Idiopathic intracranial hypertension in the British population with obesity

Various studies have found that obesity, defined as a BMI > 30 kg/m², is associated with IIH [5, 23]. Indeed, the prevalence of IIH in this study population, restricted to patients with a diagnosis of obesity only, was estimated as being 262/100,000, which is significantly higher than the previously reported general population prevalence of 10.9/100,000 [21]. There may be a number of reasons for this. Firstly, it may, in part, be due to the inclusion criteria applied to the baseline population. This study examines databases specifically pertaining to patients with a background of obesity, in whom the prevalence of IIH may be intrinsically higher than that of the general population; however, previous studies have previously explored similar “concentrated” subsets, with Raoof et al. [21] reporting the highest subgroup prevalence of 85.7/100,000 in women with obesity. This is still lower than that found in the present study. A further explanation for this may be related to stringency of diagnostic criteria. The CPRD records are GP-based records where the diagnostic criteria applied may be variable and at times loose; though the modified Dandy criteria are most often used as the gold standard for diagnosis and most patients are likely to have been diagnosed in the secondary care setting and followed-up in primary care. Finally, another important explanation for this is that with the nature of recording in the CPRD database, most patients will not have an “end” to their diagnosis. Because of this, all patients diagnosed during the 25 years of follow-up will contribute to the “prevalence” at the end of the database entries. This may therefore overestimate the prevalence as some of the patients entered may not indeed still have the disease at this time.

Very few studies have used large-scale databases to explore the factors associated with IIH, but the association between IIH and BMI is perhaps the most widely investigated [5, 23]. It has previously been reported that as many as 90% of patients with IIH are obese [26], and many studies have found associations between IIH and female gender [7] and morbidities such as PCOS, which are inherently associated with obesity and weight gain themselves, and therefore likely to be linked to BMI rather than play a direct causal role [10]. Our study, carried out on exclusively obese patients, did not find any such relationship with female gender and PCOS; this may have been due to the fact that as many as 67.3% of the obese patients in the CPRD are women. It is therefore possible that the findings of previous studies are at least partially due to the higher rate of obesity in females. The selection criteria for this study was designed to specifically address the presentation of the disease in patients with obesity; this may have effectively annulled the “obesity bias” that can be present when looking for associations with IIH in the general population without adjusting for BMI, especially when considering patient cohorts too small to carry out sensitivity analyses or multivariate models; however, the lack of this association may also arise from diagnostic inaccuracy and the lack of central specified diagnostic criteria for IIH within the dataset. As mentioned, this is a potential limitation of the study that arises from its retrospective nature.

The association between IIH and BMI was thoroughly explored in a large case-control study by Daniels et al. [5], which found a higher incidence of IIH in patients with a higher BMI when the patients were stratified on the basis of BMI categories of < 25, 25–29, 30–35 and > 35 kg/m². The results of this study showed a linear trend, which has led to the general conclusion that the incidence of IIH is likely to increase linearly with BMI.

However, our findings indicate that this linearity does not continue when considering BMI classes above 30–35 kg/m²; there may therefore not be a linear or “dose-response” relationship between IIH and BMI above a certain threshold. This tapering off may be partially related to the finding of Daniel et al. that rapid weight gain is more closely associated with IIH than weight itself, which is why IIH can oftentimes be diagnosed in patients without obesity, but after a moderate-to-high weight gain. Weight fluctuations are more frequent in patients with a BMI on the lower end of those considered in this study, in whom physical activity and motivation are more likely to be preserved. Furthermore, a significant proportion of the patients reporting weight gain in the study by Daniel et al. stated that it was related to pregnancy, and BMI is unlikely to increase to extreme levels such as > 60 kg/m² over a period of 9 months.

This study found an independent association between IIH and anemia in patients with obesity. This is in line with finding of one consecutive case note review that eight out of 77 patients with a clinical diagnosis of IIH had documented microcytic anemia [2]. The symptoms of seven of these patients were resolved by simply correcting the anemia, thus highlighting the importance of obtaining full blood counts in patients with IIH and starting appropriate treatment if they prove to be anemic. Furthermore, the symptoms of two IIH patients in an interventional case series who did not respond to other treatment improved after blood transfusions and iron replacement therapy [17].

As in previous studies [12], our univariate analysis revealed an association between steroids and IIH but, as this disappeared in the multivariate analysis, it may be an insignificant association. It has previously been reported that there is an association between IIH and steroid withdrawal [14, 18], but we did not investigate whether the disease was diagnosed during or after steroid treatment. We did not find any association between hormone replacement therapy and IIH, but there was an independent relationship between IIH and the use of NSAIDs. This may not be causal because patients with IIH suffer from frequent and debilitating headaches for which the mainstay of treatment is analgesia, but a possible alternative mechanism underlying the relationship may involve the reduced glomerular filtration and fluid retention related to the use of NSAIDs. Further work is certainly warranted to explore this relationship in more detail. Furthermore, the specific role of other medications that were not investigated in this study—such as Vitamin A derivatives, oral contraceptive pills, tetracyclines, and other analgesics—is an important field for further research.

This study has some limitations. The CPRD is a purely retrospective database and may include errors related to losses to follow-up, incorrect entries, variations in terminology, and observer bias. Furthermore, the data and patient histories depend on precise consultation notes and documentation, which may not be complete, and the analysis includes data from patients in a single country and so the findings may not be globally generalizable. The clinical diagnoses entered in the CPRD are also dependent on consultation. Patient records may either be updated by practice employees following letters from the hospitals (which are generally copied to general practitioners) or by general practitioners themselves upon follow-up visits in primary care. It is, however. Clearly, this implies that there is room for loss of follow-up, with some patients receiving a diagnosis which is not documented in primary care records if never followed-up in that setting. Furthermore, there is a time delay between initial diagnosis in secondary or tertiary care; and the updating of patient records themselves. The date of diagnosis in the CPRD may therefore not correspond to the date of true diagnosis. Furthermore, data on the BMI of patients is reliant on regular follow-up within the primary care settings, including weight measurement. The frequency of this may be variable from patient to patient, and from practice to practice.

Perhaps the most important limitation to this study rests in the patient cohort restriction, which prevents generalizability to the overall population. The patients considered all had a BMI > 30 kg/m². This was done in order to explicitly explore the trends in IIH within the population with obesity; however, this restriction may give rise to selection bias, otherwise known as “collider” bias [3], a topic which has been widely investigated especially in research in the field of obesity in relation to “obesity paradox” studies. Though stratification has been shown to introduce bias in studies, it is important to note that the size of bias introduced by colliders has been shown to be small relative to the causal relationships between the variables [22], and is significantly lower than that present when not adjusting for confounding variables to avoid the risk of a “collider” [11] as may have been the case in studies that lacked adjustment for BMI when observing trends related to IIH. Relevant to this particular work, Pizzi et al. carried out a simulation to evaluate the effect of stratification by collider factors within epidemiological cohort studies restricted to one population stratum for analysis, as this study. Very limited bias was observed [20] and the results have been further validated in numerous studies since then [27, 28]. Therefore, though an important limitation to the study, the restriction to BMI > 30 kg/m² is unlikely to have introduced significant bias in the results, and does not impact the validity of the findings when considered within the context of patients with obesity only. On the other hand, the stratification by BMI is likely to have controlled for bias that has previously been reported by associating covariates that are parent variables to BMI and not directly to IIH, without adjustment for the effect being mediated by BMI as a descendant, or intermediate variable; however, as a consequence of the selection strategy, it is important to interpret these results not as crude associations between the variables (i.e., anemia and IIH) in the general population, but as associations specific to the strata of BMI studied here, therefore restricted to population with BMI > 30 kg/m². We must therefore consider that there is a possibility that the associated variables of BMI, anemia, and NSAID use may not have a common ancestor with IIH or direct causal relationship outside the specific population with obesity.