Photodynamic Therapy With Verteporfin for Subfoveal Choroidal Neovascularizationin Age-Related Macular Degeneration: Results of an Effectiveness Study

Objective  To determine the postapproval effectiveness of photodynamic therapy(PDT) with verteporfin for the treatment of predominantly classic subfovealchoroidal neovascularization (CNV) secondary to age-related macular degeneration.

Methods  Forty-five consecutive patients treated with PDT for subfoveal CNV werecompared with an untreated historical control group. Control patients hadsubfoveal CNV and were first seen by us within 1 year before Health Canada'sapproval of verteporfin. Both groups were followed up for the developmentof significant visual loss, stability, or improvement. Multivariate modelswere constructed to evaluate the effectiveness of PDT, controlling for multiplecovariates (age, sex, baseline visual acuity, follow-up time, lesion size,and number of treatments).

Results  Significant differences were noted in the change in visual acuity betweenthose who did and did not receive PDT (χ² = 5.9, P = .048). Patients who received PDT were 2.9 times (95% confidenceinterval, 0.9-9.1) less likely to develop a moderate (>2 lines) visual loss(χ² = 3.2, P = .07). Controlling forcovariates, patients who received PDT were 13.7 times (95% confidence interval,1.4-132.6) more likely to develop a visual improvement of at least 1 line.

Conclusion  Compared with historical controls, PDT was demonstrated to be effectivefor the treatment of predominantly classic subfoveal CNV.

Age-related macular degeneration (AMD) is the leading cause of visualloss in people older than 50 years in North America¹ andis associated with significant reductions in quality of life.^2,3 Becausethis segment of the population is expected to double in size during the next25 years, the development of preventive and therapeutic strategies to combatthis disease constitutes a pressing concern.

Visual loss from AMD occurs mostly as a result of the development ofchoroidal neovascularization (CNV).⁴ Duringthe 1980s and 1990s, the Macular Photocoagulation Study group establishedthe value of thermal laser photocoagulation for patients with well-circumscribedjuxtafoveal and extrafoveal CNV, which constitute 10% to 15% of all neovascularlesions.^5,6 However, when appliedto subfoveal lesions, photocoagulation destroys the photoreceptors overlyingthe abnormal vessels, leading to immediate and irreversible loss of centralvision.^7-11 Morerecently, photodynamic therapy (PDT) was investigated as a therapeutic alternativefor these types of lesions because of its potential for reduced photoreceptordamage. Data from the Treatment of Age-Related Macular Degeneration With PhotodynamicTherapy (TAP) investigation demonstrated a reduction of visual loss in patientswith predominantly classic subfoveal CNV who received verteporfin PDT, comparedwith controls.¹² As a result, PDT has beenapproved in more than 40 countries, including Canada and the United States,for treatment of predominantly classic subfoveal CNV secondary to AMD.¹³

Randomized clinical trials represent the highest quality of evidenceavailable to evaluate treatment benefit.¹⁴ Bydefinition, these studies use randomization to assign patients to treatmentor control groups, thereby minimizing the selection bias that plagues mostother study designs. Randomized clinical trials also control for a host ofcovariates by using stringent patient selection criteria and a rigid treatmentprotocol. With this powerful experimental design, randomized clinical trialsare ideally suited to determine a treatment's efficacy, ie, the way in whichit works under ideal circumstances when delivered to select patients by providersmost skilled at providing it.¹⁵ Such rarifiedconditions, however, can severely limit the generalizability of a study'sresults to the general population in a more natural situation.¹⁶ In"real-life" settings, the patients being treated tend to have more advanceddisease and be less compliant. Furthermore, the resources available to implementa given treatment may be partially or totally unavailable.

A treatment's performance under ordinary conditions by the average practitionerand delivery system can be referred to as its effectiveness.¹⁵ Observational studies that followup patients over time to measure the strength of the association between treatmentand effect can be used to evaluate a treatment's effectiveness.¹⁷ Suchstudies, which can be prospective or retrospective in nature, usually lackrandomization of treatment and therefore cannot definitively establish a cause-and-effectrelationship. Nevertheless, they typically use a control group of patientsthat allows consideration of several important confounding variables by meansof established statistical methods. Therefore, effectiveness studies complementthe information obtained from a randomized clinical trial by informing abouta treatment's outcomes in clinical practice.

Although there is little doubt that the results of the TAP investigationdemonstrated the safety and efficacy of PDT via its 2 concurrently run randomizedclinical trials, 2 questions remain to be answered before PDT can be fullyadvocated. First, it must be established whether the observed benefits invision translate into significant gains in quality of life for patients.¹⁸ As previously shown through the use of Markov modelingand simulation, that is likely to be the case.^19,20 Second,it must be established whether an intervention's efficacy can be generalizedto other health care settings and patient populations.^21,22 Oneway of doing so is through the use of observational studies like the presentone, which can analyze a treatment's effectiveness in a real-life situation.The objective of this study was to evaluate the effectiveness of PDT withverteporfin in a tertiary care practice for the treatment of predominantlyclassic CNV secondary to AMD.

This study, which was approved by the Queen's University Research EthicsBoard, was conducted following Health Canada's approval for use of verteporfinin PDT for the treatment of predominantly classic subfoveal CNV secondaryto AMD.

Our treatment group consisted of 45 consecutive patients who underwentPDT for predominantly classic subfoveal CNV by one of us (S.S.) at a tertiarycare retinal practice, between November 12, 2000, and November 21, 2001. Patientswith non–age-related causes of CNV (eg, myopia, angioid streaks, ocularhistoplasmosis, and trauma), as well as those with greater than 50% occultCNV, were excluded from this study.

Photodynamic therapy was administered as per the TAP investigation protocol.Specifically, verteporfin was administered in a 10-minute infusion deliveredthrough the antecubital vein, followed by 83 seconds of 689-nm laser treatment.Adverse reactions observed during the procedure were documented. Patientswere followed up for at least 3 months at 3-month intervals, at which timethey had Snellen visual acuity testing, a stereoscopic fundus examination,and digital angiography performed. Patients with persistent intraretinal orsubretinal fluid (as identified by a stereoscopic contact examination witha contact lens) and persistent angiographic leakage of the lesion were offeredadditional treatment with PDT. Reasons for not offering additional treatmentincluded complete angiographic closure of the CNV, complete elimination ofintraretinal or subretinal macular fluid, or subjective worsening of visiongreater than 3 lines following treatment.

The primary clinical outcomes analyzed were (1) overall change in visualacuity, (2) likelihood of developing moderate (>2 lines) visual loss, and(3) likelihood of gaining at least 1 line of visual acuity, as measured bySnellen visual acuity testing.

We compared the visual progress of our patients with a group of historicalcontrol subjects, who were selected in a consecutive fashion from a databaseof an ongoing study evaluating the quality of life related to AMD. The patientsselected from this database had subfoveal CNV and were first seen by us beforethe approval of PDT with verteporfin by Health Canada. All of these patientshad no treatment performed for their lesions. Patients in this group manifestedlesions between June 1, 1999, and June 1, 2000, and were followed up for atleast 4 months.

An observer (A.O.-F.) who was not involved in patient treatment collecteddata from standardized forms that were filled out at the time of assessmentin the treatment group and at a point following the visit in the control group.The following information was abstracted from patients' medical charts: age,sex, baseline visual acuity, duration of follow-up, size of the lesions, numberof treatments, and visual acuities at the different follow-up times. All datawere entered into Excel 2000 (Microsoft Corp, Redmond, Wash) and SPSS 10.0for Windows (SPSS Inc, Chicago, Ill) spreadsheets.

Baseline data were analyzed by means of t testsfor continuous variables and χ² tests for dichotomous variables.Uncontrolled differences in visual acuity change and improvement in visualacuity between groups were evaluated using odds ratios and goodness-of-fitanalysis. The effect of PDT on changes in visual acuity, controlling for allknown potential covariates (age, sex, baseline visual acuity, duration offollow-up, size of the lesions, and number of treatments), was evaluated byanalysis of variance (ANOVA) testing. Logistic regression analysis was performedto estimate the likelihood of a gain in visual acuity across the range ofclinical variables. All analyses were performed by one of us (Mr Bakal) whowas not involved in patient care, data abstraction, or data recording andwho was masked to treatment allocation. (Allocation data were precoded bythe data abstractor [A.O.-F.] as 0 for PDT and 1 for natural history.)

Forty-nine consecutive patients who received PDT were recruited intothe study, 4 of whom were lost to follow-up after the initial treatment. Asearch of our quality-of-life database yielded 49 control patients. However,a review of their angiograms revealed that only 34 (69%) had the predominantlyclassic form of subfoveal CNV. Patients in the treatment group received amean of 3.3 treatments during a mean period of 13 months. As seen in Table 1, the treatment and control groupswere comparable in terms of age, sex, follow-up time, lesion size, and baselinevisual acuity.

A comparison of the net change in visual acuity of the treatment andcontrol groups showed them to differ considerably (χ² = 5.9, P = .048; log linear analysis = 5.58, P = .01) (Table 2). Although13 (38%) of 34 control patients progressed to a loss of 3 or more lines ofvision, this occurred in only 8 (18%) of 45 treated patients. This constitutesa 2-fold reduction in risk (relative risk, 2.2; 95% confidence interval, 0.9-5.1)for moderate visual loss in treated patients (χ² = 3.2, P = .07). Conversely, none of the control patients exhibitedvisual acuity gains beyond 2 lines, while 3 (7%) of those receiving PDT did.Furthermore, a statistically significant difference in the proportion of patientswho gained 1 or more lines of vision in each group was noted (3% [1/34] vs22% [10/45]; χ² = 6.1, P = .02, Fisherexact test) (Table 3).

To determine the simultaneous effect of patient characteristics andtreatment on visual acuity, we performed 2 types of multivariate analysis.First, we used ANOVA to isolate and assess the relative contribution of patientage, sex, baseline visual acuity, lesion size, length of follow-up, and numberof treatments to the changes in visual acuity that were measured during thestudy. The results of this ANOVA demonstrated that the type of treatment allocated(PDT vs observation) explained a significant proportion of the variation invisual acuity (F = 7.13, P = .008) (Table 4). This effect was second only to that attributable to thebaseline visual acuity (F = 34.02, P<.001), andno other measured attributes suggested any effect on visual improvement.

To further test the association between vision improvement and treatment,controlling for the effects of other variables, we next performed a logisticregression analysis. This test showed that patients who received PDT withverteporfin were 13.7 times (95% confidence interval, 1.4-132.6; P = .02) more likely to develop a visual improvement of 1 or more linesof vision (Table 5). In fact,of all the independent variables tested, treatment allocation was the bestpredictor of visual improvement (β = 2.62). None of the other variablesanalyzed achieved statistical significance in our model.

In the 138 administered treatments, 11 complications were documented.The most frequent complication was infusion-related back pain, which occurredin 7 instances (5%). Also documented were light-headedness (1 [0.7%]), jointpain (1 [0.7%]), hearing loss (1 [0.7%]), and transient double vision (1 [0.7%]).

Like the TAP investigation, our study assessed the proportion of patientswith predominantly classic CNV secondary to AMD who developed changes in theirvisual acuity following PDT or no treatment. Unlike the TAP investigation,our treatment allocation was not random in nature, as our study was conductedfollowing Health Canada's approval of PDT. Despite this and other implicitdifferences in study design, our study also demonstrated differences in visionbetween the treatment and historical control groups (χ² = 5.85, P = .048).

Following a mean of 3.3 treatments and 13 months of follow-up, patientsundergoing PDT with verteporfin were 2.9 times (95% confidence interval, 0.9-9.1)less likely to develop moderate visual loss (χ² = 3.2, P = .07). These results are comparable to those obtainedin the TAP investigation, with a mean of 2.7 treatments. In fact, using amultivariate analysis to control for potential confounders in our study, thosewho received PDT were 14.5 times more likely to improve their vision (P = .02). This was corroborated by ANOVA, which showedthat treatment differences explained a large proportion of the change in visualacuity (F = 7.13, P = .008) in our study population.In addition, similar to the TAP investigation, we found a low incidence oflocal and systemic adverse effects in those who received verteporfin therapy.

Our study has potential limitations. First, as with the TAP study, weare unable to predict the long-term effectiveness of this treatment; onlyongoing evaluation of these cohorts will allow for the evaluation of the long-termstability of our results. In addition, our analyses did not control for dietor the use of vitamin supplementation, both of which may have affected therelationship that exists between treatment allocation and primary outcome.

In conclusion, these findings suggest that the beneficial visual outcomesreported by the TAP investigators can be reproduced in a tertiary ophthalmicpractice. In addition, visual improvement was significantly more likely tooccur in those patients with predominantly classic subfoveal CNV.

Corresponding author and reprints: Sanjay Sharma, MD, MSc(Epid),Cost-Effective Ocular Health Policy Unit, Hotel Dieu Hospital, Brock II-224B,166 Brock St, Kingston, Ontario, Canada K7L 5G2 (e-mail: sanjay_sharma60@hotmail.com).

Submitted for publication June 11, 2002; final revision received November21, 2003; accepted November 21, 2003.

This study was supported by the Canadian Institutes of Health Researchand Canadian Foundation for Innovation, Ottawa, and the E. A. Baker Foundationand Canadian National Institute for the Blind, Toronto, Ontario; and by theJeanne Mance Foundation and Women's Auxiliary Fund, Hotel Dieu Hospital.

Dr Sharma holds a New Investigator Award from the Canadian Institutesof Health Research and a Premier's Research Excellence Award from the OntarioMinistry of Energy, Science and Technology, Toronto.