Commentary: Statistical assessment of cancer cluster evidence--in search of a middle ground

ClustR: A Space-Time Cluster Analysis R Package for Individual-level Data

BACKGROUND

Until recently, large individual-level longitudinal data were unavailable to investigate clusters of disease, driving a need for suitable statistical tools. We introduce a robust, efficient, intuitive R package, ClustR, for space-time cluster analysis of individual-level data.

METHODS

We developed ClustR and evaluated the tool using a simulated dataset mirroring the population of California with constructed clusters. We assessed Cluster's performance under various conditions and compared it with another space-time clustering algorithm: SaTScan.

RESULTS

ClustR mostly exhibited high sensitivity for urban clusters and low sensitivity for rural clusters. Specificity was generally high. Compared with SaTScan, ClustR ran faster and demonstrated similar sensitivity, but had lower specificity. Select cluster types were detected better by ClustR than SaTScan and vice versa.

CONCLUSION

ClustR is a user-friendly, publicly available tool designed to perform efficient cluster analysis on individual-level data, filling a gap among current tools. ClustR and SaTScan exhibited different strengths and may be useful in conjunction.

Detection and estimation of the increasing trend of cancer incidence in relatively small populations

BACKGROUND

Detection and estimation of trends in cancer incidence rates are commonly achieved by fitting standardized rates to a joinpoint log-linear regression. The efficiency of this approach is inadequate when applied to a relatively low levels of incidence. We compared that approach with the Cuscore test with respect to detecting a log-linear increasing trend of chronic myelomonocytic leukemia (CMML) in datasets simulated to match a province of about 700,000 inhabitants.

METHODS

For better efficiency, we replaced the standardized rate as the dependent variable with a continuous statistic that reflects the inverse of the standardized incidence ratio (SIR). Both procedures were applied to datasets simulated to match published results in the Girona Province of Spain. We also present the use of the q-interval in displaying the temporal pattern of the events. This approach is demonstrated by analyses of CMML diagnoses in Girona County (1994-2008).

RESULTS

The Cuscore was clearly more efficient than regression in detecting the simulated trend. The relative efficiency of the Cuscore is likely to be maintained in even higher levels of incidence. The use of graphical displays in providing clues regarding interpretation of the results is demonstrated.

CONCLUSIONS

The Cuscore test coupled with visual inspection of the temporal pattern of the events seems to be more efficient than regression analysis in detecting and interpreting data suspected to be at elevated risk. A confirmatory analysis is expected to weed out 75% of the superfluous significant results.

Detection and investigation of temporal clusters of congenital anomaly in Europe: seven years of experience of the EUROCAT surveillance system

Detection and investigation of congenital anomaly clusters is one part of surveillance to detect new or changing teratogenic exposures in the population. The EUROCAT (European Surveillance of Congenital Anomalies) cluster monitoring system and results are described here. Monitoring was conducted annually from 2007 to 2013 for 18 registries covering an annual birth population up to 0.5 million births. For each registry and 72 anomaly subgroups, the scan "moving window" technique was used to detect clusters in time occurring within the last 2 years based on estimated date of conception. Registries conducted preliminary investigations using a standardised protocol to determine whether there was cause for concern, and expert review was used at key points. 165 clusters were detected, a rate of 3.4% of all 4823 cluster tests performed over 7 years, more than expected by chance. Preliminary investigations of 126 new clusters confirmed that 35% were an unusual aggregation of cases, while 56% were explained by data quality or diagnostic issues, and 9% were not investigated. For confirmed clusters, the registries' course of action was continuing monitoring. Three confirmed clusters continued to grow in size for a limited period in subsequent monitoring. This system is best suited to early detection of exposures which are sudden, widespread and/or highly teratogenic, and was reassuring in demonstrating an absence of a sustained exposure of this type. Such proactive monitoring can be run efficiently without overwhelming the surveillance system with false positives, and serves an additional purpose of data quality control.

Detection and investigation of temporal clusters of congenital anomaly in Europe: seven years of experience of the EUROCAT surveillance system

Detection and investigation of congenital anomaly clusters is one part of surveillance to detect new or changing teratogenic exposures in the population. The EUROCAT (European Surveillance of Congenital Anomalies) cluster monitoring system and results are described here. Monitoring was conducted annually from 2007 to 2013 for 18 registries covering an annual birth population up to 0.5 million births. For each registry and 72 anomaly subgroups, the scan "moving window" technique was used to detect clusters in time occurring within the last 2 years based on estimated date of conception. Registries conducted preliminary investigations using a standardised protocol to determine whether there was cause for concern, and expert review was used at key points. 165 clusters were detected, a rate of 3.4% of all 4823 cluster tests performed over 7 years, more than expected by chance. Preliminary investigations of 126 new clusters confirmed that 35% were an unusual aggregation of cases, while 56% were explained by data quality or diagnostic issues, and 9% were not investigated. For confirmed clusters, the registries' course of action was continuing monitoring. Three confirmed clusters continued to grow in size for a limited period in subsequent monitoring. This system is best suited to early detection of exposures which are sudden, widespread and/or highly teratogenic, and was reassuring in demonstrating an absence of a sustained exposure of this type. Such proactive monitoring can be run efficiently without overwhelming the surveillance system with false positives, and serves an additional purpose of data quality control.