A Response to: Letter to the Editor Regarding “LIGHTSITE II Randomized Multicenter Trial: Evaluation of Multiwavelength Photobiomodulation in Non-exudative Age-Related Macular Degeneration”

Dear Editor in Chief,

With great interest we have read the Letter to the Editor from Grisanti et al. regarding the LIGHTSITE II (LTII) clinical publication on the use of photobiomodulation (PBM) in subjects with intermediate dry age-related macular degeneration (AMD). Some comments allude to poor clinical trial design or execution, and we believe these comments are misstated. The study was carefully planned as a companion study to the independent United States (US) LIGHTSITE III (LTIII) trial and was started in advance as the US trial was being planned and in formal discussions with the US Food & Drug Administration (FDA). The study was unique in evaluating patients with dry AMD and absence of geographic atrophy (GA) involving the fovea and best-corrected visual acuity (BCVA) ranging from 20/32 to 20/100. The inclusion criteria were very strict and maintained with anatomical evidence of dry AMD with vision loss and presence of drusen, and/or non-foveal GA. Fulfillment of anatomical inclusion criteria were verified by the reading center before each subject was included. The study was powered for the primary endpoint of BCVA. All anatomical assessments, such as GA lesion area, were secondary or exploratory endpoints. The European Union (EU) study mirrored the US dry AMD trial design aspects that were approved by the FDA and was to be conducted similarly to create potential synergy between studies. Global key opinion leaders and top research institutes in the field of AMD were engaged in both the LTII and LTIII trials. Their collaborative involvement was paramount from the early stages of the trial design discussions and implementation. A multifunctional relationship between trials was established.

The LTII study employed a prospective randomized controlled trial (RCT) design including double masking with the Valeda^® Light Delivery System (Valeda)—unheard of in PBM clinical trials. Randomization of subjects to either the PBM or sham treatment groups was conducted with a third-party vendor IWR (interactive web response) system. The study included an independent, leading global imaging reading center well experienced in clinical trials of AMD. The imaging center was also masked to treatment and qualified each subject prior to enrollment. Trial monitoring, database management, and statistical analyses were conducted with an outside top-tier contract research organization (CRO) with the most experience in AMD trial conduct in the EU. The trial centers were trained on all clinical endpoints by a third-party ophthalmology certification company to reduce site-to-site variability, including BCVA measurements. Equipment was standardized across centers for all clinical and imaging outcome measures. The study was composed of three retinal centers from Germany, two from the UK, one from Spain, one from France, and one from Italy, providing a multinational EU cohort.

The single most important aspect of the LTII trial was the impact of the global pandemic. Trial design and patient balances were impacted by extraordinary circumstances during this time. The trial was enrolling and in the middle of required in-clinic treatments when the global COVID-19 pandemic emerged and paralyzed Europe, requiring citizens to isolate. The disruption to the trial was immense and the sponsor, in discussion with all centers, unanimously agreed to halt the study for 3 months. No treatments were conducted as the world came to grips with COVID-19. Upon trial restart, with pending risk of further pandemic spikes, the study allowed the enrolled subjects to complete their treatments where possible, but no further subjects were enrolled. The LTII study was compromised in terms of size, treatment interval, and incomplete dosing but the underlying results were considered still reflective of clinical safety and efficacy of PBM, which is now confirmed in the LTIII results. The investigators took the approach that all data was worth analyzing but the focus was spent on the subjects that completed the trial with all three cycles of treatment.

The first criticisms of the authors include their disappointment of the size of the trial and unequal distribution in AREDS categories as a flaw, not mentioning the pandemic circumstances surrounding the trial and the subsequent termination of the study which prematurely ended enrollment. They point out group imbalances by selectively combining eyes grouped into AREDS 3 and 4. Overall, a total of 64.7% versus 68.4% of eyes in the PBM and sham groups were categorized to AREDS category 3, respectively. A numeric imbalance in AREDS category 4 eyes (PBM, n = 1; sham, n = 5) was observed, which was the minority of patients. When stratified to the full protocol analysis, the number of eyes in AREDS category 4 was reduced to one PBM and two sham-treated eyes. This numerical difference in distribution did not impact the primary analysis, which was performed using only the PBM-treated group. As there were no sham subjects in this analysis, any imbalance in AREDS between PBM and sham treatment groups would not have any effect on this analysis. The BCVA secondary analyses that included both PBM and sham subjects were numerically in favor of PBM, the hypothesis tests were not significant. Inclusion of AREDS as a covariate, if favorable towards sham as a result of imbalance, would not improve these results. The conclusion and interpretation of these data would not change.

Further Grisanti et al. were concerned about the similar amount of drusen volume in both groups despite the supposedly higher number of more advanced eyes in the sham group. This concern is confusing to understand and seemingly unjustified. There were a similar number of eyes in the AREDS 3 group, which consists of eyes categorized with either non-center-involving GA or multiple intermediate and/or large drusen. There were only a higher number of eyes with AREDS 4 category, defined by GA involving the central 1 mm ETDRS grid but sparing the central 500 µm diameter. It is well known in patients with dry AMD that with disease progression drusen area usually increases overtime, but in some cases drusen volume can also decrease, which is often followed by GA development. Given the fact that AREDS 3 eyes can present with either many medium and large size drusen (and therefore a large drusen volume) or GA (and therefore a lower drusen volume), it is not surprising that we see a similar and balanced amount of drusen volume in both groups. Thus, a higher AREDS category does not necessarily translate to a higher drusen volume. Grisanti et al. also mention and criticized the variable response in terms of drusen volume in the PBM group of the earlier LIGHTSITE I (LTI) and LTII trial. It is noteworthy here that in all three LT studies, the included patient population was different: LTI included patients with the most severe dry AMD, with the majority of eyes having dry AMD AREDS category 4; the LTII trial with the majority of eyes belonging to AREDS category 3 with presence of non-center-involving GA; the LTIII trial, which also included mainly eyes with AREDS category 3, but only 3% had any GA at baseline. Therefore, the effect of PBM on quantitative drusen volume change may have differed. What is most noteworthy is that we see a consistent beneficial effect of PBM compared to the sham treatment in all three trials and that the drusen volume of the PBM group over the course of the trials increased less than in the sham group. Drusen can regress over the course of AMD, and this can be a sign for disease progression as Grisanti et al. pointed out very nicely. Usually, an increase of drusen volume is associated with disease progression. In our LT trials we consistently observed a lower increase of drusen volume in the PBM group compared to the sham group. Even more importantly, we detected a lower incidence of iRORA and cRORA development over the course of 12 and 24 months in eyes without iRORA and cRORA in the LTIII trial and a lower incidence of a conversion from iRORA to cRORA after 24 months. Altogether, these findings support a beneficial disease-modifying effect.

We are somewhat surprised about the comment of the similar baseline BCVA and the supposedly heterogenous distribution of the AREDS category of our LTII cohort. As pointed above, only the number of AREDS 4 patients were numerically not evenly distributed. Our study included AREDS category 4 eyes which were allowed to have central 1 mm ETDRS involving GA, but this GA had to spare the central 500 µm. Thus, the foveola of all our subjects was spared. Therefore, it is not surprising that BCVA was similar at baseline in both groups. In the full protocol analysis that completed all visits, a total of one eye categorized as AREDS 4 in the PBM group and two eyes categorized as AREDS 4 in the sham group were included which presents less of a disparity than in the modified intent-to-treat (mITT) group as noted by Grisanti et al. What this demonstrates in the full protocol subset of subjects is disease progression as the sham group got worse in BCVA letter score, while the difference in BCVA increased between groups when subjects received all PBM treatments as per the protocol. This is also confirmed in the LIGHTSITE III trial results at 13 and 24 months.

We fully agree that the study was not powered to draw valid conclusions about any effect on GA progression, which is also stated in our paper. The mean baseline size of GA lesion was indeed numerically larger in the sham compared to the PBM group; however, no significant difference was observed between PBM and sham groups (p = 0.51). For our exploratory analysis, the square root transformation was utilized to account for baseline GA lesion differences and in the LTI trial (also not powered to draw any conclusion) we report a similar effect. The Fig. 5 legend presenting GA lesion growth in a representative PBM and sham subject provides an incorrect value for the PBM lesion size at month 10. The reading center has confirmed that the area of 0.78 mm² should be corrected to 0.74 mm² with the difference of 0.16 mm² reported accurately.

As noted by the authors, the mean BCVA score for the sham group in the mITT analysis and the full protocol analysis were both 70.53 (SD 5.02). The mITT data was correct and mistakenly duplicated for the sham group (full protocol) in the table. The corrected sham mean BCVA score for the full protocol group went from 70.53 (SD 5.02) to 70.50 (SD 4.98) (see corrected BCVA table values below) (Table 1). This was a misprint in the table and did not impact any analysis or other data presented in the manuscript. The study also showed the separation of the treatment groups as the sham subjects showed progression to later stages of AMD by the end of the trial. The recent completion of the LTIII data analysis [1, 2] confirms the same expected separation between the treatment groups with disease progression in the sham arm at 13 months extending to 24 months.

Table 1 BCVA baseline subject characteristics and clinical outcomes (corrected)

The authors of the letter further point out the reductions in drusen in later-stage subjects in LTI versus the stabilization of drusen in the LTII study as inconsistent, but the trials did not enroll similar patient populations. If the mechanisms of PBM are at the cellular mitochondrial level increasing ATP generation, the improved retinal health would shift the balance from extracellular deposition of drusen material to normal intracellular breakdown. This has been nicely demonstrated with PBM in β-amyloid mice models, wherein they are genetically modified to overproduce β-amyloid leading to extracellular deposits, enhanced inflammatory responses, mitochondrial dysfunction, and functional losses in learning and memory with age [3]. These degenerative anatomical, biochemical, and clinical outcomes were attenuated with PBM in the animal model. β-amyloid is a major contributor to drusen and overlapping pathology is seen across central nervous system (CNS) and retinal degenerative diseases.

Drusen is a surrogate for retinal cellular health and in all PBM studies the underlying tissue has not shown signs of phototoxicity or progression to GA with the removal of drusen following PBM treatment. Extracellular drusen removal by complement and other inflammatory pathways becomes the primary route of removal—different processes versus the normal intracellular breakdown of proteins. Drusen reductions have been seen across clinical trials in subjects with PBM treatments. In subjects with earlier-stage dry AMD treated with PBM, deposition may be less, or stabilized. This was recently confirmed in the 13-month analysis of LTIII. It replicated that lower levels of drusen deposition at baseline and reductions were seen in PBM versus sham groups at both 13 and 24 months.

Standardized outcome measures used across ophthalmology clinical trials were used in the current trial. Independent groups trained and certified each site on trial protocols and outcome measurements were conducted under masked conditions. The training records for each site were recorded, all centers were diligent in their training and collection of data, the centers were masked, patients were masked, and the imaging center was masked, so the underlying comments by the authors of the letter seem more intent on casting doubt on the trial results for reasons we cannot explain.

The further interpretation that the reading center, masked to treatment, was somehow biased in measuring drusen values in favor of PBM benefits is completely unfounded. The publication includes a clear statement that the reduction in drusen in the PBM group was not a statistically significant effect. Furthermore, a small number of GA measurements were calculated using both absolute as well as the r² analysis to remove impact of lesion size. The LTII data, despite the conditions of the pandemic, show results consistent with previous trials which now have been confirmed and expanded on in the larger 2-year LTIII trial that has completed, showing sustained improvements in BCVA and slowing of new GA.

It is worth noting that the timing of this letter to the editor is over 1 year from publication of the LTII trial. As of August 2023, a total of 9682 academic papers (editorials, lab work, clinical trials, reviews, systematic reviews, etc.) and 1263 randomized controlled trials have been conducted using PBM therapy in a variety of indications. In addition, more than 30 clinical studies using PBM have been published in ophthalmology with the vast majority showing clinical and/or anatomical benefits, no safety concerns, and an overall positive opinion to further research efforts into this arena. The safety and effectiveness of PBM has been extensively demonstrated in published nonclinical studies in animal models of ocular disease and/or safety and toxicity studies.

It is not surprising to see differing reports among clinical studies that utilize PBM. That is the reality of differing clinical trial designs, differences in the indication and subsequent pathological underpinnings of the target disease, and the nuances of establishment of optimized treatment specifics using a new technology with a multitude of modifiable parameters. It is also a reflection of the complexity of a degenerative disease with many contributing factors. None of the authors of this Letter to the Editor have any first-hand experience with Valeda nor have they reached out to request clarification on any of the points they raised in the clinical trials. However, the growing volume of PBM data being generated by many investigators is not trivial—it is a dynamic and growing field of research. The science has been shown to work both in vitro and in vivo with animal studies and has repeatedly shown benefits in clinical populations with PBM under the right conditions, appropriate wavelengths, dosing, and safeguards on the devices.

Healthy skepticism is appropriate for emerging technologies. We thank the authors for highlighting two inaccuracies in the original publication which we now correct. A small study interrupted by COVID is difficult to draw strong conclusions from, but we believe it usefully adds to the growing body of evidence suggesting that PBM has some beneficial effects in dry AMD.