Sickle cell disease (SCD) is among the most common, life-threatening monogenic disorders in the world, with recent estimates suggesting that 50,000-100,000 people in the United States suffer from SCD1 and that there are approximately 1.5 million hemoglobin S carriers and 25,000-35,000 SCD patients in Europe.18 Proliferative sickle cell retinopathy (PSCR) is a common complication of SCD. Goldberg10 described 5 PSCR stages characterized by peripheral retinal ischemia, peripheral neovascularization, intravitreal hemorrhage, and tractional or mixed retinal detachment that can lead to vision loss.2,6
Although SC genotype, lower fetal hemoglobin (HbF), and higher hemoglobin levels are associated with PSCR severity, other potentially contributory factors have not been investigated.7 We conducted the current study to examine potential associations between SCD characteristics (clinical and laboratory) and PSCR severity by analyzing SS and SC adult patients seen at our national SCD referral center.
PATIENTS AND METHODS
Patients were identified in our center's database at a university hospital, where a multidisciplinary team including ophthalmologists, hematologists, orthopedists, cardiologists, and nephrologists provides ongoing care and support. SS and SC patients had a complete medical check-up every 6 and 12 months, respectively, in the referral center. Referred patients had a complete physical examination and baseline laboratory analysis. They also had an ophthalmic examination at each visit. These patients, who came from continental France, its overseas territories (mainly Martinique, Guadeloupe, and Guyana), and sub-Saharan Africa had various types of sickle genotypes (for example, sickle β-thalassemia and Hb Sβ0-thalassemia) or other sickle cell syndromes. Fewer than 10% of the SS patients and none of the SC patients received hydroxyurea treatment. Fewer than 10% of SS patients were chronically transfused.
We selected patients with homozygous (SS) or heterozygous (SC) SCD who were seen in our center between 1990 and 2009. Patients with sickle cell trait, compound Sβ-thalassemia heterozygosity, or another red-cell disease were excluded. Patients without a precise diagnosis of homozygous or heterozygous SCD and patients without a retinal examination within 1 year of the last examination at the referral center were also excluded (Figure 1).
PSCR diagnosis and staging were based on the last ophthalmic examination that included fundus examination and fluorescein angiography. PSCR features were graded according to the Goldberg classification10 as follows: stage I, peripheral retinal ischemia (peripheral arteriolar occlusions); stage II, peripheral arteriovenous anastomosis; stage III, peripheral neovascularization (Figure 2); stage IV, intravitreous hemorrhage; and stage V, retinal detachment. Fluorescein angiography was not systematically performed for all patients, but was used in case of doubtful diagnosis of PSCR or for patients with stage III disease or worse. Peripheral retinal neovascularization was detected by fluorescein staining and leakage from the preretinal neovascularization during angiography. PSCR grading for each patient was based on the eye with the highest Goldberg stage.
Other relevant clinical data were collected as part of routine medical examinations performed at each visit, and from the updates recorded by the multidisciplinary team. These data included present or past organ involvement. Organs affected included the heart (heart failure, cardiomyopathy), lungs (pulmonary embolism diagnosed by vascular imaging, documented pulmonary hypertension with pulmonary arterial pressure ≥40 mm Hg, or restrictive syndrome >20%), bones (osteonecrosis diagnosed by magnetic resonance imaging [MRI] or radiographs), kidneys (proteinuria), central nervous system (stroke and cerebral vasculopathy diagnosed by cerebral MRI showing vascular stenosis of large arterial vascular trunks), ears (deafness or tinnitus), and gallbladder (cholecystectomy for cholelithiasis or cholecystitis). Also recorded were documented history of other clinical findings frequently associated with SCD: thrombophlebitis, acute chest syndrome, priapism, systemic infection (for example, salmonellosis), septic shock, endocarditis, bacterial pericarditis, bacterial pneumonia, tuberculosis, osteomyelitis, pyelonephritis, and bacterial meningitis; and less common manifestations, such as hypertension, myocardial infarction, valvulopathy, arrhythmia, renal papillary necrosis, chronic tubulointerstitial nephropathy, and chronic renal insufficiency.
Baseline hematology and biochemistry values included hemoglobin, HbF, and mean corpuscular volume during the first consultation before starting any treatment (such as blood transfusions or hydroxyurea). Serum iron and serum ferritin levels were recorded from the last on-site consultation if more than 3 months after the last transfusion or vasoocclusive event.
Continuous data are reported as median (interquartile range), and categorical data as number (%). Due to their skewed distribution, laboratory values were log transformed for statistical analyses; odds ratios (ORs) and 95% confidence interval (CI) are given for 1 standard deviation of the log-transformed levels. Because the prevalence of PSCR in SC and SS patients is known to be very different, with higher rates of PSCR in SC patients than in SS patients,2 we analyzed the 2 forms of SCD separately. All analyses were age-adjusted using logistic-regression models.
Patient and disease characteristics were described and compared according to the presence or absence of PSCR. Variables associated with PSCR in the first age-adjusted univariate analysis (p < 0.15) were analyzed with the patients stratified into 3 groups according to the severity of PSCR: 1) no PSCR, 2) moderate PSCR (Goldberg stage I-II), or 3) severe PSCR (Goldberg stage III-V). Multinomial logistic-regression models were used to estimate the age-adjusted OR (95% CI). The 3 PSCR strata were taken as the possible outcomes of the same nominal qualitative outcome variable, and the no-PSCR group served as the reference category. Variables associated with PSCR severity in the second age-adjusted univariate analysis (p < 0.15) were considered for multivariate analysis after testing for first-order interactions and confounding. We then conducted multivariate multinomial logistic regression to identify factors independently associated with PSCR severity. We also tested whether the β-coefficients (Ln OR) were equal across equations using the Wald test. The goodness-of-fit of the model was evaluated using the Hosmer-Lemeshow statistic;13 the Harrell c-index was used to assess discrimination.11 Kaplan-Meier curves were used to evaluate the relationship between age and the probability of remaining PSCR-free. All comparisons were 2-sided, with p ≤ 0.05 considered significant. Data were analyzed using Stata (Stata/SE 11, StataCorp 2009, College Station, TX) and LogXact 8 (Cytel Inc., Cambridge, MA) statistical software.
Among the 1361 patients in our center's database, there were 1171 patients with hemoglobin SS (815 patients; 70%) and hemoglobin SC (356 patients; 30%). The age range was 15-66 years for SC patients and 8-64 years for SS patients. Among these SCD patients, data from a retinal examination within 1 year of the last visit to the center were available for 942 patients (629 with SS and 313 with SC disease) (see Figure 1). Among SC patients, 247 (78.9%) had PSCR and 165 (52.7%) had at least stage III, including 25 patients with previous retinal detachment. Among SS patients, 404 (64.2%) had PSCR and 113 (18%) had at least stage III, including 6 with previous retinal detachment (Table 1).
Heterozygous SC patients were significantly more likely to develop severe PSCR than SS patients (OR, 3.95; 95% CI, 2.71-5.77; p < 0.001), but no difference was observed for moderate PSCR forms (p = 0.57).
Associations of PSCR and Patient Characteristics
PSCR was strongly associated with age for SC and SS patients (p = 0.005 and p = 0.0001, respectively). As shown in the Kaplan-Meier curves (Figure 3), the probabilities of remaining free of PSCR decreased with age for both SC and SS patients.
According to our first age-adjusted analysis (Table 2), PSCR in SC patients was significantly associated with pulmonary involvement, deafness or tinnitus, and acute pyelonephritis, with borderline trends (p ≤ 0.15) for baseline HbF, serum ferritin levels, systemic infection, and osteomyelitis. The same analysis for SS patients showed PSCR to be significantly associated with male sex, basal HbF, acute chest syndrome, and cholecystectomy, with borderline trends for serum ferritin level, osteonecrosis, pulmonary involvement, thrombophlebitis, acute pyelonephritis, and osteomyelitis.
No association could be established between higher hemoglobin concentrations and PSCR in SS and SC males. Indeed, the median baseline hemoglobin concentrations for men without PSCR did not differ significantly from men with PSCR in both SC (12.6 [12-13] vs. 12.8 [12-13.5]; p = 0.2 for SC with no PSCR and with PSCR, respectively) and SS groups (9 [8-10] vs. 9 [8.4-10]; p = 0.6 for SS with no PSCR and with PSCR respectively). Similar results were also observed among women in both groups (p = 0.9 and p = 0.3 for SC and SS with no PSCR and with PSCR, respectively). The effects of hydroxyurea on PSCR were not assessed.
Age-Adjusted Univariate and Multivariate Analysis: Associations of PSCR and Patient Characteristics
Univariate analysis confirmed Kaplan-Meier curves, showing that, for the SC patients, age was significantly associated with severe PSCR (p = 0.001), but not with the milder forms (p = 0.54) (Table 3). In addition, age-adjusted univariate analysis indicated that moderate forms of sickle cell retinopathy were significantly associated with pulmonary involvement, deafness or tinnitus, and acute pyelonephritis, while severe PSCR was associated with lower serum ferritin, deafness or tinnitus, and osteomyelitis, with trends observed for a higher frequency of systemic infections for moderate forms and lower baseline HbF for severe PSCR. No significant interactions were observed among the variables with p < 0.15 in univariate analysis.
Because none of the 66 reference-group SC patients (those with no PSCR) had previous acute pyelonephritis, no multivariate risk estimate could be computed. According to our age-adjusted multivariate analysis, pulmonary involvement and deafness or tinnitus were independently associated with moderate and severe PSCR. Systemic infection was not associated with any PSCR severity group because of the strong relationship with osteomyelitis (p = 0.01). Severe PSCR was also significantly associated with lower serum ferritin and the absence of osteomyelitis, and with a borderline trend for baseline HbF. Both models (moderate or severe PSCR vs. no PSCR) showed good calibration (Hosmer-Lemeshow statistic p = 0.37 and 0.36, respectively) and good discrimination (respective Harrell c-indexes of 0.76 and 0.78).
For SS patients, univariate analysis again confirmed Kaplan-Meier curves, showing that age was significantly associated with both moderate and severe PSCR (p = 0.011 and p < 0.001, respectively). The age-adjusted univariate analysis indicated that moderate and severe PSCR were significantly associated with male sex, lower baseline HbF, and lower serum ferritin, but only moderate PSCR was significantly associated with absence of thrombophlebitis and with osteomyelitis. Trends were observed toward higher frequencies of cholecystectomy for patients with moderate PSCR, and higher frequencies of pulmonary involvement, acute chest syndrome, pyelonephritis, and osteomyelitis and a lower frequency of thrombophlebitis among patients with severe PSCR.
No significant interactions were observed among the variables with p < 0.15 in univariate analysis. According to our age-adjusted multivariate analysis of SS patients (see Table 3), an older age and lower baseline HbF were independently associated with moderate PSCR, with trends observed for lower serum ferritin, acute chest syndrome, and osteomyelitis. Severe PSCR was significantly associated with older age, male sex, and acute pyelonephritis, with trends observed for lower baseline HbF and acute chest syndrome.
The parameters (β-coefficients) associated with moderate PSCR differed significantly from those associated with severe PSCR (p < 0.0001), particularly age (p < 0.0001), sex (p = 0.01), and acute pyelonephritis (p = 0.04). Both models (moderate or severe PSCR vs. no PSCR) showed good calibration (Hosmer-Lemeshow statistic: p = 0.30 and 0.45, respectively) and discrimination (Harrell c-indexes of 0.74 and 0.79, respectively).
Large numbers of SS (n = 629) and SC (n = 313) patients were followed by a multidisciplinary team in a single SCD referral center. Because the prognoses of the SS and SC forms of SCD differ and each form is associated with different clinical outcomes, we analyzed these groups of patients separately. Analysis of this large SCD population enabled us to establish several significant associations between retinal features of PSCR and other clinical features of SCD. Severe PSCR forms (stages III-V) were independently associated with older age, lower serum ferritin, pulmonary involvement, deafness or tinnitus, and absence of osteomyelitis for SC patients, and were associated with older age, male sex, lower baseline HbF, and acute pyelonephritis for SS patients. Life-span is almost normal for SC patients and is shortened for SS patients, but all are prone to ocular complications related to PSCR. Identification of clinical and/or laboratory criteria could lead to reliable prediction models for PSCR severity that would facilitate the management of SC and SS SCD.
It is noteworthy that severe PSCR (stages III-V) was more likely among people with hemoglobin SC than hemoglobin SS, although hemoglobin SS is usually associated with more severe non-ophthalmic disease. In the current study, 52.7% of SC and 17.9% of SS patients had severe PSCR. In a sickle cell center in Florida, stage III-V PSCR represented 76% of 29 SC patients and 12.8% of 109 SS patients.2 The higher rate of severe PSCR in SS patients in our cohort could be explained by the older age of our patients (mean age, 29 yr for our SS patients vs. 20 yr for the Florida patients). Although severe PSCR was less frequent in our SC patients than in the Florida study, the prevalence of severe PSCR was higher among our SC patients than in studies from Burkina Faso and Jamaica.3,4
When we considered the risk factors associated with severe PSCR, the results of our analyses showed that SS men had a 2.5-fold higher risk of developing severe PSCR than women. This protective female-sex effect was not observed in the SC group. Although a protective sex effect has not been reported in pediatric series,8,24 it was observed in another large adult series.7 The precise mechanisms by which SS women might be protected against severe PSCR have not been completely elucidated. Higher levels of HbF have been reported in the female population compared to the age-matched male population,5,23 but this was not observed in the current study. Estrogen levels could be involved in this protective sex effect through a vascular protective effect by increased endothelial nitric oxide synthase activity and decreased vascular leukocyte accumulation, as demonstrated in a mouse model of ischemia and reperfusion injury.21 Perhaps a nitric oxide deficit characteristic of SCD patients heightens the risk of coagulation system activation, leading to PSCR.22 SS patients are more prone to vascular occlusion than SC patients, and this tendency may be accentuated among SS males.
The current study showed a trend between lower baseline HbF and severe PSCR for SC and SS patients in multivariate analysis (p = 0.07 and 0.067) (see Table 3). A similar independent protective effect of higher HbF among SS and SC patients of both sexes was observed in another study.7 The beneficial effect of higher HbF in SCD was demonstrated by asymptomatic SS patients with pancellular hereditary persistence of HbF and is thought to accrue through a sparing effect on intracellular hemoglobin S-polymerization.12,20 This protective effect of HbF can be potentiated by hydroxyurea, which stimulates HbF production in SCD.17 A significant association was previously established between higher hemoglobin concentrations and PSCR in SS and SC males.7 The same result has not been verified in the present study.
Our second age-adjusted univariate analysis identified pulmonary involvement, deafness or tinnitus, and acute pyelonephritis as associated with PSCR for SC patients, and acute chest syndrome and cholecystectomy for SS patients. Our age-adjusted multivariate analysis retained pulmonary involvement, deafness or tinnitus, and osteomyelitis as associated with severe PSCR for SC patients, as opposed to only acute pyelonephritis for SS patients.
We acknowledge that these results are derived from a longitudinal study, and not from a prospective study. Although a significant linkage is clearly observed between the involvement of various organs and severe PSCR, these associations cannot presently lead to a clinical score correlated to the risk of severe PSCR, and further prospective studies are needed to validate such a clinical score.
In a 2007 editorial, Roberts and de Montalembert proposed the establishment of a clinical network and the analysis of clinical data on the natural history of SCD as major challenges and priorities for the future.18 In the current analysis we have established independent associations between PSCR severity and other organ involvement in adult SS and SC patients, yielding good predictability models for severe PSCR in both forms of SCD.
In a 2008 study that used a similar strategy, the authors concluded that neither sex nor systemic manifestations were predictive of the frequency or age of retinopathy onset in a cohort of pediatric patients.8 However, that negative finding may be partly explained by the rarity of SCD-related organ involvement in pediatric patients.
A broad spectrum of sickle cell complications ranging from the viscosity-vaso-occlusive phenotype to the hemolysis-endothelial dysfunction phenotype has been described, with each segment of the spectrum associated with specific clinical complications.9,19 Kato et al16 described a pathophysiologic model of SCD that associated certain clinical complications primarily reflecting blood viscosity and vaso-occlusion, and other complications mainly associated with hemolysis-endothelial dysfunction, that could partly explain some associations observed between PSCR and other general SCD complications. PSCR was significantly associated with pulmonary hypertension in our SS patients. The significant association between pulmonary hypertension and hemolysis-endothelial dysfunction led us to hypothesize that PSCR may be associated with this phenotype. PSCR in our SS patients was also associated with acute chest syndrome, which is associated with the viscosity-vaso-occlusive phenotype of SCD. Although these associations suggest that both mechanisms may be involved in the pathophysiology of PSCR in SS patients, our data are not complete enough to conclude that PSCR is associated with either the hemolysis phenotype or the viscosity-vaso-occlusive phenotype.
For SS patients, acute pyelonephritis was significantly associated with severe PSCR. Although this link remains unexplained, genetic susceptibility-to-infection factors may also be risk factors for retinal ischemia leading to PSCR in SS patients. Further studies are needed to confirm this result.15
We found that PSCR severity in SC patients was associated with hearing loss or tinnitus. A literature review suggests that the main etiology of SCD-related hearing loss is cochlear ischemia.14 Vaso-occlusive events occurring in SCD patients can cause occlusion of the terminal auditory internal artery, thereby generating cochlear anoxia or hypoxia, resulting in unilateral or bilateral tinnitus and/or mild to moderate high-frequency sensorineural hearing loss. Because of the high cochlear sensitivity to anoxia, these clinical outcomes could precede PSCR, suggesting that clinical assessment of hearing loss could be beneficial when screening for sickle cell retinopathy. Although these clinical associations emphasize a possible pathophysiologic link between ear, renal, and retinal SCD manifestations, they may also suggest a temporal sequence, with the ear and kidney being affected before the peripheral retina. This last hypothesis should be explored in further prospective studies.
It is puzzling that prior osteomyelitis was retained as a risk factor for stage I-II PSCR in SS patients, and that this clinical manifestation was associated with a lower risk of stage III-V PSCR in SC patients, in both age-adjusted univariate and multivariate analyses. Because more than 50% of the patients in the current study were from sub-Saharan Africa and came to France during adulthood, it is likely that they did not receive appropriate prophylactic treatment. As a result, it is possible that recruitment bias may partly explain the positive association between osteomyelitis and PSCR in SS patients.
In conclusion, the associations described here between clinical and laboratory factors and PSCR severity demonstrate the complexity of this ophthalmic disorder, the pathophysiology of which may involve several complex mechanisms. These findings should contribute to developing sight-saving preventive strategies for SCD patients. Prospective studies are needed to establish a clinical scoring system that can predict with acceptable sensitivity and specificity the risk of developing severe PSCR.
We are indebted to Janet Jacobson and Dr. John Somner for their help with the final revision of the manuscript.
1. Brawley OW, Cornelius LJ, Edwards LR, Gamble VN, Green BL, Inturrisi C, James AH, Laraque D, Mendez M, Montoya CJ, Pollock BH, Robinson L, Scholnik AP, Schori M. National Institutes of Health Consensus Development Conference statement: hydroxyurea treatment for sickle cell disease. Ann Intern Med
2. Clarkson JG. The ocular manifestations of sickle-cell disease: a prevalence and natural history study. Trans Am Ophthalmol Soc
3. Diallo JW, Sanfo O, Blot I, Meda N, Sawadogo P, Ouedraogo A, Simpore J. [Epidemiology and prognostic factors for sickle cell retinopathy in Ouagadougou (Burkina Faso).] J Fr Ophtalmol
4. Downes SM, Hambleton IR, Chuang EL, Lois N, Serjeant GR, Bird AC. Incidence and natural history of proliferative sickle cell retinopathy: observations from a cohort study. Ophthalmology
5. El-Hazmi MA, Warsy AS, Addar MH, Babae Z. Fetal haemoglobin level-effect of gender, age and haemoglobin disorders. Mol Cell Biochem
6. Fadugbagbe AO, Gurgel RQ, Mendonca CQ, Cipolotti R, dos Santos AM, Cuevas LE. Ocular manifestations of sickle cell disease. Ann Trop Paediatr
7. Fox PD, Dunn DT, Morris JS, Serjeant GR. Risk factors for proliferative sickle retinopathy. Br J Ophthalmol
8. Gill HS, Lam WC. A screening strategy for the detection of sickle cell retinopathy in pediatric patients. Can J Ophthalmol
9. Gladwin M, Sachdev V, Jison M, Shizukuda Y, Plehn JF, Minter K, Brown B, Coles WA, Nichols JS, Ernst I, Hunter LA, Blackwelder WC, Schechter AN, Rodgers GP, Castro O, Ognibene FP. Pulmonary hypertension as a risk factor for death in patients with sickle cell disease. N Engl J Med
10. Goldberg MF. Natural history of untreated proliferative sickle retinopathy. Arch Ophthalmol
11. Harrell FE Jr, Lee KL, Califf RM, Pryor DB, Rosati RA. Regression modeling strategies for improved prognostic prediction. Stat Med
12. Herman EC, Conley CL. Hereditary persistence of fetal hemoglobin. A family study. Am J Med
13. Hosmer DW, Lemeshow S. Applied Logistic Regression
. New York: Wiley & Sons; 1989.
14. Hotaling AJ, Hillstrom RP, Bazell C. Sickle cell crisis and sensorineural hearing loss: case report and discussion. Int J Pediatr Otorhinolaryngol
15. Joannes MO, Loko G, Deloumeaux J, Chout R, Marianne-Pepin T. Association of the +874 T/A interferon gamma polymorphism with infections in sickle cell disease. Int J Immunogenet
16. Kato GJ, Gladwin MT, Steinberg MH. Deconstructing sickle cell disease: reappraisal of the role of hemolysis in the development of clinical subphenotypes. Blood Rev
17. Platt OS, Orkin SH, Dover G, Beardsley GP, Miller B, Nathan DG. Hydroxyurea enhances fetal hemoglobin production in sickle cell anemia. J Clin Invest
18. Roberts I, de Montalembert M. Sickle cell disease as a paradigm of immigration hematology: new challenges for hematologists in Europe [Editorial]. Haematologica
19. Rother RP, Bell L, Hillmen P, Gladwin MT. The clinical sequelae of intravascular hemolysis and extracellular plasma hemoglobin: a novel mechanism of human disease. JAMA
20. Schechter AN, Rodgers GP. Hydroxyurea in sickle cell disease [Comment]. N Engl J Med
21. Simoncini T, Hafezi-Moghadam A, Brazil DP, Ley K, Chin WW, Liao JK. Interaction of oestrogen receptor with the regulatory subunit of phosphatidylinositol-3-OH kinase. Nature
22. Solovey A, Kollander R, Milbauer LC, Abdulla F, Chen Y, Kelm RJ Jr, Hebbel RP. Endothelial nitric oxide synthase and nitric oxide regulate endothelial tissue factor expression in vivo in the sickle transgenic mouse. Am J Hematol
23. Steinberg MH, Hsu H, Nagel RL, Milner PF, Adams JG, Benjamin L, Fryd S, Gillette P, Gilman J, Josifovska O, Hellman-Erlingsson S, Safaya S, Huey L, Rieder RF. Gender and haplotype effects upon hematological manifestations of adult sickle cell anemia. Am J Hematol
24. Talbot JF, Bird AC, Maude GH, Acheson RW, Moriarty BJ, Serjeant GR. Sickle cell retinopathy in Jamaican children: further observations from a cohort study. Br J Ophthalmol