Disease progression in women with X-linked adrenoleukodystrophy is slow

The rare inborn error of metabolism X-linked adrenoleukodystrophy (ALD, Online Mendelian Inheritance in Man entry number 300100) affects both men and women [1]. In men, the clinical spectrum includes progressive spinal cord disease in all (‘adrenomyeloneuropathy (AMN)’), primary adrenal insufficiency in 80% and cerebral inflammatory disease (‘cerebral ALD’) in 60% [2–4]. As ALD is an X-linked disease, women were previously considered asymptomatic carriers. It is now known that even though adrenal insufficiency and cerebral disease occur in less than 1% of women, more than 80% eventually develop progressive spinal cord disease [5, 6]. Although both men and women develop spinal cord disease, there are differences. In women the onset of spinal cord disease is usually later in life. Furthermore, although there are no prospective studies that have evaluated this systematically, progression is considered to be slower [5]. Current treatment options for spinal cord disease are merely supportive, however, new curative therapies are under development [1]. As ALD is a rare disease, the number of patients who can participate in clinical trials is limited. It would increase the number of patients substantially, and thus speed up drug development, if both men and women could participate, despite their clinical differences.

If women with ALD are to participate in clinical trials, progression rate of spinal cord disease and factors that determine this rate of progression need to be identified. Outcome measures usable in clinical trials should be sensitive enough to measure clinical deterioration in women over a reasonable period of time, yet they should also be clinically relevant. Previous cross-sectional research has shown that women with ALD who have spinal cord disease can be clearly distinguished from those who do not, using various clinimetric scales. These scales include the Japanese Orthopaedic Association (JOA), the Severity Score system for Progressive Myelopathy (SSPROM), the Expanded Disability Status Scale (EDSS) and the AMC Linear Disability Scale (ALDS) [5, 6]. The quality of life questionnaire Short Form (36) Health Survey (SF-36) has shown a similar trend [5]. Habekost et al. [7] provided the first longitudinal data for the JOA and SSPROM, however, change on these scales was minimal over a mean observation period of 9 ± 3 months and therefore probably not practical for use in clinical trials. Recently, Schirinzi et al. [8] illustrated clinical change (+ 0.24/year) on the Adult ALD Clinical Score in 19 symptomatic women with ALD over an observation period of 3.5 ± 2.1 years. Follow-up data from large cohorts on other outcome measures is currently unavailable. Similarly, no information exists on factors that influence the rate of progression of spinal cord disease in women with ALD, but we hypothesize that age and duration of symptoms might influence progression rate, as a positive correlation has been identified between (1) symptomatic status and age, and (2) duration of symptoms and gait disorder severity [5, 6, 8, 9].

In addition to clinical differences in spinal cord disease between men and women with ALD there is a biochemical difference. Both have a mutation in the ABCD1 gene, which encodes the ALD protein (ALDP), a peroxisomal membrane protein essential in the beta-oxidation of straight-chain very long-chain fatty acids (VLCFA; ≥22 carbon atoms) [10, 11]. Measurement of total VLCFA in plasma (specifically C26:0, the C26:0/C22:0 ratio and the C24:0/C22:0 ratio) is diagnostic in men with a nearly 100% sensitivity [12–14]. In 15–20% of women however, VLCFA levels are in the normal range [5, 14]. In the absence of elevated VLCFA an ALD diagnosis can be made by identifying a known pathogenic ABCD1 mutation or by time consuming functional studies [15]. Recently our group reported that 1-hexacosanoyl-2-lyso-sn-3-glycero-phosphorylcholine (C26:0-lysoPC) is a better diagnostic biomarker in women than C26:0 [16]. C26:0-lysoPC levels were elevated in all 49 women even though C26:0 was not. Unfortunately, the difference between the maximum control C26:0-lysoPC level and the minimum patient C26:0-lysoPC level was small, warranting the need for a superior discriminating biomarker. If such a biomarker was identified, this would enable timely diagnosis for women with normal VLCFA levels and ABCD1 variants of unknown significance. Semi-targeted lipidomics, an ultra-high performance liquid chromatography coupled to high-resolution mass spectrometry (UPLC-HRMS), allows the detection of over 10,000 lipid derivatives in one screen [17, 18]. If these lipids can distinguish between relevant clinical groups, for instance patients and controls, they could prove to be better diagnostic biomarkers.

The purpose of this follow-up study in women with ALD was to assess whether EDSS, ALDS and SF-36 can detect progression of spinal cord disease and to model the effect of age and the duration of symptoms of spinal cord disease on the rate of progression. Moreover, we performed a pilot study to assess if a semi-targeted lipidomics approach can identify possible new diagnostic biomarkers for ALD in women.

In the largest follow-up study in women with ALD to date, we provide evidence that during a follow-up period of almost 8 years the EDSS, but not the ALDS and SF-36, can detect progression of spinal cord disease, although this progression is below the rate that is generally considered as clinically relevant for clinical trial design [30, 31]. Moreover, age and the duration of symptoms seem positively associated with the rate of progression.

The significant progression that the EDSS detected is supported by our clinical observation that eight women became symptomatic during the follow-up period and the ambulation status (i.e. unrestricted walking, restricted walking and assisted walking) altered in three. Moreover, there were women who developed incontinence for urine, incontinence for feces and sensory complaints/abnormal sensation examination during the follow-up period. Only the EDSS detected this clinical change as a significant – albeit minor - increase in EDSS score over time with 0.73 points. In contrast to the SF-36 and ALDS, the EDSS score captures the degree of incontinence and sensory abnormalities in the final score. Interestingly, subgroup analysis 2 (with symptomatic women) also detected this progression, but the increase was smaller (+ 0.51 points). This is supported by the observed difference in annual EDSS progression rates, which were calculated with outcome measures not adjusted for covariates. For women with two assessments (n = 32) progression rate per year was 0.08, for women symptomatic at baseline (n = 19) 0.06 and women asymptomatic at baseline (n = 15) 0.17 (Table 5). As abnormal neurological signs in the absence of symptoms contribute to the EDSS score, this most likely reflects progression of disease in presymptomatic women. In addition, slow clinical progression on the EDSS is supported by work of Schmidt et al. (2001), who detected no progression on EDSS after 4 years in 8 women, and Habekost et al. (2015) who reported significant but slow progression on the Severity Score system for Progressive Myelopathy, a myelopathy scale that – similar to EDSS – incorporates both symptoms and abnormal neurological signs [7, 9].

Furthermore, there are theoretical considerations that reinforce the interpretation of our results. ABCD1 deficiency with defective ALDP and subsequent VLCFA accumulation are thought to be responsible for the axonal degeneration underlying the clinically detectable spinal cord disease in ALD patients [1]. Using ALD knockout mice, Gong et al. (2017) recently showed that primed ABCD1-deficient microglia are likely involved in the pathophysiology of spinal cord disease in ALD [32]. Early signs of this priming are already detectable in postnatal mice [32]. Axonal degeneration probably starts early and slowly progresses during life until it becomes clinically detectable later in life. As there have been no reasons so far to assume nonlinear progression of spinal cord disease, this supports the slow clinical deterioration we observed in our cohort [7].

Nonetheless, various uncertainties in the interpretation of our results remain. The cohort was small and heterogeneous, as the youngest symptomatic woman was 36 and the oldest asymptomatic woman was 73 years of age. The use of generalized linear mixed models allowed us to increase the cohort size because it enables the inclusion of women with only one assessment. The 19 newly identified women were included as baseline assessments and not as follow-up assessments. The age distribution of these 19 women resembled the remaining cohort at baseline more than it did at follow-up (Fig. 2). Their median age, however, was slightly higher than the group with two assessments (n = 34). As the percentage of symptomatic women increases with age, adding the 19 women as a baseline assessment could diminish the contrast between baseline and follow-up, subsequently underestimating the progression of spinal cord disease [5]. This was not supported by subgroup analysis 1 (with women with two assessments), as timing of measurement did not become a significant main effect for other outcome measures than the EDSS - for which timing was already a significant main effect. Furthermore, there were two women who could not visit the hospital, potentially causing selection bias. If the severity of spinal cord disease was the reason for not being able to visit the hospital, that could have generated an underestimation of disease progression, as the severely affected women were then not included in the follow-up assessments. In addition, theoretically, differences in symptomatic therapy (i.e. physical therapy, spasmolytics or anticholinergic medication for urge incontinence) at baseline and follow-up could have influenced assessments. However, our clinical observation is that efficacy of these treatments is limited making it unlikely that this is a relevant confounding factor.

Although the EDSS could detect significant disease progression, the ALDS and SF-36 were not sensitive enough to detect this change. As the subgroup analyses also did not show significant change, these suggest that progression was not underestimated due to adding women with only one assessment (subgroup analysis 1) or because women that remained asymptomatic during follow-up were included (subgroup analysis 2). Clinical validity at baseline was poor for all outcome measures. Even though they could differentiate between symptomatic and asymptomatic women, they could not discriminate between symptomatic women with restricted walking and aided walking. On the other hand, while the distinction between restricted walking and aided walking is straightforward, the tipping point of when a patient converts is not. When a patients chooses to start using a walking aid is subjective and is different for each individual based on their personality traits and specific circumstances. Nevertheless, ideal outcome measures would be sensitive enough to detect differences between these groups. Although construct validity was good, inconsistencies amongst the various outcome measures remained. There were only five women who showed progression on all outcome measures and some even improved over time on the ALDS and the SF-36 subdomains physical functioning and physical component summary. As ALD is a slowly progressive neurodegenerative disease it is highly unlikely that improvement over time reflects the true natural history of spinal cord disease in women with ALD [1]. Despite somewhat poor sensitivity, analysis of the construct validity showed that measures that assess disability correlated strongly with those that assess neurological impairment. As described by us in a previous study the correlation between the physical and mental status is poor, thereby underlining the fact that quality of life can be good despite having a disability [5]. Information on other psychometric properties is lacking, and despite often being problematic in rare diseases it would be desirable to assess internal consistency and test- retest reliability in a larger cohort.

Besides significant progression of disease on EDSS, generalized linear mixed model results suggest that age and the duration of symptoms of spinal cord disease are positively associated with the rate of progression. Increasing age was associated with a higher degree of disability as assessed by the ALDS. In addition, the longer the duration of symptoms, the higher the EDSS score, the lower the ALDS score (indicating more disability) and the worse quality of life was as measured on the SF-36 subdomains physical functioning and physical component summary. These findings are in concordance with the findings of others, as age and the duration of symptoms of spinal cord disease have been associated with a higher degree of disability [5, 9, 7].

Moreover, aside from being the largest follow-up study in women with ALD to date, this is also the first study to use a semi-targeted lipidomics approach in plasma for the identification of new diagnostic biomarkers for ALD in women. Our approach generated a list of 100 potential biomarker ratios with strong differentiating properties and non-overlapping data distributions between women with ALD (n = 20) and controls (n = 10). As the selected group of women with ALD included five women with either a plasma C26:0 level or a C26:0/C22:0 ratio within the normal range, these results make us feel confident that our list of potential biomarker ratios will most likely contain a ratio with a better sensitivity than plasma C26:0 levels or the C26:0/C22:0. Using a comparable lipidomics approach Ruiz et al. (2015) reported similar differences in abundancy of lipids between 13 men with ALD and 13 controls [33]. The finding that the majority of the 56 potential new biomarkers belong to the phosphatidylcholine class was not unexpected. Already in the 90’s it was reported that in normal-appearing white matter the highest VLCFA excess was found in the phosphatidylcholine fraction [34]. After validation of our results in an external or independent cohort we will make a selection of ratios for further exploration based on feasibility, the availability of internal standards and absolute abundance of the individual metabolites. Thereafter a dedicated method can be put into place for use in clinical care.

To summarize, progression after 8-year follow-up was detectable in our cohort using the EDSS, but the change in EDSS score was small. These results have implications for counseling women with ALD. These women may be informed that progression is usually very slow, with significant change occurring over years or decades, although in individual patients onset may be early and the disease more progressive. Why some women become symptomatic decades before other women or which factors influence the rate of progression remains to be elucidated in future studies. Furthermore, including women in intervention trials with clinical endpoints evaluating new treatments for spinal cord disease in ALD remains challenging. Progression seems so slow that it cannot be detected by current outcome measures such as EDSS, ALDS and SF-36 unless a study lasts for at least 8 years, which is typically not feasible from a financial perspective. Perhaps clinical outcome measures primarily focused on gait (e.g. six-minute walk test or the spastic paraplegia rating scale), sensory abnormalities (e.g. semi-quantitative vibration threshold) or incontinence (e.g. International Consultation on Incontinence Questionnaire Female Lower Urinary Tract Symptoms) can detect a more significant change between baseline and follow-up, but additional, more sensitive, quantitative measures for progression of spinal cord disease are needed to detect disease progression during a smaller and thus more practical time period [19, 35–37]. Current candidates are optical coherence tomography and spinal cord DTI [38–40], however, these modalities remain to be validated in future studies.