Comparison between tafamidis and liver transplantation as first-line therapy for hereditary transthyretin amyloidosis

Abstract

Background

By stabilizing transthyretin, tafamidis delays progression of amyloidosis due to transthyretin variant (ATTRv) and replaced liver transplantation (LT) as the first-line therapy. No study compared these two therapeutic strategies.

Methods

In a monocentric retrospective cohort analysis, patients with ATTRv amyloidosis treated with either tafamidis or LT were compared using a propensity score and a competing risk analysis for three endpoints: all-cause mortality, cardiac worsening (heart failure or cardiovascular death) and neurological worsening (worsening in PolyNeuropathy Disability score).

Results

345 patients treated with tafamidis (n = 129) or LT (n = 216) were analyzed, and 144 patients were matched (72 patients in each group, median age 54 years, 60% carrying the V30M mutation, 81% of stage I, 69% with cardiac involvement, median follow-up: 68 months). Patients treated with tafamidis had longer survival than LT patients (HR: 0.35; p = .032). Conversely, they also presented a 3.0-fold higher risk of cardiac worsening and a 7.1-fold higher risk of neurological worsening (p = .0071 and p < .0001 respectively).

Conclusions

ATTRv amyloidosis patients treated with tafamidis would present a better survival but also a faster deterioration of their cardiac and neurological statuses as compared with LT. Further studies are needed to clarify the therapeutic strategy in ATTRv amyloidosis.

Keywords:

• tafamidis
• transthyretin
• cardiac amyloidosis
• liver transplantation
• polyneuropathy

Introduction

Amyloidosis due to variant (or hereditary) transthyretin (ATTRv) is an autosomal dominant disease caused by a point mutation of the transthyretin (TTR) gene. The circulating mutated TTR produced by the liver misfolds and promotes the deposition of amyloid fibrils in target organs (mainly heart and peripheral nervous system) [1–3]. By removing the main source of the circulating mutated TTR, orthotopic liver transplantation (LT) was the first treatment that improved the course of the neuropathy and survival; it has been the reference treatment from the 90’s to 2011 [4,5]. From then, new anti-amyloid treatments have been developed to stabilize the TTR tetramer or to silence the TTR gene [6–9]. Among these compounds, tafamidis, a TTR tetramer stabilizer, was the first to receive market authorization from the European Medicines Agency (EMA) in 2011 for the treatment of adult patients with symptomatic polyneuropathy of stage 1 due to ATTRv amyloidosis [10,11]. In 2018, tafamidis proved its efficacy against placebo for patients with TTR cardiac amyloidosis in the ATTR-ACT trial and was authorized in the EU for the treatment of transthyretin cardiac amyloidosis [12]. Noteworthily, the pivotal study that led to the EMA approval and the subsequent publications evaluating tafamidis was performed versus placebo [13–15]. In an analysis on unmatched populations, Coelho et al suggested that patients with stage I V30M transthyretin familial amyloid polyneuropathy treated with tafamidis may survive longer than liver-transplanted patients [16]. Still, anti-amyloid therapies (tafamidis, LT or silencing therapies) have not been compared either prospectively by controlled studies, or retrospectively by using matching techniques that would attempt to reduce bias.

In this study, we aimed to compare tafamidis with LT as first-line therapy in ATTRv amyloidosis patients on three clinical endpoints (all-cause mortality, cardiac, and neurological deterioration).

Materials & Methods

Study design

In a single-center retrospective study, ATTRv amyloidosis patients receiving a first-line therapy with tafamidis or LT were matched on the basis of their pre-therapeutic evaluation and three clinical outcomes were compared: overall mortality, neurological and cardiac worsening. Data were collected according to the French laws and the database was declared to the competent French authority (Commission Nationale de l'Informatique et des Libertés n°1470960). This study was initiated by the investigators of the French Referral Center for Familial Amyloid Polyneuropathy and Other Rare Peripheral Neuropathies (NNERF).

Patient Selection

Patients were identified from the registry of the French National Reference Center for FAP between January 1993 and May 2018. To be included, patients fulfilled the following criteria: have biopsy-proven ATTRv amyloidosis with a positive finding of an amyloidogenic TTR point mutation [17] and be treated with either tafamidis or LT as first-line therapy. Patients who received a double transplant (liver plus kidney or liver plus heart) were excluded from the analysis (Figure 1). To take into account a possible learning curve in the results of LT for ATTRv amyloidosis patients, we decided to exclude from the analysis the patients transplanted in the first year following the initiation of LT in our center (11 patients in 1993, ESM Figure 1). Furthermore, we checked that the year of LT was not associated with a poorer outcome neither in univariable nor in multivariable analysis.

Figure 1. Study flow chart. PS: propensity score.

Pre-therapeutic evaluation

The cardiac evaluation included a physical examination and assessment based on the New York Heart Association (NYHA) functional class, a surface electrocardiogram (ECG), and echocardiography. Left ventricular filling pressures were evaluated using echocardiography or right heart catheterization if indicated and natriuretic peptide levels according to the European Society of Cardiology 2016 heart failure guidelines [18]. Cardiac dysautonomia was diagnosed if the heart response rate after the infusion of atropine was ≤11 bpm or if the heart/mediastinum ratio 4 hours after MIBG injection was ≤1.43 as previously described (see ESM for further details) [19]. Cardiac involvement was defined as previously described in guidelines or in the ATTR-ACT trial for patients who did not have DPD scintigraphy [11,12].

The neurological evaluation included the polyneuropathy disability score (PND), a recognized clinical staging score for locomotion in ATTRv amyloidosis [10,20]. Briefly, stage 0 means no impairment, stage I corresponds to sensory disturbances but preserved walking capability, stage II indicates impaired walking capability but ability to walk without a stick or crutches, and stage III coincides with only walking with the help of one (IIIA) or two (IIIB) stick(s) or crutch(es) and stage IV represents confinement to a wheelchair or to bed. The autonomic nervous system was evaluated using a dedicated and validated clinical score (Compound Autonomic Dysfunction Test) [21]. The score integrated evaluation of postural hypotension, nausea/vomiting, diarrhea/constipation, and sphincter disorders. Each component of the score ranged from 0 (severe dysautonomia) to 4 (normal autonomic nervous system).

Treatments and follow up

All patients were ambulatory and free from standard contraindications to LT [22]. For patients who received a LT as their first-line therapy, physical examinations after surgery were gradually staggered up to one visit every 3 months in the LT center. Tafamidis became available from the end of 2010 in France and was then used as first line treatment in our center (see ESM Figure 1 for the year of treatment initiation). Patients received 20mg daily. All patients visited the reference center for neurological and cardiac evaluation at least every 6 months. Fourteen patients were lost to follow-up (3 in the tafamidis group and 11 in the LT group). Follow-up database was closed on January first 2021.

Outcomes

Three clinical endpoints were studied using time-to-event analyses: overall mortality, cardiac, and neurological status worsening. Regarding the all-cause mortality endpoint, causes of deaths were sought by all possible means and reviewed by at least two physicians (PS, VA, MS, DA). The cardiac endpoint was reached if one of the following events was documented: clinical worsening of heart failure that required initiation or a substantial increase of diuretics (≥40 mg furosemide); unscheduled overnight hospitalization for the cardiovascular event; death from cardiovascular causes. The neurological endpoint was defined by a worsening of the PND score by a neurologist (CC, DA). All events were retrospectively reviewed and validated by an expert panel as previously described [23]. Patients who did not reach endpoints at the time of their last visit were censored. For patients who received a second-line anti-amyloid therapy during follow-up, the reason for changing was collected, as well as the next therapy; the endpoints remained analyzed according to this first-line therapy. Tafamidis was not given in preparation for the transplant.

Statistical analysis

Descriptive reporting: Continuous variables are presented as medians (25^th and 75^th interquartiles) and categorical variables are presented as counts and percentages. Continuous variables were compared using Mann-Whitney tests and categorical variables by Fisher tests. Unadjusted outcomes were represented with Kaplan-Meier curves and compared using log-rank test. Cox single-variable regression was used to identify the preoperative variables associated with the endpoints. Multiple variable Cox regression models were constructed including variables with a p value <.1 on the univariate analysis and using a stepwise backward elimination of the least significant variable.

Propensity score and establishment of matched cohorts: Because patients were not assigned randomly between tafamidis or LT, propensity score was used to account for confounding variables. The propensity score included 23 variables collected during the pre-therapeutic evaluation: age, gender, TTR mutation type (V30M or other mutation), familial history of ATTRv amyloidosis, length of the disease before treatment initiation, body mass index, glomerular filtration rate, hemoglobin concentration, PND score, dysautonomia score, and its components (orthostatic hypotension, gastrointestinal dysautonomia, diarrhea, dysuria), NYHA class (I or above), interventricular septum thickness, posterior wall thickness, end-diastolic left ventricular diameter, left atrial dilatation, left ventricular ejection fraction, left ventricular filling pressure, implanted cardiac pacemaker, and cardiac dysautonomia. After calculating, for each patient, the likelihood of receiving a given treatment, patients treated with tafamidis and LT were paired by making the standardized difference of the propensity score within each pair smaller than 20% (Figure 2 and Table 1) [26–26]. We used robust variance estimation to compare the outcomes on the matched populations [27–29]. Patients who could not be paired using the propensity score were excluded from the matched cohort analyses. While analyzing the cardiac and neurological endpoints, the competing risk of death was taken into account by using the Fine and Gray method and the cumulative incidence functions are provided rather than the event-free patient proportions [31]. Analyses were performed using SAS software 9.4.

Figure 2. Absolute standardized differences between the two study populations (patients treated with tafamidis and with LT) before and after the propensity score matching procedure. PND: polyneuropathy disability score.

Results

Baseline characteristics and unadjusted outcomes

The files of 345 consecutive patients with ATTRv amyloidosis were analyzed, of whom 216 received LT and 129 were treated initially with tafamidis (Figure 1). Baseline characteristics of this cohort are detailed on Table 1. Median age was 53 years (IQR: 38; 64 years), 222 (64%) were males, 223 (64%) carried the V30M TTR mutation and the median follow-up was 68 months (IQR: 37; 105 months). As compared with LT, patients treated with tafamidis were older with a lower proportion of the early V30M mutation; their neurological statuses measured with the PND score were more severe. Similarly, the cardiac evaluation indicated that patients under tafamidis had thicker interventricular septum, lower ejection fraction and higher prevalence’s of dilated left atrium, increased left ventricular filling pressure, cardiac dysautonomia, symptomatic heart failure, and finally cardiac amyloidosis.

In the unadjusted cohort, 110 (32%) patients died during follow-up, 96 (44%) in the LT group and 14 (11%) in the tafamidis group. Survival was better for patients treated with tafamidis than for those with LT (p = .0012, Figure 3(A)). Cardiac endpoint was reached by 102 (30%) patients, 60 (28%) in the LT group and 42 (33%) in the tafamidis group. Neurological endpoint was reached by 122 (35%) patients, 47 (22%) in the LT group and 75 (58%) in the tafamidis group. Patients treated with tafamidis had higher risks of cardiac and neurological worsening (p < .0001 for both outcomes, Figure 3(B,C), respectively). As stated previously, patients treated with LT or tafamidis differed significantly for numerous characteristics and a matching procedure seemed mandatory to compare them fairly.

Figure 3. Unadjusted cohort. Comparative Kaplan Meier curves of patients treated with tafamidis (black line) or with LT (gray line) for the three endpoints of the study: overall survival (panel A), neurological endpoint (panel B) and cardiac endpoint (panel C).

Propensity score matching and outcomes for the matched cohorts

A propensity score was computed for 293 patients and 144 patients (72 pairs) were finally eligible for comparison (see Figure 1 for the detailed flow chart, Figure 2 and ESM Table 1 for the pre/post propensity detailed characteristics of the patients). For this propensity-matched population, median age was 54 years (IQR: 40; 62 years), 101 (70%) were males and 86 (60%) carried the V30M TTR mutation, and 69% presented cardiac ATTR involvement.

Table 1. Baseline characteristics of the patients.

	Overall population	Liver Transplantation	Tafamidis	p	Propensity matched population
Number of patients	345	216	129		144
Follow-up, years	5.6 (3.1; 8.7)	7.4 (3.6; 11.2)	4.1 (3.0; 5.05.8)	<.0001	6.0 (3.8; 8.3)
General characteristics
Sex: male	222 (64%)	132 (61%)	90 (70%)	.10	101 (70%)
Age at treatment initiation, years	52.6 (37.5; 64.0)	42.7 (34.6; 57.8)	64.4 (51.2;71.4)	<.0001	54.0 (39.5; 62.4)
Age of onset ≥50 years old	161 (47%)	68 (31%)	93 (72%)	<.0001	73 (51%)
V30M mutation	223 (65%)	149 (69%)	74 (57%)	.029	86 (60%)
V30M ATTR with early onset form	153 (44%)	125(58%)	28 (22%)	<.0001	55 (38%)
Familial history of ATTR	248 (72%)	175 (81%)	73 (57%)	<.0001	101 (70%)
Time between first symptoms and first treatment, years	2.9 (1.8; 4.2)	3.0 (2.0; 4.5)	2.3 (1.2; 3.6)	<.0001	2.4 (1.5; 3.9)
Body mass index, kg/m²	22.7 (19.8; 25.9)	21.6 (19.5; 24.7)	23.7 (20.6; 27.2)	.0006	23.9 (20.8; 26.8)
Hemoglobin, g/dl	13.5 (12.6; 14.5)	13.5 (12.5; 14.5)	13.8 (12.7; 14.5)	.24	14.0 (12.7; 14.7)
Glomerular filtration rate, ml/min	84.2 (71.0; 96.5)	85.3 (72.1; 96.5)	83.0 (69.9; 96.3)	.41	84.1 (71.1; 96.0)
Neurological status
Polyneuropathy disability score				.041
I	194	132	62		89
II	84	47	37		38
IIIa	64	34	30		17
IIIb	3	3	0		0
IV	0	0	0		0
Polyneuropathy disability score > I	151 (44%)	84 (39%)	67 (52%)	<.017	55 (38%)
Orthostatic hypotension	129 (38%)	88 (41%)	41 (33%)	.14	56 (38%)
Compound Autonomic Dysfunction Test^a	13 (11; 15)	13 (11; 15)	13 (11; 16)	.19	13 (11; 15)
Cardiac status
NYHA class > I	37 (11%)	13 (6%)	24 (18%)	<.01	14 (10%)
Conduction disorders on surface ECG	123 (35%)	89 (41%)	34 (26%)	.011	41 (28%)
Interventricular septum thickness, mm	11 (9; 14)	10 (8; 13)	12 (10; 16)	<.0001	10 (9; 13)
Dilatation of left atrium	71 (21%)	25 (12%)	46 (36%)	<.0001	32 (22%)
Left ventricular Ejection fraction, %	65 (60;70)	65(60;70)	64(59;68)	.0258	65 (60; 70)
Increased LV filling pressures	63 (18%)	25 (12%)	38 (29%)	<.0001	144 (12%)
Cardiac dysautonomia^c	149 (43%)	103 (57%)	46 (38%)	.0012	63 (44%)
Pacemaker implanted before treatment	127 (37%)	104 (48%)	23 (18%)	<.0001	27 (19%)
Cardiac amyloidosis	158 (46%)	80 (37%)	78 (60%)	<.0001	59 (69%)

Results are expressed as median (IQR) or number of patients (%). ^a: normal value = 16.  ^c: defined by an abnormal response to atropine infusion or a low MIBG cardiac uptake (cf. Methods).

Of the 144 patients included in this analysis, the median follow-up was 71 months (IQR: 45; 100 months). As expected, patients with LT had a longer follow-up than patients under tafamidis and time-to-event analyses were conducted to compare the two groups (respectively, 95 IQR: 63; 130 months; 56, IQR: 41; 75 months; p < 0.0001).

Thirty-eight patients (26%) died during follow-up, 32 (44%) in the LT population, and 6 (8%) in the tafamidis population. As compared with LT, mortality was significantly lower for patients treated with tafamidis (HR: 0.35; 95%CI: 0.14–0.92; p = .032, Figure 4, Panel A). The divergence between the two curves arose mainly from the early mortality seen in the first year after transplantation.

Figure 4. Matched cohort. Comparisons between tafamidis (black line) and LT (red line) for survival (panel A), neurological endpoint (panel B) and cardiac endpoint (panel C). For survival, Kaplan Meier curves were used. For neurological and cardiac endpoints, the competing risk of death is taken into account and the cumulative incidence function is presented.

The combined cardiac endpoint was reached in 42 (29%) patients, of whom 20 (28%) were treated with LT, and 22 (31%) with tafamidis. As compared with LT, patients treated with tafamidis had a 3-fold higher risk of cardiac worsening (HR: 3.00; 95%CI: 1.35–6.68; p = .0071; Figure 4, Panel B).

The neurological endpoint (worsening of the PND score) was reached in 64 (44%) patients: 19 (26%) with LT, and 45 (63%) of those treated with tafamidis. As compared with LT, patients treated with tafamidis had a 7-fold higher risk of worsening their PND score (HR: 7.20; 95%CI: 3.14–16.50; p<.0001; Figure 4, Panel C).

Subgroup analyses

Exploratory analyses of the three study outcomes were conducted on subgroups of the matched cohort (Figure 5). For survival, tafamidis did better than LT, especially for males, patients with late disease onset (≥50 years old), patients with initial PND score equal to I and patients with cardiac dysautonomia (Figure 5(A)). No subgroup had a better survival with LT than with tafamidis. Regarding cardiac endpoint, LT did significantly better than tafamidis for women, patients with early disease onset (<50 years old), patients with interventricular septum less than 12mm, and patients without cardiac dysautonomia (Figure 5(B)). No subgroup had a better cardiac outcome with tafamidis.

Figure 5. Subgroup analyses in the matched cohort of the three endpoints of the study: overall survival (panel A), neurological endpoint (panel B) and cardiac endpoint (panel C). *: in situations where the occurrence of events in the two compared populations did not overlap, the hazard ratios were not reliably calculable and only the P values of the statistical tests were presented.

For the neurological endpoint, LT did better than tafamidis in all subgroups except for patients with other TTR mutations than V30M (Figure 5(C)). In patients with early-onset V30M mutation, LT did better than tafamidis and survival did not differ significantly (ESM fig2).

Predictors of amyloidosis progression and second-line therapies

The variables associated with the neurological and cardiac endpoints were determined from the overall population (n = 345). The Cox proportional hazards univariate analysis is shown in the electronic supplementary Tables 1 & 2. The corresponding multivariable analyses are presented on Table 2 for both treatment (tafamidis/LT) and endpoints (survival, neurological and cardiac endpoints). Cardiac infiltration assessed by the interventricular septum thickness was associated with worse survival and poorer neurological and cardiac outcome regardless of the anti-amyloid therapy. For patients treated with tafamidis, a low hemoglobin level measured at treatment initiation was associated with poor survival and cardiac outcome. For transplanted patients, older age at surgery was associated with worse survival and poorer neurological and cardiac outcomes.

Table 2. Predictor variables of the neurological and cardiac endpoints for the two study populations.

Treatment	Endpoint	Variable	HR	95%CI	p
Tafamidis	Neurological	PND score > I	1.96	1.02–1.21	.0064
		Interventricular septum thickness, mm	1.09	1.02–1.15	.0063
	Cardiac	Hemoglobin, g/dl	0.67	0.53–0.85	.0011
		Interventricular septum thickness, mm	1.17	1.09–1.26	<.0001
	Survival	Hemoglobin, g/dl	0.50	0.31–0.80	.0038
		Interventricular septum thickness, mm	1.21	1.06–1.38	.0040
Liver Transplantation	Neurological	Age at treatment initiation, years	1.06	1.03– 1.10	<.0001
		Orthostatic hypotension	2.23	1.24–4.03	.0077
		Interventricular septum thickness, mm	1.11	1.02–1.21	.0194
	Cardiac	Age at treatment initiation, years	1.03	1.00–1.06	.0363
		PND score > I	1.78	1.01–3.13	.0455
		Interventricular septum thickness, mm	1.15	1.05–1.25	.0014
		Increased LV filling pressures	2.43	1.16–5.07	.0183
	Survival	Age at treatment initiation, years	1.03	1.00–1.05	.0277
		Interventricular septum thickness, mm	1.13	1.06–1.21	.0004
		Cardiac dysautonomia^a	3.46	1.78–6.72	.0002

Results are expressed as hazard ratio (95%CI). ^a: defined by an abnormal response to atropine infusion or a low MIBG cardiac uptake (cf. Methods).

During follow-up, second-line anti-amyloid therapy was initiated in 57 patients, of whom 43 were in the tafamidis group and 14 in the LT group. For patients under tafamidis, second-line therapy was patisiran (n = 21), LT (n = 16), inotersen (n = 4) revusiran (n = 1), and one patient interrupted all anti amyloid therapy. For patients who received LT, second line therapy was tafamidis (n = 11), and patisiran (n = 3).

Discussion

Statement of principal findings

In this propensity-matched analysis, first-line therapy of ATTRv amyloidosis with tafamidis was associated with a better survival than with LT. Conversely, patients under tafamidis had a poorer stabilization of their cardiac and neurological statuses, which could result in initiating a second-line therapy. These results were observed in both unmatched and matched cohorts. They were also observed consistently across most of the patient’s subgroups. Main results are summarized in the graphical abstract (ESM Figure 3).

To the best of our knowledge, our study is the first to compare two anti-amyloid therapies in ATTRv amyloidosis (namely tafamidis and LT). Our results raise several points of discussion.

Comparison with other studies

The better survival observed in the tafamidis group was due to an excess mortality that arose likely from the complications classically reported early after surgery in LT [5] (graft complications, infectious episodes while under immunosuppressive treatment), whereas tafamidis was very well tolerated in our series. This difference was particularly pronounced in populations with known poor results of LT (≥50 years old, non-V30M mutation including Ser77Tyr, Ser77Phe, Ile107Val and Thr49Ala variants) but was nonsignificant amongst early-onset V30M patients (see the electronic supplementary Table 4 for the full list of TTR variants). This reinforces the need for a particularly thorough preoperative evaluation before LT [31]. It also indicates that for selected patients with low-risk profile (patients with V30M mutation early onset), LT may still represent a good therapeutic alternative if other anti-amyloid treatments are ineffective.

Cardiac and neurologic statuses of patients under tafamidis deteriorated faster than in liver-transplanted patients. While they may seem disappointing, these results are in line with those reported in the pivotal studies: for polyneuropathy, the responder rate under tafamidis, assessed with the NIS score, was 45.3% at 18 months [6]. For cardiac amyloidosis, the ATTR-ACT trial which included patients with cardiac amyloidosis and heart failure reported that 52.3% patients under tafamidis required cardiovascular-related hospitalization at 30 months but patients in our study had a much less severe profile [12]. Similar progression rates have recently been reported in real-life populations [32]. Progression of the disease in spite of LT has been explained by deposit of wild-type TTR after LT [3]. For tafamidis, one hypothesis that could explain incomplete TTR stabilization would be an underdosing of tafamidis at 20mg, which was the dose approved to treat polyneuropathy. Under this dose, the plasma concentration of tafamidis and the extent of TTR kinetic stabilization could vary by a factor of four [33]. Higher doses of tafamidis (up to 80mg) may be required to achieve adequate TTR stabilization [34]. In the ATTR-ACT trial, two tafamidis doses were tested (20 and 80mg) and done similarly on the study endpoints [12]. However, a recent analysis from the long-term extension of the study reported better survival under tafamidis 80mg than 20mg with similar tolerance profiles [35]. A second hypothesis to explain our results would be a lower efficacy of tafamidis in ATTRv amyloidosis (as compared with ATTRwt). The ATTR-ACT study included mainly patients with ATTRwt amyloidosis (76.1% in the tafamidis group vs. 0 in our study). Subgroup analyses that focused on patients with ATTRv amyloidosis were not powered to demonstrate benefits and further analyses are awaited in this specific population.

Unanswered questions and future research

Our study questions the future of non-responders to a first-line anti-amyloid therapy. Such non-responders seem to exist for all treatments (LT, tafamidis, inotersen and patisiran): in patients with polyneuropathy, positive response to treatment was documented for half of the patients under inotersen and for three quarters of the patients under patisiran [7,8]. Efficacy of these two therapies on cardiomyopathy due to ATTR amyloidosis is currently tested in phase III trials vs. placebo (NCT03997383, NCT03702829, NCT04153149) and currently, no direct comparison exists between these different treatments, which makes it all the more difficult to choose for patients with both cardiac and neurologic impairment. In favor of tafamidis, one could mention its ease to administrate (simple oral route), its safety, and the fact that it is to date the only treatment with positive results on survival demonstrated in patients with ATTR amyloid cardiomyopathy. Our study indicates however that close monitoring of these patients is mandatory as some may require alternative therapeutics.

Predictor variables associated with neurological or cardiac evolution were reported in our study. Interestingly, a low hemoglobin level was associated with poor outcomes for patients receiving tafamidis. It is known that due to a lack of erythropoietin production, anemia could affect up to one fifth of the patients with ATTRv amyloidosis [36,37]. This anemia persists despite anti-amyloid treatment (tafamidis or LT) [38,39]. However, this is, to the best of our knowledge, the first time that hemoglobin level is reported to be associated with clinical outcomes in patients with ATTRv amyloidosis under tafamidis. Finally, analysis of the predictive variables also confirms that both tafamidis and LT should be considered as early as possible in the course of the disease to be efficient.

Strengths and limitations

The design of the study has inherent limitations: it was retrospective, involved a single medical center and comparison of the two strategies (tafamidis vs. LT) was not performed using a randomization. The two therapeutic strategies were applied sequentially (and not concurrently) as tafamidis became available starting 2011, which could be seen both as a limit and a strength. A limit because patients management could change during the course of the study; a strength because there was no period when the two treatments could be chosen equally in the first line, which could have induced an additional selection bias. Some of the patients treated with LT were transplanted in the early 90s and this could represent a learning curve bias against LT. To rule out this hypothesis, we excluded patients transplanted in 1993 and we confirmed that the year of LT was not associated with survival (ESM Figure 2). The availability of the second-line treatments fluctuated during the study and was also distributed asymmetrically between the two arms. This could in theory have favored tafamidis and minimized the observed differences vs. LT, while the latter seemed to better stabilize cardiac and neurological endpoints. Our study was performed in a European population with ATTRv amyloidosis, of whom 65% patients presented with V30M mutation; the course of the disease in other FAP patient populations may therefore be different, but we observed similar results among Met30 and non-Met30 patients. Numerous variables have been shown to impact on the effect of treatment on ATTRv amyloidosis and it is still unclear whether these variables are similar under tafamidis and after LT [14,31]. To minimize these potential biases, the propensity score was built to compensate not only for the differences between groups at baseline regarding their clinical status but also included the conditions under which both therapies were administered (specifically length of the disease before treatment initiation). Given the magnitude of the effects observed on the overall survival and on the neurological endpoint, it seems unlikely that they would result only from differences induced by those biases.

Finally, the deterioration of cardiac and neurological statuses should logically be followed by a decline in survival, and one may hypothesize that the follow-up is too short in the pharma group to make any strong conclusions on long term survival between the groups.

Conclusions

In this propensity-based comparison between tafamidis with LT in patients with ATTRv amyloidosis, first-line therapy with tafamidis would be associated with a better overall survival but with a poorer stabilization of cardiac and neurological statuses. Close multidisciplinary monitoring of patients carrying ATTRv amyloidosis and treated with disease-modifying therapies including tafamidis is mandatory to adapt therapy. This hypothesis-generating analysis pleads for future studies that will assess the comparative efficacy of the anti-amyloid therapies.

Authors contributions

VA, DA and MS led the study conceptualization, development of the research question.

AB, PS, and VA led the development of advanced statistical analyses. PS and PML collected the data. VA and PS wrote the first draft of the paper. DA declared the database to the French authorities. CQ, EP, IK, AR, FR, AEL, and DS contributed to the discussion on protocol development and provided critical feedback on drafts of the manuscript. All authors had full access to all the study data, and VA had final responsibility for submission and is the guarantor. The corresponding author (VA) attests that all listed authors meet authorship criteria and that no others meeting the criteria have been omitted.

Abbreviations
ECG	=	electrocardiogram
EMA	=	European Medicines Agency
HR	=	Hazard Ratio
LT	=	liver transplantation
ATTRv	=	hereditary transthyretin amyloidosis
ATTRwt	=	wild type (non-mutated) transthyretin amyloidosis
MIBG H/M ratio	=	123-meta-iodobenzylguanidine heart/mediastinum ratio
NYHA	=	New York Heart Association
PND	=	polyneuropathy disability score
TTR	=	transthyretin

Acknowledgements

We thank the patients who were followed and treated in the NNERF since the early 90s. We thank the data managers that were involved in the study in each hospital of the NNERF, especially Sophie-Nathalie Marchal (Bicêtre), Vincent Karam (Paul Brousse), Pierre Marc Lallemand, and Cathy Ladeveze (Antoine Béclère), and Fatima Ouderg (Bichat). We thank the medical-surgical team of the Hepato-Biliary Centre of the Paul Brousse Hospital. Finally, we thank Eric Duong, who reviewed our manuscript for style and language. Eric Duong was financially compensated for his contribution.

Disclosure statement

The author(s) received no specific funding for this work.

Out of this work, VA and MS declare having links of interest with Pfizer (consulting fees, research scholardship); VA, DA and MS declare having links of interest with Pfizer and with Alnylam (for both, consulting fees, and research scholardship).

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.