PEDIATRICS Vol. 105 No. 5 May 2000, pp. 1036-1040

Does Candidemia Predict Threshold Retinopathy of Prematurity in Extremely Low Birth Weight (1000 g) Neonates?

M. Gary Karlowicz, MD^*, Peter J. Giannone, MD^*, John Pestian, PhD^*, , Ardythe L. Morrow, PhD^*, , and Justine Shults, PhD^*,

From the ^* Department of Pediatrics and the  Center for Pediatric Research, Eastern Virginia Medical School, Children's Hospital of The King's Daughters, Norfolk, Virginia.

	ABSTRACT

Top Abstract Methods Results Discussion Conclusion References

Background.  Extreme prematurity is a risk factor for both candidemia and threshold retinopathy of prematurity (ROP) and may confound the reported association between these conditions.

Objective.  To determine if candidemia is an independent risk factor for threshold ROP.

Methods.  A cohort study was conducted of infants weighing 1000 g at birth using a prospectively maintained neonatal database. The study included infants admitted to the neonatal intensive care unit at 3 days of age between January 1, 1993 and December 31, 1997. We excluded infants not screened for ROP because they died, were discharged, or transferred. Threshold ROP (ie, requiring ablative therapy within 72 hours of diagnosis) was defined by the criteria of the American Academy of Pediatrics Section on Ophthalmology ROP subcommittee. Candidemia was defined as Candida species growth from at least 1 blood culture. Cox proportional hazards regression was used to determine independent risk factors for threshold ROP.

Results.  Six hundred fourteen infants weighing 1000 g at birth, of which 165 were excluded: 120 died before ROP screening, 40 were admitted >3 days of age, and 5 were discharged or transferred before ROP screening. A total of 449 infants were included in the study; 58 (13%) developed threshold ROP. Candidemia occurred in 58 (13%) infants before developing the worst stage of ROP. Candidemia occurred in 27 of 73 (37%) at 23 to 24 weeks' gestational age (GA), 25 of 197 (13%) at 25 to 26 weeks' GA, and 6 of 129 (5%) at 27 to 28 weeks' GA, 0 of 50 >28 weeks' GA. Similarly, threshold ROP occurred in 25 of 73 (34%) at 23 to 24 weeks' GA, 26 of 197 (13%) at 25 to 26 weeks' GA, and 6 of 129 (5%) at 27 to 28 weeks' GA, and 1 of 50 (2%) >28 weeks' GA. Threshold ROP developed in 19 of 58 (33%) infants with a history of candidemia and 39 of 391 (10%) without candidemia. Proportional hazards analysis indicated that GA in weeks (hazard ratio = .75; 95% confidence interval [CI]: .61, .93) and non-black ethnicity (hazard ratio = 1.8; 95% CI: 1.05, 3.08) were significantly associated with threshold ROP. After controlling for GA and other factors, candidemia did not remain significantly associated with threshold ROP (hazard ratio = 1.6; 95% CI: .89, 2.89).

Conclusion.  Candidemia may not be an independent risk factor for threshold ROP in extremely low birth weight infants. The magnitude of the previously reported association between candidemia and threshold ROP (more than fivefold) is unlikely and much of the clinically observed association appears to be mediated by gestational age.  Key words:  retinopathy, prematurity, candidemia, screening, Candida.

Retinopathy of prematurity (ROP) is a disease of incomplete retinal vascularization in infants.^1,2 Severe ROP, which places the infant at risk for vision loss, is known as threshold ROP.^3-6 When an infant develops threshold disease, the risk of retinal detachment and poor visual outcome is approximately 50% if not treated with ablative surgery.⁶ Threshold ROP is more prevalent in extremely low birth weight (ELBW or 1000 g) and extremely premature infants.⁷

ELBW infants are also prone to Candida sepsis,^8-10 and the incidence of candidemia has increased significantly in ELBW infants as their survival has improved.¹¹ A few clinical reports suggest a potential causal relationship between candidemia and increased risk of threshold ROP. Endophthalmitis occurs in premature infants with disseminated candidiasis.¹² Candida has been associated with choroidal neovascularization in adults.¹³ In 1992, Kremer et al¹⁴ reported that 8 of 15 ELBW infants with candidemia developed threshold ROP and required cryosurgery. This was only a case series, but it suggested an association that warranted further investigation. In 1998, Mittal et al¹⁵ reported that Candida sepsis in ELBW infants was significantly associated with increased severity of ROP and a more than fivefold increased need for laser surgery.

We undertook this study to test the hypothesis that both candidemia and severe ROP are outcomes of extreme prematurity, but that candidemia, independent of extreme prematurity, is not significantly associated with threshold ROP.

	METHODS

Top Abstract Methods Results Discussion Conclusion References

Study Population

A cohort study was conducted of all ELBW infants admitted 3 days of age to the neonatal intensive care unit (NICU) at Children's Hospital of The King's Daughters between January 1, 1993 and December 31, 1997. This hospital contains the regional referral nursery for southeastern Virginia and northeastern North Carolina. Infants who were admitted >3 days of age were excluded because they were transfers from other NICUs and were not considered to be representative of our NICU population because the prevalence of candidemia, procedures, and general care may be different in other NICUs. We excluded infants who did not receive ROP screening because they died, were discharged, or transferred. Infants who died were included in the study if they lived long enough to have ROP screening.

Database Management

The neonatology division maintains several databases that consist of prospective abstraction of information from medical records of infants admitted to the NICU. Data were entered into databases by research nurses or by staff trained in data collection and entry that were supervised by research nurses. Data entry was closely monitored and periodically reviewed by research nurses and the senior clinical investigator (M.G.K.) to reduce human error. There were 3 separate databases: 1) a neonatal database that consisted of basic demographic, morbidity, and outcomes data; 2) an ROP screening database; and 3) a candidemia database containing clinical details about episodes of candidemia. Data from the 3 databases were integrated into a study database. The study was approved by the hospital's institutional review board.

Definitions

Candidemia was defined as Candida species growth from at least 1 blood culture from a peripheral or central venous sample.¹¹ Severity of ROP was staged by the International Classification of ROP.¹⁶ Threshold ROP was defined by the criteria of the American Academy of Pediatrics Section on Ophthalmology ROP subcommittee, that is, stage 3 ROP, zone I or II in 5 or more continuous clock hours or 8 cumulative clock hours with the presence of plus disease. Chronic lung disease was defined as need for supplemental oxygen >28 days.¹⁷ Infants with severe intraventricular hemorrhage (defined as grades 3 and 4) were combined with infants with periventricular leukomalacia for data analysis.

Clinical Data

Initial ROP screening was performed before 6 weeks of age by 1 of 4 board-certified ophthalmologists and staged accordingly. Follow-up ophthalmologic examinations were performed at 1- to 4-week intervals until vascularization proceeded to zone III. The screening ophthalmologists were unaware of histories of candidemia or any other potential risk factors for ROP other than very low birth weight or gestational age (GA) 28 weeks. Infants with threshold ROP identified by screening ophthalmologists were referred to 1 of 3 retina specialists who confirmed the diagnosis and performed retinal ablative therapy, if indicated. Parents, primary care physicians, and ophthalmologists were contacted to determine ROP outcome for infants who were discharged or transferred to other hospitals with advanced stages of prethreshold ROP. In addition, the hospital medical records database was searched for any study infants who had Current Procedural Terminology codes for either laser or cryosurgery for ROP to identify infants who developed threshold ROP after discharge from the NICU.

The attending neonatologist determined GA when the infant was admitted to the NICU. Blood cultures were collected and processed according to standard microbiology techniques. Ophthalmologic examinations were not routinely requested as part of diagnostic evaluation of infants with candidemia, because it would not have changed the duration of antifungal therapy.¹¹ Candidal endophthalmitis may occur less often than originally reported,¹² since Mittal et al¹⁵ found no cases in 22 ELBW infants with candidemia (95% confidence interval [CI]: 0, 14%).

Statistical Methods

Clinical data were electronically transferred from the neonatal database, ROP screening database, and the candidemia database into the study database for analysis by microcomputer. Parametric data are expressed as mean ± SEM. Nonparametric data are expressed as median (range); comparisons between groups were made with the Mann-Whitney or Wilcoxon test. Categorical data were analyzed using the ² or Fisher's exact test as appropriate. Significance was set at P < .05 and measures of relative risk with 95% CIs were computed. Cox proportional hazards regression was used to determine the hazards of threshold ROP associated with GA, birth weight, ethnicity, chronic lung disease, severe intracranial hemorrhage, Apgar scores, and candidemia.

We selected the proportional hazards model to account for the differing follow-up times (length of stay) per patient in this study. Failure time was determined to be age in days when the diagnosis of threshold ROP occurred, or total length of follow-up time for those who never developed threshold ROP. Factors that were significantly associated with risk of threshold ROP and/or candidemia from univariate analyses were entered into proportional hazards regression models. Variables that were not significant at P < .15 or less in the multivariable model were removed using a stepwise backward elimination procedure. Finally, interaction was assessed between GA and candidemia in relation to threshold ROP.

	RESULTS

Top Abstract Methods Results Discussion Conclusion References

Study Group

A total of 614 infants weighing 1000 g at birth were admitted to NICU during the study period. One hundred sixty-five were excluded: 120 died before ROP screening, 40 were admitted >3 days of age, and 5 were transferred or discharged before ROP screening. The study group consisted of 449 infants.

Twenty-three infants in the study group died but lived long enough to have ROP screening. Their median length of stay was 116 days (range: 48-321). Among the 23 infants who died, there were 4 who were diagnosed with candidemia before their worst stage ROP and 5 with threshold ROP, none of whom had candidemia.

Risk Factors for Threshold ROP

Threshold ROP occurred in 58 (13%) infants. There were 11 infants who were discharged from the NICU with prethreshold ROP, and 3 of 11 progressed to threshold ROP requiring laser therapy. Infants developed threshold ROP at a median age of 87 days (range: 50-193 days). Figure 1 shows the association of GA with threshold ROP. Table 1 lists clinical factors significantly associated with threshold ROP by univariate analysis.

View larger version (29K): [in this window] [in a new window]   Fig. 1.   The association of GA with candidemia and threshold ROP in infants weighing 1000 g at birth.

Clinical Features of Infants With Candidemia

Candidemia occurred in 58 (13%) infants before the worst stage ROP. Candidemia was detected at a median age of 27 days (range: 11-95 days). Table 2 compares clinical features of infants with candidemia to those of infants without candidemia. Infants with candidemia had a significantly higher incidence of chronic lung disease, ROP, and threshold ROP. Development of candidemia appeared to be significantly associated with extreme low birth weight and extreme prematurity. Figure 1 shows the association of GA with candidemia as well as with threshold ROP.

Candidemia and ROP

Threshold ROP developed in 19 of 58 (33%; 95% CI: 21%, 46%) infants with a history of candidemia compared with 39 of 391 (10%; 95% CI: 7%, 13%) infants without a history of candidemia (P < .001). The shortest interval between occurrence of candidemia and threshold ROP was 13 days. There were 4 cases of candidemia that occurred after the worst stage ROP, including 2 infants with threshold ROP, which were not included as cases of candidemia in the risk factor analysis, because the timing of occurrence did not support a causal association.

Proportional hazards analysis of time to occurrence of threshold ROP found a crude (unadjusted) hazards ratio of 2.15 (95% CI: 1.24, 3.73; P = .007) for candidemia (Table 3). When GA in weeks was included in the model (model 2, Table 3), candidemia was no longer significantly associated with threshold ROP (hazard ratio = 1.62, 95% CI: .91, 2.88). The full regression model including all factors associated with threshold ROP at P < .15, included candidemia, GA in weeks, non-black ethnicity, and Apgar score at 1 minute (model 3, Table 3). Of these factors, only GA and non-black ethnicity were independently significant (P < .05). After adjusting for non-black ethnicity, Apgar score at 1 minute as well as GA, the hazard ratio for candidemia was 1.60 (95% CI: .89, 2.89), virtually the same estimate obtained by adjustment for GA only. We did not find evidence of interaction with GA in the relationship between candidemia and threshold ROP.

	DISCUSSION

Top Abstract Methods Results Discussion Conclusion References

Our data suggest that candidemia may not be an independent risk factor for the development of threshold ROP. Rather, we found GA to be a common risk factor for both candidemia and threshold ROP, and that GA confounds the crude observed association between candidemia and threshold ROP. We describe a significant linear trend of decreasing threshold ROP and candidemia with increasing GA.

The only previous cohort study of candidemia and threshold ROP is that of Mittal et al¹⁵ who reported that for ELBW infants, 41% (9 of 22) with candidemia compared with 9% (10 of 111) without candidemia required laser surgery for threshold ROP. Multiple logistic regression analysis indicated that candidemia was independently associated with threshold ROP in ELBW infants with an odds ratio of 5.6 (95% CI: 1.4, 23). Thus, the association that they reported was greater than the upper 95% CI of the association reported in this study. The potential reasons for differences in results between these studies may include patient selection criteria, accounting for the timing of candidemia relative to ROP onset, patient follow-up, and methods of analysis. Mittal et al¹⁵ excluded from their study 30% of potentially eligible subjects because they were infants who died, whose records were unavailable, or who were transferred to other institutions before ROP screening. By contrast, in this study, we excluded only 1% of eligible infants because of transfer or discharge, and we included 23 infants who died but lived long enough to be at risk of ROP and to have ROP screening. Among the infants who died, there were 4 who had candidemia before their worst stage ROP and 5 who developed threshold ROP, none of whom had a history of candidemia. These data suggest that exclusion of such infants is not warranted and may lead to biased results.

These studies also differed in criteria for classifying exposure and identifying outcomes. Mittal et al¹⁵ included all cases of candidemia, even those that occurred after the worst stage of ROP when a causal relationship could not exist, and did not report cases of threshold ROP occurring in study infants after discharge, so any such infants would have been misclassified as not having threshold ROP. Follow-up of infants in our study identified 3 who developed threshold ROP after discharge. When we applied logistic regression to the analysis of our data, as Mittal et al¹⁵ had done; we found the adjusted odds ratios to be biased upward when compared with use of the proportional hazards model (adjusted odds ratio = 2.23; [95% CI: 1.08, 4.58] vs adjusted hazard ratio = 1.60; [95% CI: .89, 2.89]). In contrast to survival analysis, logistic regression ignores differing lengths of patient follow-up (ie, length of stay in the NICU). Survival analysis is the more appropriate method, because length of stay differed significantly among study infants.

We found that infants with non-black ethnicity had significantly increased risk of threshold ROP, which is consistent with the findings of other investigators that white infants have a highly significant risk of developing threshold ROP (odds ratio: 2.76, P < .001).^7,18,19 The reason black infants have less than half the risk of severe ROP is unknown. Monos et al²⁰ reported that very low birth weight infants with darkly pigmented fundi have a significantly lower risk of severe ROP. They speculated that large amounts of melanin might protect retinas by acting as scavengers of superoxide radicals.

It is recommended that infants with threshold ROP be considered for ablative therapy within 72 hours of diagnosis.¹⁶ Therefore, it is imperative that infants at high risk of developing threshold ROP be identified to ensure these infants receive properly scheduled and timely ROP screening examinations. Many reports have shown that infants of lower GA are at higher risk for threshold ROP.^7,21,22 However, these studies grouped together all infants 28 weeks as the youngest GA group. Yet, there are significant differences in morbidity and mortality in each successive week between 23 and 28 weeks.²³ We believe that outcomes studies of severe ROP should distinguish between 23 to 24 weeks', 25 to 26 weeks', and 27 to 28 weeks' GA. Important prognostic information is lost by lumping them together into 1 group. We found that infants of 23 to 24 weeks' GA have significantly higher risk of developing threshold ROP when compared with infants >24 weeks' GA (Fig 1). Other investigators have suggested similar findings.^18,24 Identifying this subgroup of infants 23 to 24 weeks' GA as those at highest risk of threshold ROP may facilitate timely detection, prevention, and treatment of threshold ROP allowing for a better visual outcome.

Consistency, strength, and temporal relationship of an association are key elements in judging the possibility of a causal relationship between a risk factor and a disease outcome.²⁵ Our findings are not consistent with those of Mittal et al,¹⁵ who reported a significant independent association between candidemia and threshold ROP. Further, the strength of association found in this study was lower and not consistent with the strength of association previously reported. The principle of temporal sequence, that the hypothesized causative factor precedes disease onset, supports our decision to exclude cases of candidemia that occurred after threshold ROP.

	CONCLUSION

Top Abstract Methods Results Discussion Conclusion References

In summary, candidemia does not appear to be an independent risk factor for development of threshold ROP. Rather, our data suggest that the lowest GAs predict the development of both threshold ROP and candidemia. It should be noted, however, that the 95% CI for the candidemia hazards ratio (1.60) ranged from .89 to 2.89. Thus, this study cannot rule out the possibility that an independent association may exist between candidemia and threshold ROP. We speculate that the remaining observed association may be attributable to chance or to other clinical factors that we were unable to measure or control for. Nevertheless, the magnitude of the previously reported association between candidemia and threshold ROP (more than fivefold) is unlikely and much of the clinically observed association appears to be mediated by GA.

	ACKNOWLEDGMENTS

We thank Marilyn Reininger, RN, Mary Fish, RN, and Christy Vaughn for meticulously maintaining the neonatal database, the candidemia database, and the ROP screening database.

	FOOTNOTES

Received for publication Mar 30, 1999; accepted Sep 1, 1999.

Address correspondence to M. Gary Karlowicz, MD, 601 Children's Lane, Norfolk, VA 23507. E-mail: gkarlowi{at}chkd.com

	ABBREVIATIONS

ROP, retinopathy of prematurity; ELBW, extremely low birth weight; NICU, neonatal intensive care unit; GA, gestational age; CI, confidence interval.

	REFERENCES

Top Abstract Methods Results Discussion Conclusion References

1. An international classification of retinopathy of prematurity. Arch Ophthalmol. 1984;102:1130-1134
2. The International Committee for Classification of the Late Stages of Retinopathy of Prematurity: an international classification of retinopathy of prematurity II The classification of retinal detachment. Arch Ophthalmol. 1987; 105:906-912 [Medline]
3. Phelps DL Retinopathy of prematurity. Pediatr Clin North Am. 1993; 40:705-714 [Medline]
4. Cryotherapy for Retinopathy of Prematurity Cooperative Group Multicenter trial of cryotherapy for retinopathy of prematurity: preliminary results. Arch Ophthalmol. 1988; 106:471-479 [Abstract]
5. Cryotherapy for Retinopathy of Prematurity Cooperative Group Multicenter trial of cryotherapy for retinopathy of prematurity: three-month outcome. Arch Ophthalmol. 1990; 108:195-204 [Abstract]
6. Cryotherapy for Retinopathy of Prematurity Cooperative Group Multicenter trial of cryotherapy for retinopathy of prematurity: one-year outcome. Arch Ophthalmol. 1990; 108:1408-1416 [Abstract]
7. Palmer EA, Flynn JT, Hardy RJ, Incidence and early course of retinopathy of prematurity. Ophthalmology. 1991; 98:1628-1640 [Medline]
8. Johnson DE, Thompson TR, Green TP, Systemic candidiasis in very low birth weight infants (<1500 g). Pediatrics. 1984; 73:138-143 [Abstract/Free Full Text]
9. Baley JE, Kliegman MD, Fanaroff AA Disseminated fungal infections in very low birth weight infants. Pediatrics. 1984; 73:144-152 [Abstract/Free Full Text]
10. Faix RG, Steven KM, Shaw TR, Mucocutaneous and invasive candidiasis among very low birth weight (<1500 g) infants in intensive care nurseries: a prospective study. Pediatrics. 1989; 83:101-107 [Abstract/Free Full Text]
11. Kossoff EH, Buescher ES, Karlowicz MG Candidemia in a neonatal intensive care unit: trends during fifteen years and clinical features of 111 cases. Pediatr Infect Dis J. 1998; 17:504-508 [CrossRef][Medline]
12. Baley JE, Annable WL, Kliegman RM Candidal endophthalmitis in the premature infant. J Pediatr. 1981; 98:458-461 [CrossRef][Medline]
13. Jampol LM, Sung J, Walker JD, Choroidal neovascularization secondary to Candida albicans chorioretinitis. Am J Ophthalmol. 1996; 121:643-649 [Medline]
14. Kremer I, Naor N, Davidson S, Systemic candidiasis in babies with retinopathy of prematurity. Graefes Arch Clin Exp Ophthalmol. 1992; 230:592-594 [Medline]
15. Mittal M, Dhanireddy R, Higgins RD Candida sepsis and association with retinopathy of prematurity, Pediatrics. 1998; 101:654-657 [Abstract/Free Full Text]
16. American Academy of Pediatrics, the American Association for Pediatric Ophthalmology and Strabismus, and the American Academy of Ophthalmology Screening examination of premature infants for retinopathy of prematurity. Pediatrics. 1997; 100:273 [Free Full Text]
17. Avery ME, Tooley WH, Keller JB, Is chronic lung disease preventable? A survey of eight centers. Pediatrics. 1987; 79:26-30 [Abstract/Free Full Text]
18. Saunders RA, Donahue ML, Christmann, LM et al Racial variation in retinopathy of prematurity. Arch Ophthalmol. 1997; 115:604-608 [Abstract]
19. Shaffer DB, Palmer EA, Plotsky DF, Prognostic factors in the natural course of retinopathy prematurity. Ophthalmology. 1993; 100:230-237 [Medline]
20. Monos T, Rosen SD, Karplus M, Yassur Y Fundus pigmentation in retinopathy of prematurity. Pediatrics. 1996; 97:343-348 [Abstract/Free Full Text]
21. Reynolds JD, Hardy RJ, Kennedy KA, Lack of efficacy of light reduction in preventing retinopathy of prematurity. N Engl J Med. 1998; 338:1572-1576 [Abstract/Free Full Text]
22. Repka MX, Hudak ML, Parsa CF, Calf lung surfactant extract prophylaxis and retinopathy of prematurity. Ophthalmology. 1992; 99:531-536 [Medline]
23. Allen MC, Donohue PK, Dusman AE The limit of viability-neonatal outcome of infants born at 22 to 25 weeks' gestation. N Engl J Med. 1993; 329:1597-61 [Abstract/Free Full Text]
24. Wright K, Anderson ME, Walker, et al Should fewer infants be screened for retinopathy of prematurity in the managed care era? Pediatrics. 1997; 102:31-34 [Abstract/Free Full Text]
25. Breslow L Some sequels to the Surgeon General's Report on Smoking and Health: thirty years later. Ann Epidemiol. 1996; 6:372-375 [CrossRef][Medline]