Estimating Meaningful Differences in Measures of Neuropathic Impairment, Health‐Related Quality of Life, and Nutritional Status in Patients With Hereditary Transthyretin Amyloidosis

ABSTRACT

Introduction/Aims

The degree of change in neuropathic impairment and quality of life (QoL) that is clinically meaningful to patients with hereditary transthyretin amyloidosis with polyneuropathy (ATTRv‐PN) is not established. This study aimed to estimate the magnitude of treatment differences that are meaningful to patients in measures of neuropathy and QoL and to determine whether eplontersen achieved a meaningful improvement versus placebo.

Methods

Data from the NEURO‐TTRansform trial on patients with ATTRv‐PN treated with eplontersen (n = 141) or historical placebo (n = 59) were used. Anchor‐based approaches were used to estimate thresholds for meaningful differences in the modified Neuropathy Impairment Score +7 (mNIS+7) composite score, Norfolk QoL‐Diabetic Neuropathy (Norfolk QoL‐DN) total score, Neuropathy Symptoms and Change (NSC) total score, and modified body mass index (mBMI). Differences between the least squares means of the treatment groups were analyzed.

Results

Meaningful improvement in mNIS+7 was estimated as −4.0 points and deterioration as 1.8 points. The estimated ranges of meaningful improvement and deterioration in Norfolk QoL‐DN were −12.8 to −4.0 points, and 5.9 to 14.7 points, respectively. For NSC, ranges were −2.4 to −1.3 points for meaningful improvement, and 0.6 to 5.8 points for deterioration. The estimated meaningful improvement in mBMI was 9.8 kg/m² × g/L and deterioration was −40.9 kg/m² × g/L. Improvements in each measure with eplontersen versus placebo were greater than the estimates of meaningful differences.

Discussion

Eplontersen demonstrated a clinically meaningful effect on neuropathic impairment, QoL, and nutritional status. Such estimates have implications for clinical practice and trials.

Abbreviations

10MWT

10 Meter Walk Test

ASO

antisense oligonucleotide

ATTRv

hereditary transthyretin

ATTRv‐CM

ATTRv cardiomyopathy

ATTRv‐PN

ATTRv polyneuropathy

COMPASS‐31

Composite Autonomic Symptom Score‐31

eCDF

estimated cumulative distribution function

FAS

full analysis set

GH01

general health item

LSM

least squares means

mBMI

modified body mass index

MCID

minimal clinically important difference

mNIS+7

modified Neuropathy Impairment Score +7

NIS

Neuropathic Impairment Score

NIS‐LL

NIS lower limb

Norfolk QoL‐DN

Norfolk Quality of Life‐Diabetic Neuropathy

NSC

Neuropathy Symptoms and Change

PCS

Physical Component Summary

PGIC

Patient Global Impression of Change

PGIS

Patient Global Impression of Severity

PND

Polyneuropathy Disability

QoL

quality of life

SD

standard deviation

SF‐36v2

36‐Item Short Form Survey version 2

TTR

transthyretin

1. Introduction

Hereditary transthyretin (ATTRv) amyloidosis is a rare, progressive, autosomal‐dominant disease that, without treatment, is fatal. It is caused by alterations in the gene encoding the transthyretin (TTR) protein that lead to an aggregation of misfolded TTR proteins [1, 2]. The most typical clinical presentations of ATTRv amyloidosis are peripheral and autonomic neuropathies, known as ATTRv polyneuropathy (ATTRv‐PN) and ATTRv cardiomyopathy (ATTRv‐CM) [2].

Clinical trials for patients with ATTRv‐PN have demonstrated improvements in neuropathic impairment and quality of life (QoL) with TTR gene silencing treatments (small interfering RNAs and antisense oligonucleotides [ASOs]) [3, 4, 5, 6]. In such trials, neuropathic impairment and QoL are commonly assessed using the modified Neuropathy Impairment Score +7 (mNIS+7) composite score [7], Norfolk Quality of Life‐Diabetic Neuropathy (Norfolk QoL‐DN) total score [8], and Neuropathy Symptoms and Change (NSC) total score [3, 4, 5, 9, 10, 11]. Nutritional status in patients with ATTRv amyloidosis can be assessed using modified body mass index (mBMI) [12].

Whilst these are well‐established measures of disease progression or improvement in ATTRv‐PN, the changes in scores obtained from these measures that are considered meaningful to patients with ATTRv‐PN are not yet established. This change between timepoints can be assessed at the patient level, which defines whether a patient is a treatment “responder,” or at the group level, that is, the clinically meaningful difference in an outcome between a treatment group and a control group. This has sometimes been referred to as the “minimal clinically important difference” (MCID). The term MCID has been used in multiple ways—referring to both a between‐group difference and to a patient‐level change—to define responders. In our analysis, the term “meaningful difference” will be used to denote the magnitude of treatment difference between groups that is meaningful to patients. When a patient‐level threshold is used to define patients who have experienced a meaningful change, this will be referred to as a “responder threshold.”

This analysis used data from the NEURO‐TTRansform phase 3 trial of eplontersen [5] to estimate the magnitudes of treatment differences in mNIS+7 composite score, Norfolk QoL‐DN total score, NSC total score, and mBMI that are meaningful to patients. The estimates obtained were also used to determine whether the treatment effects seen with eplontersen constitute a meaningful difference to patients.

2. Methods

2.1. Trial Design

The phase 3 NEURO‐TTRansform trial (NCT04136184) evaluated the efficacy and safety of open‐label treatment with subcutaneous eplontersen (45 mg every 4 weeks) until the end of the treatment period at week 85 (4 weeks after the last administration of the study drug on week 81) [5]. Results were compared with a historical placebo arm that was derived from the NEURO‐TTR clinical trial (NCT01737398) [4] at week 66, which marked the end of the historical placebo‐controlled efficacy analysis [4].

The detailed methods and results of the NEURO‐TTRansform trial were previously published [5, 13, 14]. The inclusion criteria and endpoints were designed to be comparable with those used in the NEURO‐TTR trial [4, 13].

NEURO‐TTRansform was conducted in accordance with the International Council for Harmonisation guidelines and the trial protocol was approved for each participating center by the relevant local institutional review boards or ethics committees. Written informed consent was provided by all patients before enrollment.

2.2. Patients

Patients included in the NEURO‐TTRansform trial were adult patients with Coutinho Stage 1 (independent ambulation) or 2 (ambulatory with assistance) ATTRv‐PN and had a documented TTR variant, showed signs and symptoms consistent with polyneuropathy, and had a Neuropathic Impairment Score (NIS) ≥ 10 and ≤ 130. Key exclusion criteria have been previously reported [4].

2.3. Outcome Measures in NEURO‐TTRansform

This analysis aimed to establish meaningful differences for the following four outcome measures: mNIS+7, Norfolk QoL‐DN, NSC, and mBMI. Multiple variants of the mNIS+7 exist; the mNIS+7_Ionis [7] was used in this trial. mNIS+7, Norfolk QoL‐DN, and NSC were measured at baseline, and at weeks 35, 66, and 85 of the trial. mBMI (kg/m² × serum albumin [g/L]) was assessed at baseline and at weeks 13, 35, 65, and 85, but data from week 13 were not used in this analysis, as none of the anchor measures (see subsection below) were assessed at this timepoint.

Higher scores in the mNIS+7, Norfolk QoL‐DN, and NSC indicate greater impairment or deterioration [7, 9, 11, 15]. Conversely, lower scores in mBMI indicate deterioration because, for patients with ATTRv amyloidosis, a decrease in mBMI can be an indicator of deterioration with respect to nutritional status, reflected by weight loss or lower serum albumin levels [1].

2.3.1. Anchor Measures

Exploratory assessments of the Patient Global Impression of Change (PGIC) and Patient Global Impression of Severity (PGIS) were assessed at weeks 37 and 85, whilst the general health item (GH01) from the 36‐Item Short Form Survey version 2 (SF‐36v2), the NIS, and the Polyneuropathy Disability (PND) scale were conducted at weeks 35, 65 (66 for NIS), and 85. These measures (except for the PGIC) were also assessed at baseline.

For this analysis, scores from these measures were assessed for suitability as anchor measures for the estimation of meaningful difference. The criteria used for anchor measures were that they must assess similar concepts as the outcome measures, they must not be a derivative of the outcome measures, and changes in the anchor should statistically correlate with changes in the outcomes measures [16]. Table S1 provides details of the outcome measures and anchor measures, outlining which aspect of ATTRv‐PN they are assessing.

2.4. Statistical Analysis

All analyses reported here are post hoc and exploratory. The findings are based on the full analysis set (FAS), which included all participants who were randomized to treatment and received ≥ 1 dose of trial medication, and who completed baseline and ≥ 1 post‐baseline assessment in either mNIS+7 composite score or Norfolk QoL‐DN total score. Whilst the details of these analyses were not pre‐specified in the statistical analysis plan, the inclusion of PGIC and PGIS as exploratory endpoints in the trial was solely to facilitate anchor‐based analyses of meaningful differences.

2.4.1. Anchor‐Based Estimation of Meaningful Difference

Meaningful difference estimates were derived for each of the four outcome measures using an anchor‐based approach. This approach uses independent criterion measures, for which there are predefined indicators of clinically meaningful change, comparing them to corresponding changes in the outcome measures in order to estimate a meaningful difference [17].

In this analysis, the Spearman's correlation coefficient was used to estimate the association between the outcome measures and the PGIS, PGIC, SF‐36v2 GH01 item, and PND anchors, which are ordinal measures. As the NIS is a continuous variable, the Pearson's correlation coefficient was used to estimate the association between this anchor and the outcome measures. To be deemed appropriate, an anchor should be correlated with the target outcome measure. Any cut‐off used to quantify correlation coefficients may be interpreted as arbitrary, and therefore need to be considered within the specific trial context. Two cut‐off values for correlation were used in this analysis, ≥ |0.30| and ≥ |0.20| (for sensitivity analysis), corresponding to Cohen's d effect sizes of 0.41 and 0.63 [18], to assess the sensitivity of the results to the choice of correlation requirement.

Anchor‐based estimates of meaningful difference were based on scores from the outcome measures and anchor measures collected within 2 weeks of each other to maximize the concurrence of assessment. The difference between mean change values of paired anchor and outcome measures was calculated, mapping between interpretable categories of meaningful difference in the anchor measure and the range of values in the outcome measure. The use of different anchors with varying properties is expected to result in a range of plausible estimates. For each anchor, a “stable” group of patients was established, in which the change in the anchor measure was considered equivalent to no meaningful change in the outcome measure. This stable group was then compared with groups that showed meaningful improvement and meaningful deterioration.

Of the data collected in NEURO‐TTRansform, the PGIS, PGIC, the SF‐36v2 GH01 item, and the PND were deemed appropriate as potential anchors for all outcomes. The NIS was also determined to be an appropriate potential anchor for all outcome measures, except for the mNIS+7, as the NIS is included in the mNIS+7 (alongside other components) and therefore does not meet the anchor criteria of being external to the outcome measure [7, 16].

2.4.2. Distribution‐Based Estimates of Variability

Distribution‐based estimates, based on standard deviation (SD) at baseline and SDs for changes from baseline to weeks 35, 65/66, and 85, were calculated as a measure of variability in outcome measures. Distribution‐based estimates are not related to the patient experience and therefore cannot describe meaningful differences; however, they do provide insights into the variability of the outcome measures, so can serve as secondary, supplementary evidence to anchor‐based analyses.

2.4.3. Estimated Treatment Difference of Eplontersen Versus Placebo

An analysis of difference in least squares means (LSMs) for eplontersen minus placebo was conducted, using change from baseline in mNIS+7, Norfolk QoL‐DN, and NSC at week 66, and at week 65 for mBMI. Differences in LSMs were derived from a mixed‐effects model with repeated measures adjusted and propensity score weights [5]. These differences were compared with the anchor‐based estimates of meaningful difference to determine whether eplontersen had a clinically meaningful effect on each outcome measure. If the estimated treatment difference of eplontersen versus placebo exceeded the estimated anchor‐based meaningful difference, this indicated that the treatment effect was meaningful to patients.

2.4.4. Responder Analysis

The observed treatment effect of eplontersen consists of a mixture of some patients improving and others deteriorating. A comparison of the patient‐level likelihood of improvement or deterioration was performed using the bounds of the estimated range of meaningful difference, as meaningful improvement and deterioration are not inherently symmetrical. Response to the study drug was defined as greater improvement in within‐person change scores in an outcome measure, compared to the values at either end of the anchor‐based range of meaningful improvement and deterioration. Response was analyzed using a logistic regression model, adjusting for the baseline value of each outcome measure, Coutinho stage, V30M variant, and previous treatment with stabilizers. These variables were adjusted for because they were included in the main analysis of the primary and secondary endpoints.

3. Results

3.1. Patients

This analysis evaluated patients from the NEURO‐TTRansform trial who were randomized to eplontersen (n = 144) and compared with a historical placebo comprising patient data from the NEURO‐TTR trial (n = 60). The FAS comprised 141 patients from the eplontersen group and 59 patients from the placebo group. Three patients in the eplontersen group and one patient in the placebo group did not have any post‐baseline mNIS+7 or Norfolk QoL‐DN assessments.

Baseline characteristics of patients in the NEURO‐TTRansform trial, and of patients in the historical placebo group from the NEURO‐TTR trial, have been previously reported [4, 14]. Patient characteristics were generally similar between the eplontersen and historical placebo groups; however, patients in the eplontersen group tended to be younger and were more likely to have received treatment with TTR stabilizers. Baseline demographics and characteristics in the NEURO‐TTRansform trial, including mNIS+7, Norfolk QoL‐DN, and NSC scores, mBMI, and exploratory measures are shown in Tables 1 and 2.

TABLE 1.

Baseline demographics and characteristics.

	Eplontersen (n = 141)	Placebo (n = 59)
Age, mean (SD), years	52.9 (15.0)	59.4 (14.1)
Sex, n (%)
Female	43 (30)	18 (31)
Male	98 (70)	41 (69)
Race, n (%)
Asian	22 (16)	3 (5)
Black or African American	5 (4)	1 (2)
White	109 (77)	52 (88)
Other or multiple	4 (3)	3 (5)
Geographic region, n (%)
North America	21 (15)	26 (44)
Europe	52 (37)	22 (37)
South America/Australia/Asia	68 (48)	11 (19)
TTR variant, n (%)
V30M	83 (59)	33 (56)
Non‐V30M	58 (41)	26 (44)
Previous treatment with tafamidis or diflunisal, n (%)	99 (70)	35 (59)
Coutinho stage, n (%)
Stage 1 (ambulatory without assistance)	113 (80)	42 (71)
Stage 2 (ambulatory with assistance)	28 (20)	17 (29)
BMI, kg/m2	n = 135	n = 59
Mean (SD)	24.4 (4.9)	24.3 (4.9)
Body weight, kg	n = 138	n = 59
Mean (SD)	70.5 (15.9)	71.4 (18.1)
Albumin, g/L	n = 135	n = 59
Mean (SD)	42.0 (3.0)	43.5 (3.5)

Note: Full analysis set.

Abbreviations: BMI, body mass index; SD, standard deviation; TTR, transthyretin.

TABLE 2.

Baseline values of measured outcomes and anchors.

	Eplontersen (n = 141)	Placebo (n = 59)
PGIS of amyloidosis condition, n (%) a
None	6 (4)	N/A
Mild	35 (25)	N/A
Moderate	63 (45)	N/A
Severe	23 (16)	N/A
Very severe	3 (2)	N/A
SF‐36v2 GH01 item, n (%)
Excellent	0 (0)	1 (2)
Very good	8 (6)	8 (14)
Good	52 (37)	20 (34)
Fair	60 (43)	24 (41)
Poor	21 (15)	6 (10)
NIS score, mean (SD)	45.3 (28.9)	43.4 (24.7)
PND score, n (%)
I (sensory disturbances but preserved walking capability)	55 (39)	23 (39)
II (impaired walking capability but able to walk without a stick or crutches)	60 (43)	19 (32)
IIIa (walking only with the help of 1 stick or crutch)	15 (11)	14 (24)
IIIb (walking with the help of 2 sticks or crutches)	10 (7)	3 (5)
IV (confined to a wheelchair or bedridden)	0 (0)	0 (0)
mNIS+7 composite score, mean (SD)	79.8 (42.3)	74.1 (39.0)
Norfolk QoL‐DN total score, mean (SD)	43.3 (26.2)	48.6 (27.0)
NSC total score, mean (SD)	22.7 (12.0)	22.9 (12.7)
mBMI, kg/m2 × g/L	n = 135	n = 59
Mean (SD)	1027.7 (233.8)	1053.7 (228.5)

Note: Full analysis set.

Abbreviations: GH, general health; mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; N/A, not assessed; NIS, Neuropathy Impairment Score; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change; PGIS, Patient Global Impression of Severity; PND, Polyneuropathy Disability; SD, standard deviation; SF‐36v2, 36‐Item Short Form Survey version 2.

^a PGIS was not assessed in the NEURO‐TRR trial; therefore, these data were not available for the historical placebo arm.

3.2. Estimates of Meaningful Difference

Correlation coefficients between potential anchors and the outcome measures, which informed the identification of the appropriate anchors, are shown in Table 3. The criterion for this analysis required a correlation coefficient for change of ≥ |0.30|. With this criterion, one suitable anchor was established for mNIS+7, which was PGIC at week 85, and the estimates of meaningful improvement and deterioration are presented in Table 4. For Norfolk QoL‐DN, PGIC at weeks 37 and 85, SF‐36v2 GH01 item at weeks 35, 65, and 85, and NIS at weeks 66 and 85 were established as anchors, and estimates for meaningful improvement and deterioration are presented in Table 6. PGIC at weeks 37 and 85, and NIS at weeks 35, 66, and 85 were established as anchors for NSC scores, and estimates for meaningful improvement and deterioration are presented in Tables 4 and 5. For mBMI, NIS at week 66 was the only anchor to meet the correlation criterion and the estimates for meaningful improvement and deterioration are presented in Table 5. Estimates of meaningful differences based on patient‐reported and physician‐reported anchor measures are reported in Tables 4 and 5, respectively. Anchor‐based and distribution‐based estimates of meaningful differences are summarized in Table 6.

TABLE 3.

Correlation coefficients between potential anchors and outcomes at different timepoints.

Anchor	Correlation coefficient a
	mNIS+7 composite score	Norfolk QoL‐DN total score	NSC total score	mBMI (kg/m2 × g/L)
PGIS
Baseline	0.283	0.577	0.382	0.024
Week 35/37	0.343	0.597	0.405	−0.136
Week 85	0.352	0.473	0.410	−0.199
Change at week 35/37	−0.020	0.150	0.055	0.152
Change at week 85	−0.028	0.287	0.025	0.094
PGIC
Change at week 35/37	0.255	0.455	0.310	0.164
Change at week 85	0.331	0.501	0.351	−0.063
SF‐36v2 GH01 item
Baseline	0.306	0.474	0.455	−0.238
Week 35	0.324	0.540	0.466	−0.228
Week 65/66	0.367	0.584	0.492	−0.258
Week 85	0.303	0.570	0.451	−0.194
Change at week 35	0.128	0.339	0.235	−0.231
Change at week 65/66	0.239	0.474	0.285	−0.078
Change at week 85	0.181	0.435	0.195	−0.216
NIS score
Baseline	—	0.509	0.598	−0.273
Week 35	—	0.572	0.699	−0.311
Week 65/66	—	0.607	0.704	−0.300
Week 85	—	0.603	0.684	−0.339
Change at week 35	—	0.228	0.342	−0.077
Change at week 65/66	—	0.427	0.557	−0.314
Change at week 85	—	0.324	0.446	−0.058
PND score
Baseline	0.721	0.513	0.578	−0.256
Week 35	0.706	0.584	0.656	−0.288
Week 65/66	0.775	0.605	0.666	−0.362
Week 85	0.755	0.602	0.691	−0.379
Change at week 35	0.235	0.278	0.172	0.039
Change at week 65/66	0.268	0.201	0.129	−0.146
Change at week 85	0.214	0.098	0.103	0.163

Note: Bold and underlined numbers indicate correlations for change scores that meet the ≥ |0.30| criterion. Underlined numbers indicate correlations for change scores that meet the ≥ |0.20| criterion. Dash (—) signifies no available correlation. NIS was not assessed as a potential anchor for mNIS+7 due to the NIS score being included in the mNIS+7 composite score.

Abbreviations: GH, general health; mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; NIS, Neuropathic Impairment Score; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change; PGIC, Patient Global Impression of Change; PGIS, Patient Global Impression of Severity; PND, Polyneuropathy Disability; SF‐36v2, 36‐Item Short Form Survey version 2.

^a A Spearman's correlation coefficient was used for the PGIS, PGIC, SF‐36v2 GH01 item, and PND anchors. As the NIS score is a continuous variable, a Pearson's correlation coefficient was used for this anchor.

TABLE 4.

Estimates of meaningful differences from patient‐reported anchor measures.

Anchor‐based measures	Estimates of meaningful difference
	mNIS+7 composite score	Norfolk QoL‐DN total score	NSC total score	mBMI (kg/m2 × g/L)
PGIC: week 37
“A little worse” vs. “No change”	—	6.0	2.4	—
“A little better” vs. “No change”	—	−7.9	−2.2	—
PGIC: week 85
“A little worse” vs. “No change”	1.8	5.9	0.6	—
“A little better” vs. “No change”	−4.0	−4.0	−2.3	—
SF‐36v2 GH01 item: change from baseline at week 35
“Deteriorated 1 category” vs. “No change”	—	6.0	—	—
“Improved 1 category” vs. “No change”	—	−10.5	—	—
SF‐36v2 GH01 item: change from baseline at week 65
“Deteriorated 1 category” vs. “No change”	—	14.7	—	—
“Improved 1 category” vs. “No change”	—	−11.4	—	—
SF‐36v2 GH01 item: change from baseline at week 85
“Deteriorated 1 category” vs. “No change”	—	14.0	—	—
“Improved 1 category” vs. “No change”	—	−12.8	—	—

Note: The criteria for anchors were to assess similar concepts as the outcome measures and to statistically correlate with changes in the outcome measures (correlation coefficient for change ≥ |0.30|). A Spearman's correlation coefficient was used for the PGIS, PGIC, and SF‐36v2 GH01 item anchors. Dash (—) signifies that no estimate was available.

Abbreviations: GH, general health; mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change; PGIC, Patient Global Impression of Change; SF‐36v2, 36‐Item Short Form Survey version 2.

TABLE 6.

Estimates of ranges for meaningful differences from different anchors and distribution‐based estimates of variability.

	Estimates of meaningful difference
	mNIS+7 composite score	Norfolk QoL‐DN total score	NSC total score	mBMI (kg/m2 × g/L)
Anchor‐based range for meaningful difference: Improvement	−4.0 to −4.0	−12.8 to −1.4	−2.4 to −1.3	9.8 to 9.8
Anchor‐based range for meaningful difference: Deterioration	1.8 to 1.8	5.9 to 14.7	0.6 to 5.8	−40.9 to −40.9
Distribution‐based
SEm at baseline	5.8	6.2	4.6	—
50% of SD for change from baseline at week 35	8.1	9.1	3.8	40.9
50% of SD for change from baseline at week 65/66	11.3	10.5	4.3	49.4
50% of SD for change from baseline at week 85	10.3	9.0	4.1	49.7
50% of SD at baseline	20.7	13.2	6.1	116.0

Note: The criteria for anchors were to assess similar concepts as the outcome measures and to statistically correlate with changes in the outcome measures (correlation coefficient for change ≥ |0.30|). A Spearman's correlation coefficient was used for the PGIS, PGIC, and SF‐36v2 GH01 item anchors. As the NIS score is a continuous variable, a Pearson's correlation coefficient was used for this anchor. Dash (—) signifies that no estimate was available.

Abbreviations: GH, general health; mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; NIS, Neuropathy Impairment Score; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change; PGIC, Patient Global Impression of Change; SD, standard deviation; SEm, standard error of measurement; SF‐36v2, 36‐Item Short Form Survey version 2.

TABLE 5.

Estimates of meaningful differences from physician‐reported anchor measures.

Anchor‐based measures	Estimates of meaningful difference
	mNIS+7 composite score	Norfolk QoL‐DN total score	NSC total score	mBMI (kg/m2 × g/L)
NIS score: change from baseline at week 35
“Deteriorated ≥ 2 points” vs. “Changed < 2 points”	—	—	3.1	—
“Improved ≥ 2 points” vs. “Changed < 2 points”	—	—	−2.2	—
NIS score: change from baseline at week 66
“Deteriorated ≥ 2 points” vs. “Changed < 2 points”	—	10.7	5.6	−40.9
“Improved ≥ 2 points” vs. “Changed < 2 points”	—	−6.7	−2.4	9.8
NIS score: change from baseline at week 85
“Deteriorated ≥ 2 points” vs. “Changed < 2 points”	—	6.5	5.8	—
“Improved ≥ 2 points” vs. “Changed < 2 points”	—	−1.4	−1.3	—

Note: The criteria for anchors were to assess similar concepts as the outcome measures and to statistically correlate with changes in the outcome measures (correlation coefficient for change ≥ |0.30|). As the NIS score is a continuous variable, a Pearson's correlation coefficient was used for this anchor. Dash (—) signifies that no estimate was available.

Abbreviations: mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; NIS, Neuropathy Impairment Score; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change.

Anchors and estimates of meaningful difference using a less strict criteria of correlation coefficient for change of ≥ |0.20| are presented in Table S2. Using this criterion, PND was a suitable anchor for mNIS+7 and Norfolk QoL‐DN. Deterioration by one category on the PND score corresponded to a deterioration of between 17.4 and 24.2 points in the mNIS+7, and between 8.4 and 12.5 points in the Norfolk QoL‐DN (with ranges reflecting multiple timepoints of assessment).

3.3. Estimated Treatment Difference of Eplontersen Versus Placebo

Figure 1 shows the LSM difference between eplontersen and placebo for mNIS+7, Norfolk QoL‐DN, NSC, and mBMI alongside the anchor‐ and distribution‐based estimates of meaningful difference. The improvements seen in each measure with eplontersen treatment versus placebo are greater than the estimates of the meaningful differences. Therefore, the difference between the effect of eplontersen versus placebo can be considered meaningful.

FIGURE 1.

LSM difference (eplontersen minus placebo) in change from baseline, per outcome measure. Red lines indicate anchor‐based estimates of meaningful difference, with red shading demonstrating the range. Blue lines indicate distribution‐based estimates of meaningful difference. All estimates of meaningful difference are on the same side of zero because the difference between treatment versus placebo is not one of “improvement” or “deterioration,” it is a mixture of the two that reflects the clinically meaningful difference. CI, confidence interval; LSM, least squares mean; mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change.

The estimated cumulative distribution function (eCDF) of eplontersen and placebo, and the difference between these eCDFs, are reported in Figures S1–S8.

3.4. Responder Analysis

When using the anchor‐based estimates of meaningful difference for mNIS+7, Norfolk QoL‐DN, NSC, and mBMI to define responder thresholds, there was a significant difference between the proportion of responders in the eplontersen versus placebo group for every estimated threshold for each measure (Table 7). In the eplontersen group, a significantly greater proportion of patients responded with improvement and a significantly lower proportion responded with deterioration, compared with the placebo group.

TABLE 7.

Responder analysis by estimated thresholds for meaningful change.

Responder definition	Placebo, number of responders (%)	Eplontersen, number of responders (%)	Eplontersen vs. placebo, OR (95% CI)	p
mNIS+7 composite score
Improvement < −4	7 (13.5)	51 (39.8)	4.2 (1.7, 10.2)	0.002
< 0	10 (19.2)	68 (53.1)	5.1 (2.3, 11.5)	< 0.001
Deterioration > 1.8	41 (78.8)	52 (40.6)	0.2 (0.1, 0.4)	< 0.001
Norfolk QoL‐DN total score
Improvement < −12.8	5 (9.6)	41 (32.0)	8.4 (2.6, 26.3)	0.001
Improvement < −1.4	11 (21.2)	82 (64.1)	8.6 (3.8, 19.6)	< 0.001
< 0	12 (23.1)	83 (64.8)	8.0 (3.5, 17.9)	< 0.001
Deterioration > 5.9	34 (65.4)	25 (19.5)	0.1 (0.1, 0.3)	< 0.001
Deterioration > 14.7	20 (38.5)	15 (11.7)	0.2 (0.1, 0.4)	< 0.001
NSC total score
Improvement < −2.4	4 (7.7)	50 (37.9)	7.8 (2.6, 23.7)	< 0.001
Improvement < −1.3	6 (11.5)	62 (47.0)	7.0 (2.7, 17.9)	< 0.001
< 0	6 (11.5)	70 (53.0)	9.2 (3.6, 23.6)	< 0.001
Deterioration > 0.6	41 (78.8)	54 (40.9)	0.2 (0.1, 0.4)	< 0.001
Deterioration > 5.8	31 (59.6)	18 (13.6)	0.1 (0.0, 0.2)	< 0.001
mBMI (kg/m2 ×  g/L)
Improvement > 9.8	7 (13.5)	59 (45.4)	5.5 (2.3, 13.5)	< 0.001
Deterioration < −40.9	32 (61.5)	39 (30.0)	0.3 (0.1, 0.5)	< 0.001

Abbreviations: CI, confidence interval; mBMI, modified body mass index; mNIS+7, modified Neuropathy Impairment Score +7; Norfolk QoL‐DN, Norfolk Quality of Life‐Diabetic Neuropathy; NSC, Neuropathy Symptoms and Change; OR, odds ratio.

4. Discussion

This analysis established estimates of meaningful improvement and worsening for four measures of neuropathic impairment, health‐related QoL, and nutritional status in patients with ATTRv‐PN.

When assessing these estimates of meaningful change in comparison to the possible range of scores for each outcome, some of the estimates may seem relatively small. For example, as the mNIS+7_ionis range is –22.3 to 346.3 points [5], the estimated thresholds for improvement and deterioration may seem like small changes that fall within the range of random variability. However, the estimates derived were based on differences between categories within the anchor measures, indicating that they are meaningful and distinguishable to patients (for patient‐reported anchors) or clinicians (for clinician‐reported anchors). Further support for the estimates derived from the mNIS+7_Ionis is provided by the NEURO‐TTR trial, which reported that the intra‐rater reliability of mNIS+7_Ionis was 0.97 [19], demonstrating the high precision of this measure in an ATTRv‐PN patient population.

These analyses demonstrated that the estimates of meaningful improvement in each measure differed from the estimates for deterioration. Interestingly, the estimate of meaningful deterioration in mNIS+7 was 1.8, similar to the two‐point minimum degree of change in NIS lower limb (NIS‐LL) composite score that is detectable by a neurologist's examination [7]. The precision of the NIS‐LL score was also demonstrated in the NEURO‐TTR trial, with a high intra‐rater reliability of 0.98 [11]. It should be noted that the NIS‐LL has a score range up to 88 points, comprising only a quarter of the range of the mNIS+7 (up to 346.3 points). Additionally, estimates in this analysis were based on change from baseline at various timepoints (i.e., week 35/37, week 65/66, and week 85) and, in some cases, estimates derived from the same anchor differed between these timepoints. This suggests that the thresholds for meaningful differences can vary over time.

Although PND score was not an appropriate anchor for the outcome measures, which required a correlation coefficient ≥ |0.30|, the correlation coefficient for the change in PND and mNIS+7 composite score was between 0.2 and 0.3, as was the correlation coefficient for change in PND and Norfolk QoL‐DN (at two timepoints). As the PND score is widely used in clinical practice, it is worth highlighting that deterioration by one category in the PND score was associated with deterioration of 17.4 to 24.2 points in the mNIS+7 composite score, and 8.4 to 12.5 points in the Norfolk QoL‐DN (with ranges reflecting multiple timepoints of assessment). These estimates for Norfolk QoL‐DN fall within the range of estimates of meaningful deterioration generated by anchors meeting the criterion of a correlation coefficient ≥ |0.30|. Conversely, the relatively high estimate of deterioration for mNIS+7 based on PND is unlikely to represent the “minimum” meaningful change in this outcome, but rather reflect the fact that PND scoring is markedly less sensitive to functional change than scoring in the outcome measures. Therefore, a patient could have a meaningful change in health status without registering a change in the relatively broad PND staging. In anchor measures that have similar levels of granularity (in terms of margins of difference between each point on the anchor scale), such as the PGIC and the SF‐36v2 GH01, there was consistency of the results across the two correlation cutoffs, ≥ |0.30| and ≥ |0.20|, supporting the robustness of the findings (Table 3 and Table S2).

A previous analysis used data from the NEURO‐TTR trial to establish patient‐level thresholds of meaningful change (in the direction of deterioration only) in scores for mNIS+7 and Norfolk QoL‐DN [19]. For mNIS+7, a distribution‐based estimate for meaningful change of 12.2 was established using a different methodology than the anchor‐based approach used for this measure in the current analysis. For Norfolk QoL‐DN, an anchor‐based approach was used, as in the current analysis, in addition to distribution‐based estimates. SF‐36v2 measures (including the GH01, as in the current analysis) as well as the PCS, were used as anchors, yielding an estimated range of meaningful change from 7.2 to 8.8 [19]. This falls within the range of estimates for deterioration that were established in the current analysis (5.9 to 14.7), using the SF‐36v2 GH01 and PGIC as anchors.

There are limitations to this analysis. Primarily, no anchor is completely aligned with the outcome measures, as demonstrated by the correlation coefficients. This is an inherent limitation of the anchor‐based approach but should be taken into consideration when interpreting the results of this analysis. The anchor measures in this study are subject to variability and potential bias as the concept of “meaningfulness” is inherently subjective to each patient's perspective. There is also a risk of over‐reliance on the correlation coefficient data to determine what constitutes a suitable anchor, so a range of correlation cut‐offs was employed. Another potential source of misalignment is the difference between the time of assessment of the anchors and of the outcome measures. PGIC at week 37 was an anchor for Norfolk QoL‐DN and NSC scores; however, these were assessed at week 35. Although it is understood that neuropathic symptoms are unlikely to change on a clinically detectable level over a period of < 3–6 months, it is possible that patients may respond differently at the two different timepoints, and so the lack of concurrent assessment may compromise the anchor and outcome measures.

Some features of the measures assessed as anchors could also be limiting factors. For example, NIS is a clinician‐graded measure which could affect how appropriate this measure is for establishing differences that are meaningful to patients. Additionally, although a change in NIS of two points has been previously reported as the minimum a physician could detect, this is not empirically supported, and therefore remains a point of controversy [7, 11].

A further potential limitation is the use of a historical placebo in this trial. The potential heterogeneity of patients within the eplontersen and historical placebo groups may limit the extrapolation of these results to real‐world ATTRv‐PN patient populations.

5. Conclusion

Meaningful difference estimates were derived for four commonly used outcome measures in ATTRv amyloidosis. For each of these measures, eplontersen improved scores beyond their respective meaningful difference estimates, supporting that eplontersen had a clinically meaningful effect on neuropathic impairment, QoL, and nutritional status. The implications of this analysis go beyond understanding the effect of eplontersen. Meaningful difference estimates established for commonly used measures of neuropathic impairment and QoL in patients with ATTRv‐PN can be used to support interpretation of observed changes in mNIS+7, Norfolk QoL‐DN, NSC scores, and mBMI with treatments in future trials. This can help clinicians assess treatment effects to contextualize their impact on patients, as well as helping clinicians identify disease progression and offer educated support to patients about their continued care and treatment.

Author Contributions

Folke Folkvaljon: conceptualization, software, formal analysis, writing – original draft, writing – review and editing, visualization, supervision. Morie Gertz: investigation, resources, writing – review and editing. Julian D. Gillmore: writing – review and editing. Sami Khella: investigation, resources, writing – review and editing. Ahmad Masri: investigation, resources, writing – review and editing. Mathew S. Maurer: writing – review and editing. Márcia Waddington Cruz: investigation, resources, writing – review and editing. Jonas Wixner: investigation, resources, writing – review and editing. Jersey Chen: writing – review and editing. Barry Reicher: writing – review and editing. Jesse Kwoh: writing – review and editing. Aaron Yarlas: writing – review and editing. John L. Berk: investigation, resources, writing – review and editing.

Ethics Statement

We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Conflicts of Interest

John L. Berk has participated in ad hoc advisory committees for Alnylam, AstraZeneca, BridgeBio, Intellia, and Ionis. Folke Folkvaljon is an employee of AstraZeneca. Jersey Chen and Barry Reicher are employees of, and hold stock in, AstraZeneca. Julian D. Gillmore has acted as Advisor for Alnylam, AstraZeneca, ATTRalus, BridgeBio, Intellia, Ionis, and Pfizer. Morie Gertz reports receiving personal fees from Ionis, Alnylam, Prothena, Janssen, Sanofi, Juno, Physicians Education Resource, Johnson & Johnson, Celgene, and Research to Practice; serving on the data and safety monitoring board for AbbVie; receiving grants from Aptitude Health; receiving meeting fees from Ashfield and Sorrento; and developing educational materials for i3Health. Sami Khella reports receiving consulting fees from Ionis and Alnylam. Jesse Kwoh is an employee and stockholder of Ionis. Ahmad Masri has received research grants from Attralus, Cytokinetics, Ionis, and Pfizer; and consulting fees from Alexion, Attralus, BioMarin, Bristol Myers Squibb, Cytokinetics, Eidos, Haya, Ionis, Lexicon, Pfizer, and Tenaya. Mathew S. Maurer reports grant support from NIH R01HL139671 and R01AG081582‐01; grants and personal fees from Alnylam, Pfizer, Eidos, Prothena, and Ionis; and personal fees from AstraZeneca, Akcea, Intellia, and Novo Nordisk. Márcia Waddington Cruz is a principal investigator for the NEURO‐TTRansform trial and a consultant for Alnylam, AstraZeneca, Ionis, and Pfizer. Jonas Wixner reports receiving consulting fees from Akcea, AstraZeneca, Alnylam, Pfizer, and Intellia. Aaron Yarlas is an employee and stockholder of Ionis.

Supporting information

Data S1.

Data Availability Statement

Data underlying the findings described in this manuscript may be obtained in accordance with AstraZeneca's data sharing policy described at https://astrazenecagrouptrials.pharmacm.com/ST/Submission/Disclosure. Data for studies directly listed on Vivli can be requested through Vivli at www.vivli.org. Data for studies not listed on Vivli could be requested through Vivli at https://vivli.org/members/enquiries‐about‐studies‐not‐listed‐on‐the‐vivli‐platform/. AstraZeneca Vivli member page is also available outlining further details: https://vivli.org/ourmember/astrazeneca/.