria).

Figure 5
Figure 5 Regression of mediastinal lymphadenopathies and lung nodule in a patient with anti-Hu PNS.A woman in her 50s, current smoker, subacutely developed a clinical picture compatible with limbic encephalitis leading to the detection of anti-Hu antibodies.The first cancer screening, performed 4 months after clinical onset, found strongly hypermetabolic para-tracheal, precarinal and superior right hilar lymphadenopathies (Ai-Aii) and a weak hypermetabolism (Aiii) in a superior right lung nodule of 6 mm diameter (Aiv).The result of a biopsy performed 2 months later was inconclusive.Subsequent screening, performed 7 months after clinical onset, found an almost complete regression of mediastinal hypermetabolic lymphadenopathies and the lung nodule (Bi-Biv).Subsequent screening performed 14 and 25 months after clinical onset were completely normal (Ci-Div).

Table 1 Demographic, clinical features and diagnostic findings according to anti-Hu PNS cluster LE(+) group n = 107 Neuropathy group n = 126 Mixed group n = 174 P-value
P-values correspond to either Fisher's exact test or Kruskal-Wallis test as appropriate.Statistically significant values are presented in bold.Abs, antibodies; CI, confidence interval; CPM, cyclophosphamide; CSF, cerebrospinal fluid; IVIG, intravenous immunoglobulins; mRS, modified Rankin Score; OCB, oligoclonal bands; PE, plasma exchange; RTX, rituximab.a Post hoc testing significant between LE(+) versus neuropathy and mixed groups but not between neuropathy and mixed groups.b Post hoc testing significant between neuropathy versus LE(+) and mixed groups but not between LE(+) and mixed groups.c Other treatments included n = 5 patients receiving azathioprine, n = 5 mycophenolic acid (one of them received also methotrexate) and n = 1 mitoxantrone.d Post hoc testing with Holm correction; non-significant.

Table 2 Demographic, clinical features and diagnostic findings of anti-Hu PNS patients according to whether they had a cancer diagnosis before or after 2 years from clinical onset Cancer ≤2 years (n = 280) Cancer >2 years (n = 13) P-value
-values correspond to either Fisher's exact test or Mann-Whitney U-test test as appropriate.Statistically significant values are presented in bold.There was no clear date information for n = 2 patients, for whom the delay to cancer diagnosis was not calculated.Abs, antibodies; mRS, modified Rankin Score; OCB, oligoclonal bands; n, number. P