Article Text

Survival of retinoblastoma in less-developed countries impact of socioeconomic and health-related indicators
1. S Canturk1,
3. V Khetan3,
4. Z Ma4,
5. A Furmanchuk5,
6. C B G Antoneli6,
7. I Sultan7,
8. R Kebudi8,
9. T Sharma3,
10. C Rodriguez-Galindo2,
11. D H Abramson9,
1. 1Department of Ophthalmology, Faculty of Medicine, Istanbul University, Istanbul, Turkey
2. 2Department of Oncology, St Jude Children's Research Hospital, Memphis, Tennessee, USA
3. 3Medical and Vision Research Foundation, Sankara Nethralaya, Chennai, India
4. 4Department of Pediatric Oncology, West China Second University Hospital, Chengdu, China
5. 5Department of Oncology, Belarussian Research Centre for Pediatric Oncology-Hematology, Minsk, Belarus
6. 6Department of Pediatric Oncology, Hospital AC Camargo, Sao Paulo, Brazil
7. 7Department of Pediatric Oncology, King Hussein Cancer Center, Amman, Jordan
8. 8Division of Pediatric Hematology-Oncology, Cerrahpaşa Medical Faculty and Oncology Institute, Istanbul University, Istanbul, Turkey
9. 9Department of Surgery, Memorial Sloan-Kettering Cancer Center, New York, New York, USA
10. 10Department of Hemato-oncology, Hospital JP Garrahan, Buenos Aires, Argentina
1. Correspondence to Guillermo L Chantada, Hemato-Oncology Department, Hospital JP Garrahan, Combate de los Pozos 1881, C1245 AAL, Buenos Aires, Argentina; gchantada{at}yahoo.com

## Abstract

Background The survival of retinoblastoma in less-developed countries (LDCs) and the impact of socioeconomic variables on survival are not widely available in the literature.

Methods A systematic review of publications from LDCs was performed. Articles were from multiple databases and written in seven languages. Results were correlated with socioeconomic indicators. Lower-income countries (LICs) and middle-income countries (MICs) were included in our analyses.

Results An analysis of 164 publications including 14 800 patients from 48 LDCs was performed. Twenty-six per cent of the papers were written in languages other than English. Estimated survival in LICs was 40% (range, 23–70%); in lower MICs, 77% (range, 60–92%) and in upper MICs, 79% (range, 54–93%; p=0.001).Significant differences were also found in the occurrence of metastasis: in LICs, 32% (range, 12–45); in lower MICs, 12% (range, 3–31) and in upper MICs, 9.5% (range, 3–24; p=0.04). On multivariate analysis, physician density and human development index were significantly associated with survival and metastasis. Maternal mortality rate and per capita health expenditure were significantly associated with treatment refusal.

Conclusions Important information from LDCs is not always available in English or in major databases. Indicators of socioeconomic development and maternal and infant health were related with outcome.

• Retinoblastoma
• survival
• metastasis
• risk factors
• early diagnosis
• treatment refusal
• retina
• public health
• neoplasia
• child health (paediatrics)

## Statistics from Altmetric.com

For the past decades, the survival of retinoblastoma has been >80–90% in developed countries.1 As opposed to most other malignancies, these excellent results may be more related to early diagnosis than to the availability of sophisticated treatments. Despite 90% of children affected with retinoblastoma who live in less-developed countries (LDCs), the published literature from this region has received little attention.2 The reasons for this are manyfold. A publication bias against retrospective studies showing negative results or from LDCs has been reported for other conditions.3 In the case of retinoblastoma, the problem is further compounded by the lack of a prospective approach in many treatment centres as well as lack of a staging system for extraocular disease.4 Therefore, we speculated that relevant information about retinoblastoma outcome in LDCs may be available in languages other than English and published only in regional journals or as abstracts for specialised meetings.

Therefore, the aims of this study were to perform a systematic review of retinoblastoma literature from LDCs, carry out an analysis of the retrieved data and describe the outcome of those patients. Early diagnosis of retinoblastoma is influenced by socioeconomic and maternal educational factors,5 so we speculated that indicators of childhood and maternal health could affect the outcome of retinoblastoma in LDCs. Thus, for this study, we also sought to identify the socioeconomic factors that potentially correlate with outcome.

## Methods

### Search strategy

We searched the PubMed database at http://www.ncbi.nlm.nih.gov/pubmed/using the terms “Retinoblastoma AND (Humans(Mesh)) AND ((infant(MeSH) OR child(MeSH) OR adolescent(MeSH))))” and retrieved publications from January 1998 to December 2008. Countries classified as low-income (LICs; gross national income (GNI) <US$755), and middle-income (MICs); which are further subdivided into upper-MICs (GNI US$2996–9265) and lower-MICs (GNI US\$756–2995), per the World Bank (http://web.worldbank.org/WBSITE/EXTERNAL/DATASTATISTICS/0,contentMDK:20420458∼menuPK:64133156∼pagePK:64133150∼piPK:64133175∼theSitePK:239419,00.html) report of economies for the year 2000 were eligible. The abstracts of all publications from the eligible countries were screened and analysed by two authors (SC and GC).

Papers dealing with clinical features, treatment results, epidemiological studies and prevention or early diagnosis of retinoblastoma were eligible for this study, and the entire paper was analysed. Additional data were retrieved using the same strategy and time intervals at the following electronic data bases: EMBASE (http://www.embase.com), Literatura Latinoamericana y del Caribe en Ciencias de la Salud (http://bases.bireme.br/cgi-bin/wxislind.exe/iah/online/?IsisScript=iah/iah.xis&base=LILACS&lang=i&form=F), Scientific Electronic Library Online (http://www.scielo.org), Index Medicus for the Eastern Mediterranean Region and Eastern Mediterranean Region Office (http://www.emro.who.int). Meeting proceedings from the International Society of Pediatric Oncology, American Society of Clinical Oncology, Association for Research in Vision and Ophthalmology, International Congress of Ocular Oncology, American Academy of Ophthalmology and International Society of Genetic Eye Disease were searched. When electronic versions of meeting proceedings were not available, abstract books were hand searched. Additional sources for grey literature (unpublished studies, with limited distribution) included http://www.scholar.google.com and http://www.cure4kids.org. At the former, the term “Retinoblastoma” and each of the eligible countries were used for retrieval of papers.

In addition, each co-author obtained abstracts presented at local or regional meetings from their region. Attempts to further increase the number of retrieved publications from journals only available locally or regionally included search of the Web sites of publications about paediatrics, cancer, ophthalmology and pathology from the different regions. All of the references quoted in the retrieved manuscripts were also searched for additional publications. Finally, we contacted by email some investigators of the retrieved publications and those listed in the International Retinoblastoma Staging Group hosted at http://www.cure4kids.org, librarians and leaders of local, regional or global organisations providing treatment for retinoblastoma and asked them to provide material.

### Data analysis

For each selected paper, we retrieved the following information to complete our database.

• Publication characteristics: (full paper, abstract, monograph, protocol, multimedia presentation and language of the publication)

• Study characteristics: patient number, design (prospective, retrospective), length of follow-up, survival (and methodology of calculation), stage and staging system used.

• Treatment information: chemotherapy used, deaths caused by treatment-related toxicity, percentage of patients who abandoned or refused treatment.

After the analysis of each paper, the data were introduced into a custom-designed database in English including the information detailed above that was available to all of the authors. Publications in languages other than English were read and analysed by local co-authors knowledgeable in the language used at each paper, and the results were included in the abovementioned database in English. All data were checked by two authors (SC and GC) to detect repetitions before analysis. If duplicate information was detected, only the source with the more complete dataset was included. When more than one value was available for a given parameter, the final result was calculated as the mean, combining results from different series. Once the analysis for each region was undertaken, the senior authors (GC, DHA) discussed the data with the local representative to estimate the accuracy and consistency of the data before the final statistical analyses were done.

### Survival estimation

For each country, we calculated a single estimate of survival. Survival data were reported in a non-systematic manner across publications. For example, different durations of follow-up were included, and on occasion, results were not analysed by the Kaplan–Meier method but were reported as survival proportions. Therefore, it was not possible to pass strict statistical methodologic criteria for many papers. When survival data from a given country were available from more than one source, we analysed all of the information from the different centres. When more than one paper from a single centre or group was available, we selected the source with the most complete dataset for analysis. Thus, for each country, we calculated an estimate of survival that was derived from the mean of survival data from independent publications. This result was analysed by each author from that region and matched with the percentage of patients with metastatic disease to be considered accurate.

### Definitions

The following socioeconomic and health-related indicators were obtained from the WHO (http://www.who.int/whosis/en/index.html) report for 2000, except for maternal mortality rate and fertility rate, which were derived from 2005 data:

• Demographic and socioeconomic: Adult literacy rate and gross domestic product

• Health service coverage: 1-year-old children immunised with three doses of DPT vaccine

• Health system resources: Per capita government and total expenditure on health, physician density and nursing and midwives personnel density

• Maternal–child mortality and burden of disease: Infant mortality/maternal mortality rate and <5 mortality rate.

We also obtained the Human Development Index (HDI) from the UN development program (http://hdrstats.undp.org/en/indicators/74.html) as an indicator of global country development. El Salvador, Guatemala, Honduras and Nicaragua comprised a cooperative group, and their data are analysed together as Central America. Treatment refusal, treatment abandonment and cases lost to follow-up were analysed together as a single parameter, which was denoted as “treatment refusal”. These variables were correlated with the estimated values for survival, metastatic rate and treatment refusal for each country.

### Statistical analysis

The correlation between calculated survival, the proportion of patients with metastatic disease and treatment refusal with socioeconomic indicators were calculated with the Pearson correlation coefficient (R and R2). Multiple regression models were used to ascertain the correlation between combinations of predictive variables and survival, the presence of metastatic disease and the occurrence of treatment refusal. χ2 or Fisher exact tests were used for categorical variables. The Mann–Whitney test was used for continuous variables. p Values ≤0.05 were considered significant.

## Results

### Data retrieval

A total of 456 (235 at PubMed and 221 from other sources) references from the countries under study were initially screened. Of those, after selecting those eligible and deleting duplicated information, we analysed 164 references from 48 countries describing 14 800 patients that fulfilled our criteria. Forty-five (27.4%) papers were identified after searching PubMed (including four abstracts initially retrieved by our search were subsequently published as full papers), and 119 (72.6%) were retrieved from other sources. A total of 118 references (73.5%) were written in English, and 46 (26.5%) were written in other languages. Thirty-two papers (19.5%) included patients treated before 1990, and the remaining ones include only children treated after that time.

### Estimated survival

Estimated survival, percentage of metastatic patients at diagnosis and treatment refusal varied among countries (tables 1 and 2) and significantly among the different countries groups (figure 1). Estimated survival was 40% (range, 23–70%) in LICs, 77% (range, 60–92%) in lower MICs and 79% (range, 54–93%) in upper MICs (p=0.001). The percentage of patients presenting with metastatic disease was 32% in LICs (range, 12–50%), 12% in lower MICs (range 3–31%) and 9.5% in upper MICs (range, 3–24%; p=0.04). The percentage of patients that refused treatment was LICs was 35% (range, 30–50%), in lower MICs 11% (range, 2–37%) and in upper MICs was 5% (range, 1–25%; p=0.002).

Table 1

Outcome measures among patients with retinoblastoma in lower-income countries (LICs)

Table 2

Outcome measures among patients with retinoblastoma in middle-income countries

Figure 1

Estimated survival, percentage of metastatic disease at diagnosis and treatment refusal rate according to countries classified according to the World Bank.

### Impact of socioeconomic and health-related variables

Several socioeconomic indicators correlated significantly with survival, the occurrence of metastasis and treatment refusal (table 3). However, after multivariate analysis, only HDI was independently correlated with survival (p=0.0001), and this indicator together with physician density were correlated with the occurrence of metastatic disease (p=0.0001 and 0.006, respectively). Factors that independently correlated with treatment refusal were physician density (p=0.03), per capita total expenditure on health (p=0.002) and maternal mortality rate, which had the strongest correlation (p=0.001).

Table 3

Correlations between socioeconomic indicators and the presence of metastastic disease at diagnosis, estimated survival and treatment refusal in patients with retinoblastoma in less-developed countries

## Discussion

Our study showed that significant information about the outcome of retinoblastoma in LDCs is available in non-English reports, published as meeting abstracts or papers in local journals, confirming that the outcome of retinoblastoma in LDCs is poorer than that in more affluent countries. The inclusion of publications in other languages than English in systematic reviews of randomised clinical trials increases their precision and completeness.6 In addition, the exclusion of grey literature in systematic reviews of randomised interventions may show an exaggerated impact of the intervention effectiveness.7 However, most systematic reviews and meta-analysis are performed to analyse the validity of a randomised treatment intervention. Because there are few prospective studies and no randomised trials for retinoblastoma in LDCs, the quality of the available evidence is sub-optimal. Therefore, as suggested for other diseases with similar features,8 we designed this review including grey literature publications in many languages which helped us to improve the completeness of our data. However, the lack of a commonly used staging system among groups,4 the retrospective design of most studies and the rarity of the disease impeded the use of meta-analysis methodology, and we only report descriptive outcome data under a variety of treatments. So, by analysing our data, we found that among LDCs, the outcome and clinical characteristics of patients with retinoblastoma differ based on the degree of socioeconomic development of the country. We also found that some socioeconomic and health-related indicators correlated with estimated survival. The creation of cooperative international groups, the discussion and publication of standards for staging9 and the update of staging systems, including the experience in LDCs, like the TNM could be helpful towards a better description of the current situation and possibly future clinical trials addressing problems prevalent in LDCs. In addition, the recent creation of national retinoblastoma registries in some LDCs like Malaysia and India as well as global registries will help to give a more reliable description of the situation by providing a population-based estimation of survival.

As reported for other paediatric malignancies, socioeconomic and health-related indicators were significantly correlated with outcome, but some peculiarities have been found.10 A recent study showed that total annual government per capita healthcare expenditure was the strongest predictor of outcome of children with cancer in LDCs.10 Most paediatric malignancies require a rather sophisticated facility to be successfully treated, and such facilities require a high healthcare expenditure. However, indicators of mortality and burden of disease played a more prominent role in predicting survival in our study. Specifically, we found that indicators of societal development, especially those related to the health status of the mother and child, correlated significantly with survival and the occurrence of metastatic disease at diagnosis. The annual per capita healthcare expenditure was correlated with the occurrence of metastatic disease at diagnosis and with survival, but significance was lost in the multivariate analysis. Retinoblastoma can be cured by enucleation of the affected eye(s), a simple surgical procedure if the diagnosis is timely and the families accept it. We speculate that factors related to maternal health and societal development substantially influence the likelihood of early diagnosis before the onset of metastatic disease and, hence, outcome regardless of the facilities available. However, annual per capita healthcare expenditure was a stronger predictor of treatment refusal. Treatment refusal is a common problem, occurring in up to 50% in our study.11 This factor may be related to the type of cancer, and until recently, it has not received much attention in the retinoblastoma literature.11 12 Thus, treating institutions in LDCs should consider developing programs to counteract this problem.12–14

Our data showed that the situation in most LICs differs from that in MICs. In the former, patients tend to present with metastatic disease; families abandon therapy more frequently and survival is in the range from 20% to 30%. In MICs, survival is better, and significantly fewer patients present with metastatic disease. The use of prospective treatment protocols has improved the survival of patients with this condition in MICs.15 16 However, some MICs like South Africa and Malaysia showed lower than expected survival figures.17 18 Despite being relatively affluent, these countries show a lower per capita healthcare expenditure and lower physician and nurse density. Also, contrarily to other countries with similar development scores, no national retinoblastoma strategy was developed until recently in Malaysia, and treatment refusal plays a significant role there.17 Also, inequity in the access to treatment might be important.19 20 On the other hand, countries with relatively lower income figures such as Cuba or Costa Rica show better results. In this setting, a better organised and egalitarian healthcare system compared with lower gross national income figures in a small and homogeneous country, even under different political regimes, could be responsible for the good results.21 22

Despite collecting the best-available published evidence about retinoblastoma in LDCs, our data still have limitations. Most studies were retrospective, and no randomised trial was included. Estimating survival was a challenge because of the inconsistencies we found in reporting follow-up times and the various methods used to estimate survival. However, the main limitation of our study is the possibility of having been missed significant publications, especially from the African continent, where our access to regional reports was more limited. Therefore, for this study, we considered data from all possible sources, despite the quality of survival reports, which was not optimal in many cases. It is also important to consider that about 20% of the papers included patients treated before 1990, so the current results might be better in some settings. In addition, some countries moved from one country category to another during the study period.

Early diagnosis campaigns directed at the general population have been carried out in some countries in order to improve the situation.2 14 The important role of maternal and child health indicators in the survival of patients with retinoblastoma justifies developing campaigns targeted to the public to increase awareness. In all cases, a cost-effective initiative would be to reduce the treatment refusal rate and improved screening of familial cases, especially in MICs.12 23 Twinning programs between institutions or organisations in developed countries and centres in LDCs; the creation of centres of excellence in such settings has been successfully implemented, and they are also an alternative for improving results.24

## Acknowledgments

We are indebted to the following colleagues who contributed their data to this study: Drs Alegria Totah, Imelda Pifano, Gary Mercado, Carlos Leal-Leal, Evandro Lucena, Rita Sitorus, Julia Palma, Sandra Luna Fineman, Ofelia Cruz, Clara Perez, Doris Calle, Abby White, Trijn Israels, Juan Luis Garcia, Maja Beck-Popovic, Luis Castillo, La-Ongsri Atchananeeyasakul, Alp Ozkan, Najeeb Al Shourbaji, Mr Hatem Nour and Ms Aida Farha. We are also indebted to Dr Raul Ribeiro for his critical review of this manuscript and Mrs Angela Mc Arthur for editorial review.