Resumen de Applying sequential methods in order to detect and interpret a possible increased rate of myeloid malignancies in the girona province
Background: Based on preventive and precautionary dictates, the availability of procedures to detect increased morbidity rates in relatively small populations or clusters of rare diseases stands paramount. In this context, cost-aware, instructive and timely analyses must provide guidance thereby determining whether cases relate causally and the nature of their causal mechanisms. In addition, these analyses must indicate whether further assessments are needed and/or if any active environmental intervention in the pertinent community is required. Genuine aggregates of infrequent chronic diseases may be rendered biased or unnoticed by just measuring general secular trend statistics. Discretionary surveillance of these conditions entails the appearance of incidental clusters and costly supervening elucidation. In the foreground of the present study lies a report of ascending secular trends for leukaemias of myeloid lineage (MMs), in a province of 730,000 people. This warrants an inquiry plan aimed at unveiling causative clues. The main aim of this thesis is to demonstrate the use of a sequential-based approach in detecting an elevated rate that may have initiated at some unknown point of time during the study period.
Objectives: 1) To assess the temporal clustering of MMs through sequential procedures by waiting time. 2) To probe into the intensity, time pattern and overall place-dependent behaviour of any detected cluster. 3) To inquire into person variables to help informing possible causative paths in the communities bearing indicative clustering.
Methods: Analyses were based on the registered cases of the main MM categories occurring at preselected municipalities of Girona Province (Spain) over a 15-year period. Every municipality with at least 10 diagnoses for some category was selected; this totalled 35 series in 15 communities. Subtle clustering was validated using a cumulative score test (CUSCORE) that signals an alarm for any stretch during the analyzed diagnoses. A CUSCORE test hinges on the Relative Interval (RI) statistic, which reflects the waiting-time-to-event; this constitutes a continuous variable that controls for size and profile differences between populations and sub-periods. Following the CUSCORE test, a graphical display of the temporal pattern and a confirmation test were conducted. These procedures were preceded by ascertaining standardized incidence ratios (SIR). Occurrences of MM-subtypes within the temporal agglomerations of cases were assessed too.
Results: Eighteen series (11 communities) evinced excess of observed MM cases by the 15-year SIR. The RI-based sequential procedures unveiled the temporal patterns of the clusters over the multiyear period. Their detection yield even proved at sub-unity event counts per year and at mild-intensity epidemic rises. Six communities registered one or more indicative time clusters. There is no reasonable chance of observing more than one cluster of those in each community during 15 years. Sometimes 2 MM-categories overlapped for the epidemic spell. Depictions of waiting intervals evidenced embedded huddles within the series. Once a stretch of these 10 clusters was sensed, the focused analyses on morphological subtypes showed selective involvement by de novo subtypes in 8 of them. Some causative hints emanated from observed deviations from usual gender ratios or ages at diagnosis in the clustering communities. And so did 'within-cluster' subtype frequencies that were divorced from anticipated prevalence Conclusions: Addressed beforehand, upon paucity of occurrences, these sequential procedures worked usefully for the ad hoc test of alarm signal, post alarm deciphering of time pattern, and cluster confirmation. Where the descriptive epidemiology failed to show an increased secular trend, the aforementioned methods uncovered small community-based clustering and helped in tailoring realistic intensity estimates related to unobtrusive ballooning rates. Whereas the integrated assessments used herein focused on cancer agglomeration, this rather economical approach could be extended to other non-return maladies if high quality morbidity data derived from population-based registries are available. The insights, achieved as plausible causation understandings of the signalled clusters, hint what ought to be investigated at the next elucidation phase, and outreach to mending interventions and pre-empting preventable chronic maladies.
