Standardizing the unstandardized: The international system reaches the cerebrospinal fluid

Précis

Two new studies now make the most substantial case yet that the framework of the International System for Reporting Serous Fluid Cytopathology can be carried across the biologic boundary to include cerebrospinal fluid, and current work on the second edition of the system will address cerebrospinal fluid for the first time as a special category of body fluid—an expansion the international community has explicitly endorsed. The contributions of these authors will stand among the empirical foundations on which a coherent, cerebrospinal fluid–specific—and, ultimately, fluid‐agnostic—reporting standard can be built.

Few developments in diagnostic cytopathology over the past decade have been as consequential as the migration from free‐text, laboratory‐specific reporting toward standardized, tiered classification. The International System for Reporting Serous Fluid Cytopathology (TIS), published in 2020, extended to pleural, peritoneal, and pericardial effusions the same five‐tier logic that had already reshaped thyroid, cervical, pancreaticobiliary, and urinary cytology: a small set of mutually exclusive categories—nondiagnostic (ND), negative for malignancy (NFM), atypia of undetermined significance (AUS), suspicious for malignancy (SFM), and malignant (MAL)—each anchored to an actionable risk of malignancy (ROM). ¹ The value of that architecture is no longer in serious doubt. A systematic review and meta‐analysis of 16 studies comprising more than 19,000 effusions found a clean, monotonic ROM gradient—approximately 27% for ND, 11% for NFM, 49% for AUS, 90% for SFM, and 100% for MAL—with excellent specificity when SFM and MAL were treated as positive. ² Single‐institution and community hospital series have reproduced the same ordering with notable consistency despite differences in case mix and preparation technique. ³ , ⁴

Conspicuously absent from that consolidation was the cerebrospinal fluid (CSF), and the omission was deliberate. CSF is not a serous effusion but an ultrafiltrate of the choroid plexus, normally near‐acellular and populated by a sparse background of small lymphocytes and monocytes rather than by mesothelium; its differential is dominated by leptomeningeal dissemination of hematologic malignancy, metastatic solid tumors, and primary central nervous system neoplasms set against a reactive lymphoid backdrop. Yet CSF has labored under precisely the unstructured, idiosyncratic reporting that TIS was created to abolish. Two new studies now make the most substantial case yet that the TIS framework can be carried across that biologic boundary, and they do so from complementary directions. Wang and Cai reclassified 854 CSF specimens from a single North American academic center, integrating flow cytometry and estimating a category‐specific ROM. ⁵ Yildiz Aktas and colleagues went broader still, applying TIS to 1137 specimens from three institutions on two continents and linking each diagnostic category to overall survival. ⁶ Together with the earlier feasibility study of Vidali and colleagues, who classified 618 CSF samples in Athens, ⁷ these reports move the question from whether TIS can be applied to CSF to what it tells us once it is applied. Their appearance is especially timely because work has now commenced on the second edition of TIS, in which CSF will be addressed for the first time as a special category of body fluid—an expansion the international community has explicitly endorsed: in a recent preparatory survey for TIS 2.0, respondents were broadly supportive of extending the system to additional body sites. ⁸

The two studies are usefully complementary in design and in what they validate. Wang and Cai offer depth: a single laboratory, 1 year of accruals, and, crucially, systematic correlation with flow cytometry, which allowed them to convert morphologically indeterminate specimens into definitive diagnoses and to compute an ROM for each tier. ⁵ Yildiz Aktas and colleagues offer breadth and an outcome anchor: a large, deliberately heterogeneous cohort spanning United States and Chinese centers and a survival end point that tests whether the categories carry prognostic—not merely descriptive—weight. ⁶ Where Wang and Cai ask how often each category proves malignant, Yildiz Aktas et al. ask how long patients in each category live; these are two faces of the same underlying validity question and, as I shall argue, two complementary solutions to the problem that has long made CSF resistant to a ROM‐based system.

The category distributions are revealing precisely because they differ. Wang and Cai report NFM 87.2%, AUS 7.1%, SFM 0.6%, MAL 5.0%, and ND 0.0% ⁵ ; Yildiz Aktas and colleagues report NFM 67.1%, AUS 6.2%, SFM 2.7%, MAL 20.8%, and ND 3.2% ⁶ ; whereas the figures in the report by Vidali et al. (NFM 78.8%, MAL 19.4%) sit between them. ⁷ The malignant fraction therefore ranges from 5% to nearly 21% across the three series. Case mix, not diagnostic disagreement, accounts for most of this spread. The Wang and Cai cohort is dominated by serial surveillance of patients with established hematologic malignancy—a population that generates large numbers of true‐negative specimens and dilutes the malignant fraction—whereas the Yildiz Aktas cohort is weighted toward solid tumors, which accounted for 57% of patients with clinical data and fully 93% of all malignant diagnoses, lung primaries foremost among them. The composition of the MAL category thus inverts between the two articles: predominantly hematologic in the surveillance‐heavy, single‐center series and overwhelmingly epithelial in the multi‐institutional series. This pattern is consistent with the broader CSF literature, in which an earlier multi‐institutional series identified metastatic solid tumors in 35%–40% of patients who had malignant CSF with hematolymphoid disease, who comprised the single largest group. ⁹ The message for the second edition of TIS is clear: CSF category frequencies will track local referral patterns even more strongly than effusion frequencies do, and published distributions must be read as epidemiology, not as performance benchmarks.

The most conceptually important divergence, as in the effusion literature, concerns adequacy and the ND category—and here the two articles, read together, expose a subtlety that neither resolves alone. In serous effusions, an acellular or blood‐only specimen is conventionally ND, and the meta‐analytic ROM for ND is nontrivial (around 27%) precisely because an inadequate effusion may conceal disease. ² Wang and Cai argue, persuasively, that this logic cannot be transplanted wholesale to CSF: normal CSF is paucicellular by definition, so an acellular or red cell–only specimen, in most contexts, is positive evidence of the absence of a cellular malignant infiltrate rather than a sampling failure. They accordingly classified 56 acellular and 75 red cell–only specimens as NFM, assigned ND to no case, and justified the decision by a benign follow‐up course and an NFM ROM of just 0.3%. ⁵ Yildiz Aktas et al. retained a small ND fraction of 3.2%—but the instructive detail is why. Their ND rate proved to be largely an artefact of preparation method: at the center that used cytospin preparations, blood‐only specimens could be recognized and called ND; whereas, at the two centers that used liquid‐based processing, which strips background red cells, ND could be assigned only on the basis of cellular degeneration and was correspondingly rare. ⁶ In other words, one and the same biologic specimen may be ND under one preparation and NFM under another. This is a finding of real consequence for a standard intended to harmonize reporting across laboratories. It suggests that the second edition of TIS should not merely redefine CSF adequacy—restricting ND to specimens rendered uninterpretable by severe degeneration, clotting, or preservation artefact rather than by low cellularity, as Wang and Cai propose—but should make that definition explicitly preparation‐aware, so that an identical specimen does not migrate between categories according to whether a laboratory favors cytospin or liquid‐based cytology.

The indeterminate categories raise a related concern. Across the three series, the combined AUS and SFM fraction varies from <1% (Vidali et al.) to roughly 8%–9% (Wang and Cai; Yildiz Aktas et al.), a spread too large to attribute to true biologic differences and more plausibly reflecting divergent thresholds for invoking atypia and suspicion. Wang and Cai confronted the practical root of this variability directly: the specimen reported descriptively as hypercellular, with increased small lymphocytes and/or monocytes, a noncommittal phrase that maps to no TIS tier and conceals everything from florid reactive pleocytosis to occult, low‐grade B‐cell lymphoma. Sixty‐five such specimens were redistributed—54 to NFM, eight to AUS, and three to MAL—on the basis of re‐review and flow cytometry. ⁵ That disciplined dissolution of the descriptive gray‐zone report, to my mind, is among the most valuable contributions of either report because it demonstrates that the indeterminate report so corrosive to clinical communication can be resolved into the TIS architecture without loss of information. The lesson for the second edition is that publishing category labels is necessary but not sufficient: CSF‐specific morphologic criteria, and an illustrated atlas of the reactive versus neoplastic lymphoid distinction, will be needed to anchor reproducible use of AUS and SFM.

Both articles grapple, in different ways, with the deepest obstacle to importing an ROM‐based system into CSF: the near absence of a tissue gold standard. Leptomeningeal disease is seldom biopsied, so truth cannot be established the way it is for an effusion with a corresponding pleural biopsy. Wang and Cai address this by assembling a composite reference standard—serial cytology, flow cytometry, and clinical course—and report an ROM of 0.3% for NFM, 4.5% for AUS, 66.7% for SFM, and 100% for MAL, with 95% sensitivity and 100% specificity among cases carried to a definitive end point. ⁵ That gradient mirrors the pooled effusion estimates almost step for step, ² but it rests on small denominators—the SFM figure derives from three specimens—and on a reference standard that inevitably entangles the index test with its own verification. Yildiz Aktas et al. sidestep the ROM problem altogether by substituting overall survival as the validating end point, an approach with direct precedent in the World Health Organization System for Pancreaticobiliary Cytology, in which diagnostic categories were shown to stratify not only malignancy risk but survival. ¹⁰ In my view, the result is the single most persuasive piece of evidence yet that TIS categories are biologically meaningful in CSF rather than merely descriptive.

The survival data are striking in their orderliness. Across 579 patients with follow‐up, restricted mean survival over 5 years fell monotonically across the categories—53.9 months for NFM, 48.4 for AUS, 40.3 for SFM, and just 18.7 for MAL—and 5‐year survival ranged from 78.8% in NFM to 10.4% in MAL. ⁶ In unadjusted analysis, every category carried a significantly higher hazard of death than NFM, culminating in a 14‐fold hazard for MAL; after adjustment for age, sex, disease group, and diagnostic category, MAL remained the strongest independent predictor of mortality at roughly a 6.5‐fold hazard, with SFM independently associated at a 2.7‐fold hazard. ⁶ Two features of the adjusted model deserve comment. First, the ND and AUS categories lost their independent association with survival once disease group was taken into account—appropriately so, because the prognosis of an indeterminate CSF result is governed largely by the underlying disease rather than the cytologic label itself. Second, primary tumor type exerted its own powerful effect: patients with solid tumors fared roughly 2.5‐fold worse than those with hematologic malignancies, echoing a consistent literature in which leptomeningeal disease from solid tumors carries a grimmer prognosis than that from hematologic primaries. ¹¹ , ¹² The implication is that, in CSF, the diagnostic category and the clinical context are not competing variables but complementary ones, and a mature reporting framework should present them together.

These prognostic data also begin to convert TIS categories into management signals, which is where a reporting system ultimately earns its keep. A negative CSF result in a patient with no malignant history—overwhelmingly the outcome in such patients across all three series—can reasonably terminate the diagnostic pursuit rather than prompt reflexive serial sampling; an AUS or SFM result is best read as an instruction to obtain further material for ancillary testing rather than as a diagnosis in its own right; and an MAL result, particularly from a solid primary, identifies a patient whose survival is measured in months and for whom timely intrathecal or systemic therapy is paramount. The second edition of TIS has an opportunity to pair each CSF tier with consensus, evidence‐graded management suggestions while making clear that these are starting points for multidisciplinary discussion rather than substitutes for it.

If the study by Yildiz Aktas and colleagues supplies the prognostic validation, the Wang and Cai study supplies the mechanism by which indeterminate CSF diagnoses can be resolved, and that mechanism is flow cytometry. CSF cytomorphology alone is notoriously insensitive for leptomeningeal hematologic disease, and multiparameter flow cytometry materially improves detection. ¹³ In the Wang and Cai cohort, it was decisive: 10 specimens originally read as negative or atypical were reassigned to MAL on the strength of a clonal flow result. ⁵ This is precisely how adjunct testing reduces indeterminate diagnoses—not by replacing morphology but by driving the AUS and SFM tiers toward a definitive end point. The unresolved question for the second edition is not whether to use flow cytometry but when and how often; published algorithms already provide a scaffold, recommending paired cytology and flow cytometry at the initial assessment of suspected hematologic involvement and repetition until results are consecutively negative. ¹⁴ Flow cytometry was applied unevenly across the three studies, reflecting local custom rather than protocol, and codifying its role within TIS would do much to rationalize that variation. The second edition should likewise acknowledge the rapidly widening molecular landscape, in which cell‐free tumor DNA increasingly detects and characterizes leptomeningeal disease in specimens whose cellularity would defeat conventional cytology; a modular system that defines, for each category, the threshold at which ancillary interrogation is indicated will age better than one that prescribes a fixed test menu.

Quantitative adequacy remains unsettled, and here the two articles diverge instructively once more. The effusion literature shows malignancy detection rising with fluid volume, and a foundational CSF series reported sensitivity climbing from roughly 68% at 3.5 mL to 97% at 10.5 mL. ¹⁵ Yet Yildiz Aktas et al. found no significant association between specimen volume and diagnostic category across their 1137 cases and could not establish a minimum volume for reliable interpretation. ⁶ The discrepancy is most plausibly explained by the distinctive biology of CSF—the intermittent, sparse exfoliation of tumor cells and the small volumes typical of lumbar puncture—rather than by any true independence of yield from volume. The pragmatic position the second edition is well placed to codify. therefore, is a soft one: that 1 mL be regarded as a minimum acceptable volume for basic assessment, with 3 mL or more preferred and larger volumes actively solicited whenever ancillary or molecular studies are anticipated, while recognizing that CSF, unlike pleural fluid, cannot be expected to obey a simple volume‐to‐sensitivity rule. Both articles also underscore that preparation technique is not a neutral choice: cytospin and liquid‐based methods enrich cells differently, render red cell contamination differently, and—as the adequacy discussion above shows—can move a specimen between categories.

Several limitations are common to these studies and temper any claim that the question is settled. All are retrospective; none formally assessed interobserver reproducibility, which is the acid test of any classification intended for global use and which the very divergence in AUS and SFM use across the three cohorts suggests will be nontrivial to achieve. The multi‐institutional design that lends Yildiz Aktas and colleagues’ study its generalizability also introduces heterogeneity in preparation, staining, and threshold that the authors rightly flag, and survival distributions in such a cohort are vulnerable to selection and referral bias. Composite and surrogate reference standards—whether the ROM scaffold of Wang and Cai or the overall survival end point of Yildiz Aktas et al.—are pragmatic responses to the missing tissue gold standard but are not equivalent to it. None of this undermines the central thesis; it defines the agenda for the prospective, multi‐institutional, and explicitly reproducibility focused studies that must follow, ideally incorporating the troublesome lymphoid‐predominant specimen and spanning laboratories on more than one continent, much as the maturation of the effusion criteria once required.

Taken together, the two studies accomplish more than their individual results. Read in series, they demonstrate that the TIS scaffold is robust enough to bear a specimen for which it was not designed ⁵ , ⁶ ; that its categories retain both a graded ROM ⁵ and, independently, a graded survival signal ⁶ ; and that the points at which CSF demands bespoke rules are now identifiable with some precision. Therefore, a CSF chapter for the second edition of TIS can be specific rather than aspirational. It should redefine adequacy so that it does not penalize physiologic paucicellularity, and make that definition preparation‐aware; provide CSF‐specific criteria and illustrations for the indeterminate categories; embed an explicit, tiered role for flow cytometry and emerging molecular assays; pair each category with both a risk estimate and a prognostic expectation, presented alongside the clinical context that so strongly modifies them; and state plainly that category frequencies will track local case mix. That the international community already favors such an expansion ⁸ only strengthens the case for getting the details right.

The broader implication deserves emphasis, and it is more than rhetorical. The same reasoning that justifies extending TIS to CSF applies to other body fluids that, even now, remain without any standardized reporting framework. Synovial fluid is reported in an almost entirely free‐text fashion, its questions—crystal arthropathy, septic arthritis, the occasional synovial or metastatic malignancy—are poorly served by a vocabulary that varies from laboratory to laboratory. Aqueous and vitreous specimens occupy an even more precarious position: they are scant, technically demanding, and yet asked to resolve the high‐stakes distinction between vitreoretinal lymphoma and inflammatory uveitis, often on a handful of fragile cells. In each instance, the core problem is identical to the one TIS already solves for effusions and now, on the evidence of these studies, for CSF—the need for mutually exclusive categories, explicit and preparation‐aware adequacy criteria, transparent risk stratification, and an agreed role for ancillary testing. A modular international system that preserves the five‐tier spine while permitting fluid‐specific annexes would let each of these specimens inherit decades of accumulated reporting discipline without being forced into an effusion mold that does not fit. The CSF experience is the proof of concept: it shows that the system can adapt.

Standardization in cytopathology has never been about uniformity for its own sake; it is about making a report mean the same thing in New Haven, in Beijing, in Athens, and everywhere else it is read, and about tying each category to a defensible expectation—of risk, and now of outcome—that a clinician can act upon. The studies by Wang and Cai ⁵ and Yildiz Aktas and colleagues, ⁶ set against the earlier work of Vidali and colleagues, ⁷ show that this ambition is achievable for CSF: that TIS categories can be applied across continents and preparation methods, that they stratify both malignancy risk and survival, and that flow cytometry and molecular testing sharpen them further. Their differences are not contradictions but a map of the work that remains. As the second edition of TIS takes up that work, these contributions will stand among the empirical foundations on which a coherent, CSF‐specific—and, ultimately, fluid‐agnostic—reporting standard can be built.

CONFLICT OF INTEREST STATEMENT

The author declares no conflicts of interest.