Exa-cel on review at the FDA

Just last week, on 31 October, the FDA’s Cellular, Tissue, and Gene Therapies Advisory Committee met to discuss Vertex Pharma’s Biologics License Application for Exagamglogene autotemcel (or exa-cel, and formerly known as CTX-001) — a cell-based gene therapy designed to treat sickle-cell diseases.

I have written about exa-cel many times before, both on this blog (here) and in published academic writing too. I have also spoken about it in this podcast. Exa-cel is a therapeutic product that is composed of the patient’s own (autologous) hematopoietic stem cells; however, those cells — specifically differentiation 34+ (CD34+) cells, have been edited using CRISPR/Cas 9 editing machinery (CRISPR).

In short, a CRISPR-Cas endonuclease system (CRISPRs) is a naturally occurring adaptive immune system that exists in most bacteria. These systems prevent bacteria from being infected by foreign genetic elements, such as viruses and phages. Where an infection exists, the Cas9 protein in the CRISPR will cleave or cut one strand each of double-stranded DNA to cause a double-strand break and thus decrease production of progeny viruses. Following that DSB, the genome will be repaired naturally, usually through a naturally occurring process called non-homologous end joining or NHEJ.

But this CRISPR/Cas 9 system can be ‘hijacked’ by science for therapeutic purposes. Using a guided template, the DSBs made by the Cas9 protein can be repaired in a precise and controllable manner, allowing the editing machinery of cell repair to be redirected toward doing repairs or edits that it would otherwise not do unguided.

This is how exa-cel works (in a nutshell). After the patient’s cells have been extracted, they are subject to guided disruption and repair by CRISPR. The CRISPR system make precise DSBs at the erythroid lineage specific enhancer region of B-cell lymphoma/leukemia 11A (BCL11A) gene on chromosome 2. In turn, this process disrupts GATA1 binding and abrogates BCL11A expression. Having turned off the expression of BCL11A, another gene, γ-globin (HBG1/HGB2) is expressed, creating fetal hemoglobin (HbF) production. It is quite complicated; but, in essence, the production of fetal hemoglobin allows people with sickle-cell diseases (red blood cells shaped like sickles because the cells are starved of oxygen) to be restored to health.

There are many issues and risks with this therapy, including the fact that sickle-cell disorder could be a protective disorder against malaria. But one especially concerning prospect, which is really at the core of the BLA on review currently, is the chance that the CRISPR may create cuts or DSBs at a site on the genome locus that is not in the right place. These misplaced or unforeseen cuts are known as ‘off-target effects’ or, alternatively, as indels — which means (usually unintended) insertions or deletions. The BLA puts the risks well:

One of the main concerns related to genome editing technology is risk of cleavage of genomic DNA at unintended sites due to imperfect pairing between the gRNA and the target DNA sequence. A subset of these imperfectly paired sites can be cleaved by the Cas9 endonuclease resulting in unintended edits across the genome. These sites can tolerate up to 6-mismatches between the gRNA and the genomic DNA. Since unintended edits can disrupt gene expression if present in the coding or regulatory DNA sequences, it is critical that the specificity of the gRNA be thoroughly screened to ensure off-target genome editing is minimized.

https://www.fda.gov/media/173414/download

I am still trying to get my head around the recent report of the results of the off-target analysis present in the BLA. The FDA’s BLA report states as follows:

For the cellular off-target analysis, the Applicant used three samples from healthy donors and three samples from subjects with SCD of African American ethnicity. Given the impact of the SCD on [hematopoietic stem cell] function, which can potentially change the chromatin landscape and can impact off-target editing, the merits of using healthy donor samples for such analysis is not clear.

Additionally, it is not clear if the small number of samples used in the cellular GUIDE-seq offtarget analysis is sufficient to adequately assess off-target editing in exa-cel.

https://www.fda.gov/media/173414/download

The report then continues:

4.1.1.1 In Silico Analysis Off-Target Analysis Data for Exa-cel

The Applicant used three publicly available in silico algorithms to nominate potential off-target sites for the sgRNA SPY101 (Figure 6) based on its homology to the reference sequence.

https://www.fda.gov/media/173414/download

Notably, however, when you get to the next page on the analysis of these risks, there are a number of redactions, no doubt because these are commercially protected contents that the regulator must not disclose. On first view, it appears that these ‘in silico algorithms’ to nominate potential off-target sites is, as is said below, a ‘part of the tool.’ I am not quite sure whether that means that the tool — the CRISPR system used be Vertex — is also the same system that conducts the off-target search, or something else. Have a read:

In any case, it looks like the so-called ‘indel frequency’ is very low. As the report noted later, “In this analysis, there were no statistically significant off-target editing events observed at any of the off-targets nominated using in silico analysis.” Although it remains unclear to me precisely how the indel assessment takes place, it is worth noting that the report’s view of the findings of the sponsor are very ambiguous, and tend towards a finding that the results of the study are inadequate. As the report notes in its conclusion of the safety summary section (4.1.2):

These changes have the potential to impact the chromatin landscape of SCD donor derived CD34+ HSPCs. Since chromatin accessibility can influence off-target activity, it is not clear if GUIDE-seq analysis of healthy donor derived CD34+ HSPCs can adequately capture potential off-target editing occurring in patient cells. However, availability of SCD donor cells can be limited and should also be considered. The Applicant used a total of four samples that were from donors of African American ethnicity. Three of these samples were from SCD donors that were used in the GUIDE-seq experiment and hybrid capture sequencing experiment, and one sample was from healthy donor that was used in the hybrid capture sequencing experiment. Given the limited number of SCD samples that were used in the cellular off-target analysis, it is not clear if the GUIDE-seq analysis adequately assessed the potential off-target editing by exa-cel.

Given the ambiguity of the FDA’s assessment, which states that Vertex’s pharmacovigilance plan is still under review, it remains to be seen whether more studies will need to be provided before the FDA consider exa-cel ready for the clinic.

Updates for September: Cosmetics and COVID-19

This month has been a busy month for my semi-scholarly work. At the start of the month, I wrote a piece for the Conversation about the so-called crackdown on the cosmetics procedures sector announced by AHPRA around 7 September 2023.

Following the publication of that piece, I was lucky to be asked to appear on a long interview for Mamamia’s The Quicky. It was a great talk and I thank host Claire Murphy for the great interview.

Over the last few days, when it was announced that a COVID-19 inquiry would be run by the Commonwealth government, its terms of reference came under scrutiny. Having co-authored and published an article earlier in the year on COVID-19 mandates — an article that I was very proud to have written, especially when it was given plaudits from now Justice of the High Court of Australia Hon Robert Beech-Jones — I was keen to provide my view on the terms of reference.

Several commentators around the country have argued that the terms of reference are unduly restrictive and a cause for complaint. That, in some ways, is understandable. After all, it is essential that critical minds and governments examine how the public health legislation was deployed during the pandemic, especially where profound human rights issues arise from coercive interventions.

However, as I argued ion ABC News Radio on Friday 22 September, the Inquiry expressly will cover questions of governance, including

the role of the Commonwealth Government, responsibilities of state and territory governments, national governance mechanisms (such as National Cabinet, the National Coordination Mechanism and the Australian Health Protection Principal Committee) and advisory bodies supporting responses to COVID-19.

See Inquiry TORs

Given that our Constitution does not confer on the Commonwealth a competency to make laws with respect to public health, however, the terms of reference strike me as sensible and practical.

After all, it is the states, including New South Wales, that hold the power to regulate public health within their polities. Sections of the states’ public health Acts confer different conditional powers on their relevant health ministers; and, in New South Wales, it is section 7 that gives the Minister massive powers to do whatever they consider to be necessary when, on reasonable grounds, they consider that a situation has arisen that is likely to be a risk to public health, and they seek to deal with the risk and its possible consequences. My phrasing, above, is a little garbled, because it seeks to paraphrase the wording of the provision, which is as follows:

7 Power to deal with public health risks generally
(cf 1991 Act, s 5)

(1) This section applies if the Minister considers on reasonable grounds that a situation has arisen that is, or is likely to be, a risk to public health.

(2) In those circumstances, the Minister–

(a) may take such action, and
(b) may by order give such directions,

as the Minister considers necessary to deal with the risk and its possible consequences.

All of the NSW public health orders were made pursuant to section 7, above.

The Victorian Act, by contrast, requires that a ‘pandemic declaration’ be made before the minister’s powers are enlivened. Following the making of any such declaration, the Minister of Health may exercise their powers to make orders under s 165AI of the Public Health and Wellbeing Act 2008 (Vic), which states as follows:

PUBLIC HEALTH AND WELLBEING ACT 2008 – SECT 165AI

Minister may make a pandemic order

(1) The Minister may, at any time on or after the making of a pandemic declaration, make any order (a pandemic order) that the Minister believes is reasonably necessary to protect public health.

As is obvious, these powers are held by state ministers. They are not controlled or controllable by the Commonwealth government, or any Commonwealth minister. Arguably, any influence exerted by a Commonwealth minister over a state minister to exercise their powers in a particular way may give rise to a misuse of the power, inasmuch as the requisite ‘consideration’ (New South Wales) or ‘belief’ (Victoria) of the health minister has to be held by that relevant state minister. Sure, a health minister’s belief or consideration may be reconfigured by additional information from other actors, including their equivalent ministers in other Australian governments. But the legal fact is that the Commonwealth cannot really repeal, remake or otherwise alter or even influence this power; and this is a point that I have found has been missing from the public debate about the scope of the COVID-19 inquiry.

What would be the best outcome of an inquiry into the way the state health ministers exercised their powers? Would the Commonwealth recommend that the state ministers must have approval from the Commonwealth before exercising these powers? Would they be required to consult? This would seem to be completely contrary to the design of the Constitution, which, as Quick and Garran point out time and again, was premised on a fair compromise between the new central government and the colonies: the states would retain independent powers across a range of areas, including, as I think it must be inferred, health. I am not opposed to more centralisation; however, I think it;’s important to understand what we are proposing when we ask for a Commonwealth inquiry into the exercise of state power. And so I was pleased to feed that into the debate via national radio.

My request, alas, remains with ABC Archives for a copy of the recording! (It cannot be replayed like some other programs on ABC listen, etc).

In the meantime, I feel more compelled than ever to revise and rewrite the serious consideration that I developed in 2018 on the division of health power in Australia — an article that, for various reasons, did not make it through peer review at the top journals but which I have been told several times since would make a sorely needed contribution to the literature in this area. The best articles on the topic I have found are two oldies and a newie:

Updates on welfare law

I have been writing on welfare law over on my substack blog Welfare Law in Australia. One recent post and video deals with the question of ‘unlawfulness’ as it arises in respect of the former section 1073B of the Social Security Act 1989 (Cth). The way that that section was applied by the welfare departments between 2003 and 2020 was, as the Ombudsman has confirmed, unlawful. But what does unlawful mean here?

As I suggest, it essentially means that the departments acted in a way that was not expressly authorised by the legislation. But, perhaps more faultily than this, the departments misinterpreted or misconstrued their powers under the section; and, accordingly, they acted without power on an assumption that they did have power. As I note (perhaps somewhat agonisingly), this is a little like playing football with ‘boots on your hands’ — essentially doing something that is within the rules (‘kicking the ball’) but completely misunderstanding what the rules mean (ie, the rules require players to kick the ball with their feet; not simply ‘with boots on one’s limbs’). For mine, the lesson to arise from this (the Ombudsman’s statement, after all, is titled ‘Lessons in Lawfulness’) is that reading legislation carefully is absolutely indispensable. The more sophisticated version of that lesson is that:

  • internal calculators or automation tools in government departments need to be crafted in accordance with legislation to the letter;
  • the calculator in this case, the Earnings Apportionment Tool (within the Adex Debt Statement) should never have been permitted to create apportionments that went beyond the statutory limit of one ‘instalment period’; and
  • cross-disciplinary legal practitioners must be identified and employed, or alternatively trained — those who have skills in statutory interpretation as well as computer and software design — in order to avoid internal tools being run and applied that perform operations that are contrary to law.

The issue that arose in relation to the legal construction of the provision was simple enough — at least in hindsight. The provision provided as follows:

1073B Daily attribution of employment income

(1) If:

(a) a person is receiving a social security pension or a social security benefit; and

(b) the person’s rate of payment of the pension or benefit is worked out with regard to the income test module of a rate calculator in this Chapter; and

(d) the person earns, derives or receives, or is taken, either by virtue of the operation of section 1073A or any other provision of this Act, to earn, derive or receive, employment income during the whole or a part of a particular instalment period of the person;

the person is taken to earn, derive or receive, on each day in that instalment period, an amount of employment income worked out by dividing the total amount of the employment income referred to in paragraph (d) by the number of days in the period.

As I have noted in my analysis in this video, it was perfectly lawful for the agencies to apportion income, as the provision provides, by ‘dividing the total amount of the employment income… by the number of days in the [instalment] period.’ However, it was not lawful for the agencies to divide the ‘total amount of the employment income … by [a greater number] of days [than exist] in the [instalment] period,’ which is what they did. It’s all there in the last sentence of the provision.

The word ‘period’ as it is used in the last clause of the provision appears to have taken on an expanded meaning on the agencies’ construction so that it could apply to the entire payslip period; however, as is clear on a plain language reading of the provision, the term ‘period’ can only refer back to the word ‘instalment period’ that appears earlier in the provision. Otherwise it has no fixed meaning, which is clearly unbearable.

I was very disappointed by the Deputy Secretary of DSS, as well as by the Ombudsman, when they appeared recently in the Senate inquiry on poverty in Australia, because both these senior governmental figures bemoaned the ‘complexity’ of the matter, and stressed how difficult it was to explain. I was disappointed because it really need not be described in this way. It can be explained quite tersely and clearly. And if senators who are not legally trained have questions, they can be answered pretty well. What needs to be explained is that the agencies made a very rudimentary reading error. They identified the word ‘period’ in the provision as a word that has no fixed meaning but could potentially expand to any quantum. When one thinks about it, that was a very amateurish mistake. And one imagines, as in the case of robodebt, the mistake was bordering on one that used to be described, in the old actions on the case, as faulty for ‘want of due care.’

I will be aiming to write a more complete analysis later on why this is not an error that is wholly or even at all distinguishable from robodebt, as seems to be the message from all and sundry now, from the Ombudsman himself, through the department heads, through to the relevant ministers (Hons Shorten and Rishworth). But it is an untextured and, ultimately, untrue message. As the appendix to the robodebt Royal Commission clearly illustrates, more than twenty of the earliest robodebts that came before the AAT1 were subject to recalculation orders by AAT members who found them to be ‘unlawful’ precisely because they were calculated impermissibly under s 1073B.

This project to distinguish miscalculated debts under s 1073B from robodebts is something of a legal nonsense. Yes: robodebts involve debt letters that alleged debts based on ATO data. And, yes, the s 1073B problem predates the OIC and other version of the robodebt programs. But, in terms of the legal problems presented by robodebts, these debts were unlawful, in the end, for two reasons: first, they were insufficiently evidenced in terms of their quantum, as the Federal Court observed in its settlement orders. The Court (Davies J) put it this way (at paragraph 9):

there was no material before the decision-maker capable of supporting the conclusion that a debt had arisen pursuant to s 1223(1)(b) of the SS Act.

But the second and very much related reason why robodebts were unlawful is because they were calculated contrary to s 1073B. The fact that this second reason was not explored in the Royal Commission nor articulated in the reasons of Davies J in Amato, and had not been aired publicly before this month, does not make it any less true that robodebts were unlawful because of this problem. To be sure, the smoking gun in this respect is the fact that the Amato submissions made the s 1073B impermissibility argument (emphasis mine):

56.4 Section 1073B provided for averaging of income earned by a person receiving social security payments over an “instalment period”. An “instalment period” could not exceed 14 days. The effect of s 1073B was to produce an average daily rate of income for each day in the instalment period. Section 1073C then provided that “(a) the rate of the person’s employment income on a fortnightly basis for that day may be worked out by multiplying that amount by 14; and (b) the rate of the person’s employment income on a yearly basis for that day may be worked out by multiplying that amount by 364”. Sections 1073B and 1073C do not authorise apportionment of the kind done under the EIC program. They simply provide a mechanism for working out “the rate of the person’s employment income on a yearly basis” where it is relevant, under the SS Act, to calculate the rate on a yearly basis. The relevant period for Austudy payments was a fortnight, not a year. As the relevant explanatory memorandum states, the main purpose of ss 1073B and 1073C was to facilitate the “working credit” rules in the SS Act, the function of which “is basically to allow a person’s ordinary income to be reduced before it is put through [the] income testing process”. The working credit rules provided for the calculation of “the participant‘s rate of employment income on a yearly basis” for the purpose of working out the effect on working credit balances of social security pensioners. The working credit rules did not apply to Austudy recipients.

But, as I say, more on that later.

Another recent post deals with another issue that arose this month regarding historical errors on the part of the welfare agencies. Services Australia’s IT system committed errors that affected almost 50,000 child support payment recipients in an unspecified period prior to and around 2018. While the IT issue is said to have been resolved in 2020 (some two years after the agency had been consulted on the issue by the Ombudsman), only around 35,000 people had by that time had their assessments and entitlements re-assessed and remediated, while another roughly 15,000 recipients’ entitlements were not reassessed or remediated. Instead, they were simply ‘written off’ as inconsequential and unnecessary to address by the welfare delivery department.

This decision was made by the agency despite the fact that it had committed to reassessing all the affected files in 2019 in representations it had made to the Ombudsman. All of this, of course, was occurring during the robodebt settlement and through the Royal Commission. Clearly concerned that this backsliding on a prior commitment could result in unfair outcomes, the Ombudsman this year sought to re-engage with the department to ‘make things right’ — and that, as it happens, is the name of the statement published by the Commonwealth Ombusdman in respect of this matter.

This short statement is not to be confused with a report of the same name published by the Ombudsman with respect to the Department of Education back in 2015. (I suppose, when one is the Ombudsman, titles like Making things Right or Lessons in Lawfulness will probably need to be recycled every few years.) As I argue in the post, it is just another example of the recalcitrance of the agency among several that have emerged since the Robodebt Royal Commission.

If you have not been following my activity on Welfare Law in Australia, please do subscribe. Relatedly, my media appearances in recent times on welfare law are as follows: