Registry & Real-World Data Digest — Issue 01

CF, Registry & Real-World Data Digest

This edition focuses on registry-based research conducted outside the United States, drawing on data systems from China, Australia, Sweden, and Norway. Across five studies spanning oncology, emergency medicine, toxicosurveillance, orthopedics, and perinatal epidemiology, a consistent set of methodological themes emerges: the value of large, well-curated registries for generating real-world evidence, the persistent challenges of confounding and incomplete clinical data, and the importance of honest interpretation when effect estimates are wide or effect sizes are small.

---

PM2.5 Reduction Prolongs Survival in Beijing Lung Cancer Patients, but Rising Ozone Erodes Nearly Half the Benefit

Lei Yang, Ning Kang, Tao Xue, et al.

Summary: This registry-based cohort study used data from 172,845 lung cancer patients diagnosed in Beijing between 2003 and 2022 to examine the independent and joint associations of long-term PM2.5 and ozone (O3) exposure with all-cause mortality. Both pollutants were independently associated with higher mortality risk (HR ~1.11 per 10 µg/m³ increase for each). Counterfactual modeling estimated that PM2.5 reductions since 2013 were associated with approximately 13,229 avoided deaths and a 7.8-month gain in 5-year expected lifespan—but rising O3 and pollutant interaction effects offset roughly 40% of those estimated gains.

Why it matters: China's "War on Pollution" has produced measurable PM2.5 reductions, and this study attempts to quantify survival benefits in a clinically vulnerable population—lung cancer patients—where air pollution exposure plausibly compounds disease-related mortality risk. The finding that O3 and the PM2.5–O3 interaction substantially erode PM2.5-related gains is a policy-relevant signal: single-pollutant control strategies may be insufficient, and the trade-off between PM2.5 and O3 (which can increase as certain precursor emissions are reduced) deserves explicit regulatory attention. For oncology and public health communities, the 7.8-month gain in 5-year expected lifespan, if credible, is clinically meaningful for a disease with poor prognosis.

Expert perspective: The study is methodologically ambitious, and several strengths deserve acknowledgment: the sample is large and drawn from an established cancer registry spanning two decades, exposure assignment uses a validated high-resolution spatiotemporal dataset (ChinaHighAirPollutants), and the analytic approach—Cox models with tensor-product splines for joint exposure-response and counterfactual simulation—is appropriately sophisticated. However, several limitations deserve careful scrutiny before placing too much confidence in the specific effect estimates and survival projections.

Confounding is the central concern. Pre-diagnosis air pollution exposure is correlated with socioeconomic position, smoking prevalence, occupational exposures, healthcare access, cancer stage at diagnosis, and treatment quality—all of which independently affect lung cancer survival. Registry data almost never capture smoking history, stage, histology, and treatment with sufficient completeness and granularity to eliminate residual confounding. If PM2.5 exposure is a proxy for broader socioeconomic or behavioral disadvantage, the associations may be attenuated or inflated in unpredictable ways.

Exposure assignment is ecological: a three-year average pre-diagnosis concentration is assigned from gridded model output, not measured at the individual level. This introduces exposure misclassification that is likely non-differential but can bias estimates depending on the measurement error structure. Residential mobility, indoor time, and personal behavior are unaccounted for.

The counterfactual simulations deserve particular skepticism. Estimating "avoided deaths" by comparing observed outcomes to a counterfactual 2013 exposure scenario assumes that the association between pollution and mortality is causal and stable over time, that no other temporal trends confound the comparison (treatment advances, screening uptake, and changes in cancer subtype distribution all shifted substantially between 2013 and 2022 in China), and that the model coefficients generalize across exposure ranges. These are strong assumptions. The 13,229 avoided deaths figure and the 7.8-month expected lifespan gain should be understood as model outputs conditional on those assumptions—not as empirically observed survival gains.

The multiplicative interaction between PM2.5 and O3 at higher O3 concentrations is biologically plausible, and the statistical approach is appropriate, but interaction modeling at the extremes of joint distributions is inherently uncertain and sensitive to model specification.

From a registry and real-world data perspective, the completeness of follow-up for mortality ascertainment, the accuracy of cause of death, and the availability of clinical covariates (stage, treatment, comorbidities) are not discussed in detail. The absence of cause-specific mortality analysis is also notable: all-cause mortality in lung cancer patients is dominated by cancer-specific mortality but not entirely, and pollution effects on cardiovascular and other non-cancer mortality may differ.

Overall, this is a well-executed observational study addressing a genuinely important public health question. The directional finding—that PM2.5 reduction benefits lung cancer patient survival but that rising O3 partially offsets these gains—is plausible and policy-relevant. However, the specific numerical estimates should be treated as illustrative rather than precise, given the confounding challenges inherent to this design.

Relevance: Registry and Real-World Data Methods

Source: https://pubmed.ncbi.nlm.nih.gov/42778299/

---

No Clear Winner Between Lidocaine and Amiodarone for In-Hospital Refractory VF/pVT: A Chinese Registry Analysis

Xu Niu, Tangxing Jiang, Wen Zheng, et al.

Summary: This multicenter prospective registry study from China compared lidocaine and amiodarone in 612 adult in-hospital cardiac arrest (IHCA) patients with refractory ventricular fibrillation or pulseless ventricular tachycardia, drawn from 32,545 cases across 40 hospitals. No statistically significant differences were observed between the two antiarrhythmic agents in survival to discharge, sustained return of spontaneous circulation, or favorable neurological outcome, both before and after multivariable adjustment. The authors present this as the first rigorously adjudicated IHCA-specific comparative effectiveness analysis of these two drugs using national registry data.

Why it matters: Current resuscitation guidelines—including those from the American Heart Association and the European Resuscitation Council—permit either amiodarone or lidocaine for shock-refractory VF/pVT, largely on the basis of out-of-hospital cardiac arrest (OHCA) data, most notably the ALPS trial. IHCA differs meaningfully from OHCA in etiology, monitoring intensity, time to first shock, and patient comorbidity profile, so direct extrapolation has always been an assumption rather than a demonstrated fact. This study provides real-world evidence specifically in the IHCA context, offering modest reassurance that the guideline's permissive stance on either agent is not obviously wrong in this setting.

Expert perspective: The study is methodologically conscientious in several respects: it uses a prospective, Utstein-style registry; it performs careful case adjudication against guideline indications; it applies multilevel mixed-effects logistic regression to account for hospital clustering; and it reports prespecified subgroup analyses. These are genuine strengths for a registry-based comparative effectiveness study.

That said, several important limitations deserve direct acknowledgment. The most fundamental is sample size. Of 32,545 IHCA cases, only 612 met inclusion criteria—approximately 1.9%—and these were split unevenly between the two groups (259 lidocaine, 353 amiodarone). The adjusted odds ratios for survival (1.22, 95% CI 0.63–2.37) and neurological outcome (1.44, 95% CI 0.70–2.99) carry wide confidence intervals that are entirely consistent with clinically meaningful differences in either direction. A null result here is not strong evidence of equivalence; it is substantially a consequence of underpowering. The authors do not report a formal power calculation or minimum detectable effect size, which would help readers calibrate how much uncertainty remains.

Treatment-selection bias is a persistent concern in any non-randomized antiarrhythmic comparison. Although multivariable adjustment was performed, the variables available in a registry are unlikely to fully capture the clinical gestalt that drives a resuscitation team's choice between lidocaine and amiodarone—factors such as perceived hemodynamic trajectory, prior drug exposure, formulary availability, and institutional habit. In China, lidocaine has historically been more widely used than in Western settings, which may reflect formulary patterns rather than clinical discrimination, and also limits generalizability to non-Chinese healthcare systems where amiodarone predominates.

The data collection window (July 2019 to December 2020) predates the widespread adoption of newer resuscitation bundle approaches and coincides partially with the COVID-19 pandemic period, which may have affected IHCA characteristics and hospital resuscitation practices in ways that are difficult to adjust for retrospectively. The registry's 40-hospital coverage, while impressively broad for a Chinese IHCA dataset, raises questions about representativeness, particularly for rural and lower-resource settings. Completeness rates for individual variables and missing data handling are not reported in the abstract—both relevant concerns for any registry analysis.

On balance, this is useful incremental real-world evidence that neither contradicts current guidelines nor provides sufficient statistical precision to conclude true equivalence. It is appropriately framed as supportive of the existing permissive guideline recommendation rather than as a practice-changing finding. The careful case adjudication process and multilevel modeling approach are worth noting as best practices applicable beyond cardiac arrest research—including for those working with small denominators in rare-disease registries.

Relevance: Registry and Real-World Data Methods

Source: https://pubmed.ncbi.nlm.nih.gov/42767592/

---

Fatal Drug Toxicity in Victoria: A Clinical Registry Snapshot Highlighting Polysubstance Risk and Pre-Hospital Mortality

James White, Sarah J Hampton, Shaun Greene

Summary: This retrospective observational case series characterizes 46 fatal cases recorded in the Emerging Drugs Network of Australia-Victoria (EDNAV) clinical registry between September 2020 and December 2025, drawn from 5,026 total presentations. Fatal cases were predominantly male (89%), with a median age of 33 years, and nearly all involved polysubstance exposure (91%, median three substances detected). Methamphetamine and heroin were the most frequently detected substances; two-thirds of deaths involved out-of-hospital cardiac arrest, and hypoxic brain injury was the terminal complication in 72% of cases. Novel psychoactive substances (NPS) appeared in fewer than 8% of detections and were never found in isolation.

Why it matters: The near-universal polysubstance pattern—particularly mixed stimulant-sedative combinations—challenges single-substance harm-reduction frameworks and complicates clinical management and forensic attribution. The predominance of out-of-hospital cardiac arrest and irreversible hypoxic brain injury at presentation underscores that most of these deaths are effectively determined before emergency services arrive, making community-level interventions (bystander cardiopulmonary resuscitation, naloxone distribution, rapid emergency activation) the most actionable levers for reducing mortality. For toxicosurveillance systems more broadly, EDNAV demonstrates how a registry anchored to high-sensitivity mass spectrometry can generate actionable epidemiological intelligence that routine death-certificate data cannot. The relatively low prevalence of NPS in this cohort is also informative, tempering concern that novel synthetic substances have yet displaced traditional illicit drugs as primary drivers of fatal toxicity in Victoria.

Expert perspective: EDNAV is a well-designed, clinically anchored state-wide toxicosurveillance registry that combines emergency department case capture with liquid chromatography-quadrupole time-of-flight mass spectrometry (LC-QTOF-MS)—a methodology capable of detecting a broad range of substances, including NPS that would be missed by conventional immunoassay screening. This analytical rigor is a genuine strength and meaningfully increases confidence in the polysubstance findings.

Several limitations nonetheless deserve attention. This is a case series of 46 deaths—a small absolute number accumulated over five years. While the 0.9% case-fatality rate within the registry provides useful context, the denominator itself reflects only emergency department presentations with illicit drug-related toxicity captured by participating sites—an inherently selected population. Deaths that never resulted in an ED presentation (the majority, given the 67% out-of-hospital cardiac arrest rate) are structurally outside the registry's ascertainment window, meaning EDNAV likely captures the less immediately fatal end of the out-of-hospital death spectrum, or cases where resuscitation was attempted but ultimately unsuccessful. The registry is therefore better understood as a surveillance tool for near-fatal and fatal presentations reaching the ED than as a comprehensive population-level mortality registry.

Detection of a substance by LC-QTOF-MS confirms exposure but does not establish causation or contribution to death. The analytical findings describe toxicological co-occurrence, not necessarily the pharmacological driver of each fatality. Relative contributions of individual substances cannot be reliably apportioned from blood concentrations in post-arrest samples, which are subject to redistribution artifacts.

Geographic and demographic representativeness is also constrained by Victorian state-wide but not national scope, and by the pronounced male predominance (89%) in this fatal subcohort—a pattern that may reflect true epidemiological sex differences, differential healthcare-seeking behavior, or case-ascertainment patterns.

The recommendation to prioritize community-based interventions—particularly naloxone distribution given the heroin and benzodiazepine co-detection pattern—is well-supported by the data and consistent with international evidence. The opioid-benzodiazepine combination is a recognized high-lethality pairing where naloxone addresses one pharmacological arm but not the other, a nuance that merits explicit mention in harm-reduction messaging.

Overall, this is a useful, methodologically transparent contribution to real-world drug toxicity surveillance. Its value lies less in statistical power—the sample is too small for multivariable analysis—and more in demonstrating how a clinically integrated registry with rigorous analytical toxicology can characterize the phenotype of drug-related fatality in ways that inform targeted public health responses.

Relevance: Registry and Real-World Data Methods

Source: https://pubmed.ncbi.nlm.nih.gov/42716535/

---

Hip Arthroscopy in Sweden: Higher Socioeconomic Status and Substantial Regional Variation Revealed Through National Registry Data

Miriam G Wadström, Anders Stålman, Tobias Wörner, et al.

Summary: This nationwide Swedish registry study characterized 4,633 individuals undergoing hip arthroscopy (HA) between 2006 and 2018, comparing them with 23,060 age- and sex-matched controls. Individuals undergoing HA were more likely to be employed, higher-income, and more educated than the reference population, yet carried substantially greater compensated sick leave burdens both before and after surgery. Regional variation in HA incidence was pronounced, clustering around metropolitan areas.

Why it matters: The socioeconomic skew in HA utilization raises equity questions: is access to this procedure genuinely driven by clinical need, or by the ability to navigate specialist care pathways? The finding that sick leave was elevated even before surgery suggests that occupational burden and functional impairment may be important drivers of care-seeking, but the substantially higher post-operative sick leave (31% vs. 8%) also highlights real workforce and economic implications. Regional variation in incidence—independent of clinical indication—signals potential inequities in access or, alternatively, geographic clustering of specialist capacity and referral cultures, neither of which is benign from a health systems perspective.

Expert perspective: This is methodologically sound descriptive epidemiology. Linkage of surgical records to multiple national administrative registers—covering education, income, employment, and compensated sick leave—is a genuine strength of the Nordic registry infrastructure and produces a dataset that would be nearly impossible to replicate in most other countries. The matched-cohort design is appropriate for its descriptive purpose, and the 13-year coverage provides reasonable temporal breadth.

Several limitations deserve attention. This is explicitly descriptive; causal inference is not attempted and should not be inferred. The association between higher socioeconomic status and HA utilization could reflect better access to specialist care, greater health literacy, more physically demanding recreational activities common in higher-income groups, or differences in symptom tolerance and care-seeking thresholds—the data cannot distinguish among these. Clinical information is notably absent: diagnosis, symptom severity, imaging findings, prior conservative treatment, and surgical indication are not captured in administrative registers, making it impossible to determine whether socioeconomic differences in utilization reflect appropriate triage or inequitable access.

The sick leave data, while striking, require careful interpretation. Pre-operative sick leave elevation likely reflects the functional impairment that prompted surgery; the post-operative elevation conflates expected recovery time with potentially prolonged rehabilitation or surgical failure. Without clinical outcomes data linked to these records, it is impossible to assess whether the post-operative sick leave burden represents appropriate convalescence or suboptimal results. The residence-based incidence measure may also not perfectly capture where care was actually received, given that Swedes may travel to metropolitan centers for specialist procedures.

The finding that higher-income, employed individuals dominate the HA population is consistent with evidence from other elective orthopedic procedures globally—not surprising, but worth documenting rigorously. The regional variation is perhaps the most actionable finding, as it suggests that practice variation—rather than epidemiological differences in hip pathology—may be driving utilization patterns. This warrants prospective investigation with clinical data, ideally through linkage to the Swedish Hip Arthroscopy Register or equivalent clinical registry, to determine whether regional variation correlates with outcomes. This is useful population-level descriptive work that raises important equity and health-systems questions without being positioned to answer them—which is an honest and appropriate contribution.

Relevance: Registry and Real-World Data Methods

Source: https://pubmed.ncbi.nlm.nih.gov/42776610/

---

Gestational Age Weakly Predicts Physical Activity Levels in Adolescents and Adults: A Registry Linkage Study from Norway

Kristina Anna Djupvik Aakvik, Silje Dahl Benum, Marjaana Tikanmäki, et al.

Summary: This Norwegian study linked accelerometer-based physical activity data from 12,896 participants in the population-based HUNT Study with birth records from the Medical Birth Registry of Norway. Using linear regression with fractional polynomials, the investigators found that shorter gestational age modestly predicted lower physical activity—particularly light physical activity among females and total physical activity among adolescents. Birth weight showed no independent association with physical activity outcomes.

Why it matters: Understanding whether perinatal factors influence long-term physical activity patterns has public health relevance, as activity levels are tied to chronic disease risk across the life course. The registry linkage approach demonstrated here—combining a high-quality population cohort with a national birth registry—represents a methodologically useful model for longitudinal life-course research. For rare disease registries, this design illustrates how birth record linkage can enrich phenotypic datasets without requiring prospective enrollment from birth. The findings themselves, however, are modest in magnitude and require careful contextual interpretation.

Expert perspective: The study's methodological foundation is solid: the HUNT Study is a well-characterized, large population cohort with device-measured (rather than self-reported) physical activity, and the Medical Birth Registry of Norway is one of the world's most complete birth registries, with nationwide coverage since 1967. The linkage of these two sources is the main scientific and operational contribution of this work.

Effect sizes are small. The reported regression coefficients—for example, B = 1.53 MET-min/day per additional gestational week for light physical activity in the full sample—are statistically significant but clinically trivial across the range of gestational ages represented. The sample spans 13 to 52 years of age and encompasses both term and moderately preterm births; truly extreme prematurity (the group most studied in prior literature) is likely underrepresented, since very preterm infants born in earlier decades had higher mortality and may have been systematically absent from the adult cohort. This survivorship consideration is not fully addressed.

Confounding remains a concern. The analysis adjusts for age and sex, but gestational age is correlated with a range of maternal, socioeconomic, and perinatal factors—preeclampsia, multiple gestation, smoking, socioeconomic status—that independently predict physical activity trajectories. In observational registry-linked designs, residual confounding is almost always a material threat.

The age range is unusually wide. Pooling 13-year-olds and 52-year-olds into a single analytical sample with gestational-age exposures that occurred between roughly 1970 and 2010 introduces secular trends in obstetric practice, gestational age assessment (ultrasound versus last menstrual period), and population physical activity norms that could confound associations. The birth weight null finding is noteworthy and somewhat at odds with a fetal programming hypothesis; the authors interpret this as consistent with literature suggesting gestational age rather than size-for-gestational-age is the key determinant, but this deserves more scrutiny given known correlations between these variables.

From a registry and real-world data methods standpoint, this paper is primarily interesting as a demonstration of linkage feasibility and analytic approach rather than as definitive epidemiology. The fractional polynomial modeling is appropriate for detecting non-linear effects, and the finding of broadly linear associations is a useful result. The accelerometry protocol—two tri-axial devices worn for seven days—is a methodological strength relative to questionnaire-based studies.

Overall, this is a carefully conducted but incremental study. The associations are consistent with prior literature on preterm birth and physical activity, the effect sizes are small, and causal inference is not warranted. The registry linkage methodology is the primary contribution.

Relevance: Registry and Real-World Data Methods

Source: https://pubmed.ncbi.nlm.nih.gov/42776930/

---

Editor's takeaway

What stands out across this edition is how much rigorous registry science is being conducted outside the United States—and how much the quality and scope of that work depends on the underlying data infrastructure. The two Chinese studies illustrate what becomes possible with large-scale, nationally coordinated cancer and cardiac arrest registries: sample sizes that dwarf what most single-country rare-disease programs can achieve, yet persistent vulnerabilities around clinical covariate completeness and residual confounding that limit causal inference regardless of scale. The Australian, Swedish, and Norwegian studies each reflect the distinct advantages of smaller, higher-linkage national ecosystems—where administrative records, birth registries, and clinical cohorts can be routinely joined to answer questions that require decades of follow-up and population-level denominators. Together, the five studies reinforce a theme that applies equally to CF and rare-disease registries everywhere: a large and well-curated registry creates the conditions for meaningful real-world evidence, but it does not by itself resolve the fundamental challenges of confounding, incomplete clinical data, and the gap between statistical association and causal inference.