Safety and efficacy of autologous cell-based therapies for autism spectrum disorders

Discussion

The heterogeneity of the published therapies, particularly in separation and reported cell characteristics, limits comparability between studies. The reasons for the divergent efficacy results remain unclear, and several factors may contribute. Each approach has specific advantages and limitations.

Cell source: cord blood versus bone marrow

• Umbilical cord blood [1–5] contains young, immunologically naive cells and requires no harvest procedure: the infusion is simple and needs no anaesthesia. However, it is only available if a unit was banked at birth, the cell number is fixed and decreases per kg body weight as the child grows, and treatment usually cannot be repeated. Cryopreservation requires DMSO, which can cause transient side effects.
• Bone marrow [6–16] is available to every patient, and harvesting can be adapted to body weight and repeated. Its drawbacks are the need for bone marrow aspiration under sedation or general anaesthesia and a cell yield that varies between individuals and with age.

Processing: BMMNC, BMAC and culture-expanded MSCs

• Bone marrow mononuclear cells (BMMNC) [6–11] are isolated in a laboratory by density-gradient centrifugation (e.g. Ficoll). This yields a defined mononuclear fraction of lymphocytes, monocytes and haematopoietic progenitor cells (CD34⁺), with few MSCs, while red cells, granulocytes, platelets and plasma are removed. Harvest and application can take place on the same day, but the method requires laboratory infrastructure and more processing time.
• Bone marrow aspirate concentrate (BMAC) [12–15] is produced by automated centrifugation in a closed system directly in the operating room. Harvest, concentration and application take place in a single session under one anaesthesia (one-step procedure). The concentrate contains all nucleated cells, including granulocytes and platelets, and the plasma supernatant, which is rich in cytokines and growth factors, can be returned intravenously [13, 15]. Because the concentrate is applied immediately, cell counts are less precisely defined before application.
• Culture-expanded MSCs [16] allow a high and well-characterised dose of a single cell type. They require bone marrow harvest, several weeks of GMP laboratory culture and separate application procedures. In [16], this meant three general anaesthesias (aspiration and two intrathecal injections), premedication with hydrocortisone and an antihistamine, prophylactic intravenous antibiotics and a 72-hour hospital stay per injection. Culture-expanded MSCs also lack the monocytes and haematopoietic progenitors contained in BMMNC and BMAC.

Route of administration

• All cord blood studies used intravenous infusion [1–5]. Intravenous application is simple and needs no lumbar puncture, but it has two main disadvantages. First, a large proportion of cells is retained in the lungs, liver, and spleen (pulmonary first-pass effect). Second, the few cells that remain in circulation must cross the blood–brain barrier, which largely prevents cell entry into the central nervous system. Effects may therefore be mediated mainly by systemic immunomodulation or by paracrine factors (e.g. cytokines, extracellular vesicles) rather than by cells acting directly within the brain. Intrathecal application bypasses both obstacles by delivering cells directly into the cerebrospinal fluid.
• Intrathecal application [6–16] delivers cells directly into the cerebrospinal fluid but requires a lumbar puncture, usually under sedation in children, and is associated with headache, back pain and rarely seizures. A combined intrathecal and intravenous application [13, 15] aims to address both the central nervous system and systemic immune dysregulation.
• However, the MSC trial [16] also used the intrathecal route and showed limited efficacy, so the route alone does not explain the differences. Possible reasons include the older age of participants (5–15 years), a cell type without monocytes and haematopoietic progenitors, a small sample size (n = 32), and the fact that both groups received risperidone and rehabilitation, which may have masked an additional effect.

Number of applications

• Several studies applied cells repeatedly, two to five times [8, 10, 12, 14–16]. In the largest BMAC cohort, the greatest improvement occurred after the first procedure [15]. The optimal number and interval of applications have not been established.

Cell dose

• Cord blood provides considerably fewer cells per kg body weight than bone marrow preparations, and its composition differs. In the cord blood studies, about 1.6–3.8 × 10⁷ TNC/kg were infused, with a median of only about 0.3 × 10⁵ CD34⁺ cells/kg [1]; in the bone marrow trials, about 4.2–4.8 × 10⁷ MNC/kg and 2.1–2.6 × 10⁶ CD34⁺ cells/kg were applied per procedure [8, 10].

Patient selection

• Subgroup analyses suggest that younger children and children without intellectual disability or with less severe ASD respond better [1, 3, 5, 10]. Heterogeneous study populations may therefore dilute treatment effects.

Study design and the ethics of sham procedures

• The negative trials were randomised and partly blinded [2, 3]. The positive randomised trial [10] was open-label, without sham treatment of the control group. Expectation effects may therefore contribute to the differences, particularly on parent-reported scales.
• At the same time, sham-controlled designs are ethically highly questionable for bone marrow-derived therapies: a sham bone marrow aspiration and lumbar puncture under anaesthesia would expose children to procedural risks without any potential benefit. A placebo control is more easily justified for intravenous cord blood, which is already banked and requires no additional intervention [2, 3]. Alternatives include blinded outcome assessment, delayed-treatment or crossover designs, and objective biomarkers such as EEG, eye-tracking or cerebrospinal fluid cytokines [3, 14].