Is there value in using randomized placebo controlled trials in neuromuscular disease?

References to placebos and the placebo effect have come down to us from ancient Egypt. The placebo effect on outcomes is considerable and increasing, even in neuromuscular disease (NMD) studies. A transient placebo effect was observed in the CHERISH study of Nusinersen for SMA types 2 and 3. In one of our previous placebo-controlled SMA studies the placebo controls’ muscles also strengthened [1]. As a result of positive placebo effects, there is lower separation between the outcomes of randomized double-blinded placebo-controlled trials (RDBPCTs), especially on subjective parameters. Up to 90% of new antidepressants, antipsychotics, pain medications, and even statins are demonstrating less and less benefits over placebos [2]. Further, unlike for infectious and some oncological diseases, there have been no medical cures for the relatively rare neuromuscular diseases (NMD), for which adequate recruitment to establish subtle differences is especially daunting and problems with placebos and randomization take on additional dimensions.

The RDBPCTs have become the basis of ‘evidence-based-medicine (EBM)’ and are ‘transforming medicine into a scientific discipline’ [3]. However, this comes at enormous cost, ethical ambiguities, and logistical difficulties. In addition, adherence to it is causing morbidity and mortality. Although EBM has been defined as, ‘the conscientious, explicit and judicious use of current best evidence in making decisions about the care of individual patients, whether RDBPCTs or not,’ [4] the Cochrane group goes on to classify ‘no controls, case-series only’ as Level V evidence that ‘ … can only hint at efficacy’ [5]. This ‘hint at efficacy’ is even taken for granted for interventions that take the place of vital bodily organs and functions that would be lethal if substituted for by placebos.

The primacy of RDBPCTs in EBM may be an overcompensation for previous abuses by which cures of muscular dystrophy were claimed by administering strychnine, electrical stimulation, and cod liver oil [6]. While France and the English colonies were provisioning naval vessels with citrus fruit since the early 1600s, over 100.000 British sailors were dying from scurvy. In 1747, James Lind performed a controlled study by prospectively provisioning a British naval vessel with citrus fruit while others had only the sulfuric acid solutions sanctioned by the British College of Physicians to balance the four bodily humors. In 1753, Lind published a 400-page ‘Treatise of the scurvy’ but was ignored. British physicians could not see how lemons could affect humoral balance despite the prospective-controlled study. For the next 40 years, tens of thousands more British sailors died and the colonies won independence with thanks to British physicians [7].

Today, we have the other extreme, that of accepting only evidence from controlled studies even when placebo controls would be lethal. Survival is impossible without breathing and coughing. Yet studies on interventions that substitute for them are routinely rejected from peer review journals and meta-analyses for lacking controls [8]. Rose et al. [9] in 2017 reviewed 2686 citations, ignored observational studies on patients who could neither breathe nor cough, and found only two small-controlled trial failures of mechanical insufflation exsufflation (MIE) that, in addition, was used at one-half of effective settings. They and others concluded ‘the non-randomized trials of MIE (and noninvasive ventilatory support or NVS) for NMD,’ were rated as ‘very low-quality evidence’ for ‘the usual obvious reasons’ [9–11], ignoring the fact that anyone who could never cough or take in air would asphyxiate and die. Further, successful extubation and decannulation of patients having no functioning respiratory muscles and 0 ml of vital capacity was also reported, obviously without controls, and only possible by using continuous NVS and MIE [12]. The systematic rejection and denigration of interventions that support life is no less ironic than considering the benefits of parachutes unproven for skydivers [13]. The result is clinicians not learning how to prolong quality survival of their patients without resort to harmful and unnecessary tracheotomies for the support.

Historically, advances in medicine occurred from observational studies. Even James Lind’s study was observational. Penicillin, and Streptomycin for tuberculosis, were never subject to RDBPCTs, nor, for that matter, were kidney transplants, hip replacements, invasive mechanical ventilation, or parachutes for gravitational challenge [13]. On the other hand, to demonstrate relatively minor statistically significant benefits, most often only slightly slower rates of deterioration, or noninferiority with drugs already available, even without further clinical benefits, RDBPCTs are necessary and power analyses mandate large cohorts for this. This has resulted in the average cost per approved new drug increasing from 70 USD million in 1970 to 2.4 USD billion in 2013 and is increasing by 8.5% per year above inflation. Single clinical trial costs are from 2.1 USD million for a 4 patient study on uridine triacetate, to 346.8 USD million for larger studies [14], and quite possibly well more than 1 USD billion for larger studies but drug companies are reluctant to discuss this proprietary information. Adequate cohort size is especially difficult to achieve and costly and may deleteriously delay therapies for rare NMDs. Clinical trials have cost a median of 41,117 USD per subject and 3562 USD per visit with costs highest when placebos are used [15].

Besides logistical difficulties, RDBPCTs are also difficult to interpret when NMD subjects, who are naturally getting stronger (pre-plateau), are randomized with those who are weakening (post-plateau). For example, normal physiological plateaus occur around age 20, following which, VC and organ function decrease by >1% per year. The VCs of patients with Duchenne muscular dystrophy (DMD) plateau at age 13 (range = 9–17); speed rising from the floor plateaus from ages 3 to 7, and 6 to 12 minute walk distances also optimize at varying ages. The DMD studies typically lump together patients over age 4, that is, pre- with post-plateau subjects. This can require very large cohorts to attain statistical significance and result in performance bias and randomization selection effects. Plateaus in function in NMD patients are variable. Thus, ‘change over time,’ the critical variable in RDBPCTs in NMD, is not easily plotted, and complicated by inability to quantitate placebo effects. The ultimate irony is that what we think we know about placebo effects is purely observational.

On the other hand, any parameter observed to improve on any patient post-plateau would be explained by a treatment effect. Large cohorts are unnecessary. Perhaps the greatest ethical dilemma for placebos in NMD is that many medications are far more beneficial the earlier that they are administered. Gene therapy and agents that modulate splicing of the SMN2 gene messenger RNA, are most beneficial when administered at or very shortly after birth. For placebos, their administration would be delayed and greatly compromize treatment effectiveness.

Biologic systems have multiple redundancies, or a biologic cushion, to achieve a functional task. If various bodily systems are degraded, then a non-linear relationship between parameters can become the dominant driver of success or failure to achieve a function. Thus, using ‘function’ as a NMD endpoint can be complex to evaluate as opposed to pain or a twitch response, because it is a highly integrated event. NMDs affect multiple bodily systems, not just strength. Hence, achieving a ‘function’ can have an erratic relationship to strength, in which very small changes may create disproportionate responses in function, or no response in function if the same % increase in strength is inadequate. For example, if a patient has a cough peak flow of 140, he may fail to clear secretions and get pneumonia. If MIE increases the flow to 180 L/m, this may or may not be adequate to avoid pneumonia. However, pathophysiological alterations of mucus characteristics, the immune and cardiac systems, ability to humidify sputum by nose or mouth breathing, coaguability, factors affecting mucociliary transport, and airway patency effects make the results of interventions on function non-linear.

Different NMDs can have very consistent or different disease evolutions. Patients with motor neuron disease (MND) can have extremely variable courses making significant changes in rate of deterioration very difficult to determine by RDBPCTs or observation. While this has been done for Radicava, its 150,000 USD price tag per year for a statistical one-third decrease in rate of functional deterioration, for a condition where spontaneous improvements occur, might make its use difficult to justify [16]. In 1995 Riluzole can onto the market to treat ALS at a 9500 USD annual cost for a statistical prolongation of tracheostomy-free survival of a few months. That seemed expensive at the time. Today there are 500,000 USD annual costs for medications for diabetes mellitus, DMD, Pompe disease, and SMA. Some medications for SMA, like Zolgensma, can cost 2.1 USD to 5.1 USD million for one injection; nusinersen for SMA costs almost 1 USD million the first year then about 400,000 USD annually with many patients showing no benefit from either.

Unlike for MND, SMA has consistent evolution. Untreated patients with SMA type 1 die before 18 months of age. No child with SMA type 1 with 2 copies of the SMN 2 gene will be able to sit independently or have a VC exceed 300 ml. Any such patient who undergoes tracheostomy will remain continuously ventilator dependent unless there is a treatment effect [17]. Yet, CNVS and MIE treated patients are surviving over 25 years with no volitional muscle movement and without tracheostomy tubes or medical therapies. Any child with SMA type 1 who attains independent sitting must have had a medication treatment effect. Placebo controls are unnecessary. Any post-plateau child with SMA or DMD whose VC or manual muscle testing grades increase over time must have had a treatment effect. Patients as controls in RDBPCTs should have data points before study onset to determine whether the plateau has or has not been reached so that any reversal of deterioration must be a treatment effect for any individual patient. We would argue that any medication that can not yield a post-plateau improvement in strength and/or function but that requires statistical analysis to demonstrate slight decreases in rate of deterioration is not worth the hundreds of thousands of dollar price tag necessitated by the cost of the RDBPCTs used to study it.

Receiving nusinersen since infancy would have amounted to a $12 million price tag for our 25 year old patients thus far and perhaps another 12 USD million for each in the future. These unsustainable expenses are due to the non-market-based nature of the American medical system and excessive drug development costs. Pre-Covid19 we paid 23% of our gross domestic product (GDP) for health care, were rated 37th for quality by the World Health Organization, and die 4 years younger than in a country whose government pays 1% of GDP, is rated 6th, and guarantees health care for all its citizens [18]. With the same system, we would have an annual 1.5 USD trillion surplus instead of the opposite which could sustainably be invested into drug development. While there is no question that small studies are necessary to demonstrate drug safety. The enormous expense of RDBPCTs to demonstrate small statistical benefits and noninferiority rather than obvious improvements or cures, excessively increases health-care costs in a country where those costs are already unsustainable. When pilot studies do not suggest benefits on the current status of each patient rather than statistically slightly slowing relentless deterioration, RDBPCTs should not be supported to the tune of $ billions. Sometimes an exceptional grasp of what it takes to survive is better than ‘cook book’ algorithms. If no RDBPCTs were necessary for Penicillin and Streptomycin, why do we need them when medications clearly convert infants with SMA type 1 into type 2 or 3?

Expert Opinion

For rare diseases for which it is difficult, time consuming, and expensive to recruit subjects for placebo controlled studies and for which pilot studies suggest that outcomes depend on the earliest possible initiation of treatment, the use of placebos can delay treatment and have a very detrimental effect. For example, when the natural history is certain, as it is for patients with spinal muscular atrophy type 1 with 2 copies of SMN 2, controlled studies are unnecessary when the treatment results in type 1 patients becoming type 2 or 3. In fact, placebo controlled studies are really only important to demonstrate statistical bioequivalence between competing medications and barely perceptible ameliorations in disease course and not important reversals of pathology or cures. As a result, their enormous cost greatly adds to the expense of medications and the unsustainable expense of our health care systems.

Declaration of interest

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

Reviewer disclosures

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Additional information

Funding

This paper was not funded.