Obesity paradox in subarachnoid hemorrhage: a systematic review

As the number of obese people is globally increasing, reports about the putative protective effect of obesity in life-threatening diseases, such as subarachnoid hemorrhage (SAH), are gaining more interest. This theory—the obesity paradox—is challenging to study, and the impact of obesity has remained unclear in survival of several critical illnesses, including SAH. Thus, we performed a systematic review to clarify the relation of obesity and SAH mortality. Our study protocol included systematic literature search in PubMed, Scopus, and Cochrane library databases, whereas risk-of-bias estimation and quality of each selected study were evaluated by the Critical Appraisal Skills Program and Cochrane Collaboration guidelines. A directional power analysis was performed to estimate sufficient sample size for significant results. From 176 reviewed studies, six fulfilled our eligibility criteria for qualitative analysis. One study found paradoxical effect (odds ratio, OR = 0.83 (0.74–0.92)) between morbid obesity (body mass index (BMI) > 40) and in-hospital SAH mortality, and another study found the effect between continuously increasing BMI and both short-term (OR = 0.90 (0.82–0.99)) and long-term SAH mortalities (OR = 0.92 (0.85–0.98)). However, according to our quality assessment, methodological shortcomings expose all reviewed studies to a high-risk-of-bias. Even though two studies suggest that obesity may protect SAH patients from death in the acute phase, all reviewed studies suffered from methodological shortcomings that have been typical in the research field of obesity paradox. Therefore, no definite conclusions could be drawn. Electronic supplementary material The online version of this article (10.1007/s10143-019-01182-5) contains supplementary material, which is available to authorized users.

The PICO format was also used to determine inclusion criteria: aneurysmal, nontraumatic SAH patients (P), reported obesity variable (I) and reference group for comparison (C) were all mandatory for inclusion. Additionally, mortality (O) was defined as an outcome measure. Case reports, case series, letters, commentaries, book chapters and animal studies were excluded. Descriptive studies without calculated odds ratios (ORs), hazard ratios (HRs) or risk ratios (RRs) were also excluded.
English was the language used in literature searches.

Quality of selected studies
The quality of selected studies was evaluated systematically by the Cochrane Collaborator Handbook 1 and Critical Appraisal Skills Program (CASP) 2 . Based on the checklists, we created six specific domains in which study biases and methodological shortcomings were most likely and influential. Depending on whether the study fulfilled the criteria, studies were classified into low, unclear or high-risk-of-bias by each domain. To reach the overall classification of high-quality, all six domains had to fulfill low-risk-of-bias criteria.

Sudden-death SAHs
Since half of fatal SAHs occur before hospitalization, there is a risk for selection bias in studies using hospital-based registers. Since the risk factor status seems to be the worst among SAH patients dying before hospitalization, inclusion of sudden-death SAHs was required to reach low-risk-of-bias in this domain.

Obesity measurement
The second domain evaluated the potential risk for measurement bias. To reach the classification of low-risk-of-bias, obesity had to be measured by validated and generally known variables, such as BMI, and the measurements had to be performed at the time of SAH (for example, at admission).

Obesity analysis
Since obesity and SAH-related mortality have a possibly nonlinear association, combining extreme weight categories into a normal weight category or analyzing obesity only as a continuous variable may also bias analyses. Thus, we created the domain to evaluate valid analysis for obesity variables. To reach the classification of low-risk-of-bias, different categories for underweight, normal weight, overweight and obese/morbidly obese (e.g., according to WHO guidelines) were required.

Comprehensive short-term follow-up
Since the majority of fatal SAHs seems to occur within the first month 3 , we selected comprehensive short-term follow-up as a domain to avoid possible detection bias.
Either in-hospital, 30-day or a comparable follow-up time was required to achieve the classification of low-risk-of-bias.

Confounding control
Because consideration of all general risk factors for short-term mortality is overly challenging, we decided to focus on factors that have been previously associated with both higher/sudden SAH mortality and obesity. Smoking and high blood pressure are reported by high-quality studies 4-6 to interfere with both factors; therefore, to reach the classification of low-risk-of-bias in the domain of confounding control, these two variables needed to be adjusted for in multivariate analysis. Even though several other variables, such as SAH severity, amount of bleeding, aneurysm location and size, intracerebral hemorrhage, rebleeding and multiple post-operative complications, have been related to higher mortality after SAH, we excluded them in our quality assessment due to a lack of evidence suggesting that these variables associate with obesity categories.

Sufficient sample size
Since the theoretical protective effect of obesity is unlikely very strong, we included sufficient sample size as a domain to evaluate possible biases. Calculations for significant results were based on directional power analysis. Details are described under the subtitle Statistical analysis.