More information about text formats
We had an opportunity to read the article by Lakomkin et al regarding systematic literature review of LVO prevalence. Since one of our studies is part of this review we feel compelled to comment on the paper. We do appreciate the authors’ efforts in conducting this analysis which is important in understanding the burden of disease – but, with respect offer some criticisms. The major limitation of the paper which the authors recognize is the heterogeneity of the included studies. Unfortunately, this limitation is so critical that it yields unreliable information at best and misleading at worst.
The paper intends to study the prevalence of large vessel strokes. However, apart from a couple of population based studies in their review, the rest are a heterogenous mix describing an LVO rate from very selective cohorts of patients from single centers. Several are centered around validation of clinical scales for detecting LVOs. The key features of a population based study include a defined catchment population, access to a large part of that population and a reliable marker of disease. Without these a “prevalence” constitutes a report of a center’s experience of disease rate as it pertains to their patient intake. While still valuable it is not an estimation of the disease burden in the population that the center serves unless an overwhelming majority of that population comes to that center.
The authors determine an average rate of about 30% LVO amongst acute isch...
The authors determine an average rate of about 30% LVO amongst acute ischemic stroke (AIS) patients based on their review. The critical factor here is the denominator, i.e. the number of AIS patients from which the LVO rate is derived. It can be misrepresentative to extrapolate disease rate to a larger denominator derived from a different methodology. For instance, the oft quoted denominator of about 700,000 ischemic strokes is based on population studies using very specific ICD discharge codes in well-defined populations. Examples include the GCNKSS and the BASIC projects. Unless a study uses the same discharge codes in its methodology for determining the AIS denominator it cannot transplant LVO percentages calculated from its selective cohort to the larger denominator and derive absolute numbers of disease. For instance, a 30% LVO rate in a cohort of patients assessed as having an ischemic stroke by the EMS is not the same as 30% LVO rate amongst all AIS based on specific ICD discharge codes. Using the percentage from one cohort and applying to a different denominator could yield inaccurate absolute numbers.
The study from our center that is included in this analysis was designed to capture the LVO incidence in a unique well defined population of which the vast majority (85%) was served by our hospital system based on each county data reported to the DHHR. Thus, similar to the studies used to derive the total AIS estimates (e.g. GCNKSS, BASIC) we had a defined population, had significant access to the population and used the same ICD discharge codes to determine the denominator as in those studies. We also used a reliable marker of LVO, i.e. CTA performed on every AIS patient as identified by these codes. In our first paper an LVO was defined as ICA-T, MCA and BA to restrict it predominantly to the occlusion sites considered as LVO in the major clinical trials. In our second paper, not included in this analysis, we separately determined an incidence of M2 occlusions and combined with ICA-T, M1, BA this yields a rate of 16% (95CI 14-17). In our population, this gives an incidence of 31 (95CI 26-35)/100,000 people/year. Our second paper also includes a chart showing the location of occluded sites for other vessels not considered as LVOs i.e. ACA, PICA, SCA, PCA etc.
The authors comment on our inclusion of TIA codes in the denominator. We did that because it had been part of previous population studies evaluating AIS incidences. TIA comprised at most 1% of our denominator and if we exclude these patients, the incidence of 16% LVO does not change by more than a percentage point. Another variable to consider in a single center report is that a tertiary level comprehensive center may get transfers of sicker stroke patients from referring hospitals which can further skew the denominator and hence the derived LVO rate.
In summary, it is important to differentiate between population studies and single center experiences – especially when compiling these in a systematic review. It is critical when extrapolating disease incidence from one center’s report to the national disease burden that the methodology of deriving the larger denominator is the same. Nonetheless we do appreciate this review and commend the authors on their efforts. This highlights the need for collaborative efforts to collect population data based on a uniform methodology. These efforts should include state and federal registers that collect health data based on specific disease codes. Such collective ventures can provide more realistic estimates of the LVO burden and help shape systems of care.