Although the incidence of polio acute flaccid paralysis (AFP) is coming down in India, the non-polio AFP (NPAFP) rate has increased. Nationwide, the NPAFP rate is 13.7/100,000 where the expected rate is 1-2/100,000. We examined the correlates of NPAFP, to discern explanations for the increase.
National Polio Surveillance data 2000-2012 was used. Differences between states and changes over time were examined. Demographic factors and polio programme parameters were assessed. Multiple linear regression analysis adjusting for region/state, literacy rate, female literacy rate, population density and per-capita GDP was performed.
NPAFP increased with the OPV doses used. (R2=32.1%;P<0.001). Correlation was best when cumulative dose received over the previous 4 years was considered (R2=62.5). Per capita income of the state, female literacy and overall literacy showed negative correlation with NPAFP. This disappeared in a multivariable analysis when the number of doses of OPV was considered. On multiple regression analysis, the number of OPV doses was the only factor that showed a positive correlation with the NPAFP rate. NPAFP in UP and Bihar decreased in 2012 coinciding with a reduction in OPV administered.
Our observation showed positive association between NPAFP and the number of OPV doses. In 1994 the National Academy of Sciences, Washington noted a causal relation between Guillain-Barre syndrome and OPV. We hope our findings will stimulate further work to bring down the NPAFP incidence and if needed, to rationalize and optimize the dose schedule of the OPV.
Key words: Acute flaccid paralysis, surveillance, vaccination, polio
Acute Flaccid Paralysis (AFP) surveillance helps identify poliovirus circulation promptly and also provides certification quality evidence that wild polio transmission is not occurring. To qualify as AFP for polio surveillance, there must be acute onset of focal weakness or paralysis with reduced tone in the absence of other obvious cause (like trauma) in children under 15 years (1). The full list of causes of non-polio AFP were reviewed and listed by Marx and colleagues (2). Transient weakness (postictal paralysis for example) is not included (3). In the absence of wild polio transmission, the WHO estimates that there is a background annual incidence of at least 1 case of AFP per 100,000 children under 15 because of diseases like Guillain–Barré syndrome (GBS). In other words, once polio is eradicated, it is expected that the AFP rate (made up of polio and non-polio AFP) would come down to 1 to 2 per 100,000 as there would only be non polio AFP left (4).
The surveillance performance of India has been excellent, and the incidence of polio AFP is coming down in the country (5). As a consequence the total AFP rate (which includes polio and non-polio causes of AFP) must come down with the non-polio AFP rate remaining steady. Inexplicably however, the non-polio AFP rate has shown a trend to increase. Nationally, the non-polio AFP rate is 13.7 per 100,000 (2012 data) where the acceptable rate is 1 to 2 per 100,000 (6). In the state of Uttar Pradesh (UP) the non-polio AFP rate is 24.6 per 100,000 and in Bihar this is 37 per 100,000 (2012 data) (5). It has been said that the increase in AFP in recent years is the result of a deliberate effort (that began in 2004) to intensify surveillance and reporting in India (7). However, a surveillance program no matter how good, can only record every case of AFP, but it cannot exaggerate the numbers or explain the 20 to 40 fold increase in the non-polio AFP rate, as seen in the state of Bihar (assuming that the natural non-polio AFP rate should lie between 1 to 2 per 100,000 as per internationally accepted norms) (6). The lowest non-polio AFP rates are seen in the states were polio was eliminated earliest. These are the states which must have had the best implementation of the polio control programme and the best surveillance. If good surveillance was the reason for the increase in non-polio AFP, it is paradoxical that the well-performing states should have the lowest non-polio AFP rates.
Follow-up of these cases of non-polio AFP is not done routinely. However a fifth of these cases of non-polio AFP in the state of Uttar Pradesh were followed-up after 60 days, in 2005. 35.2% were found to have residual paralysis and 8.5% had died (total residual paralysis or death 43.7%). (8) This suggests that the pathology in children being registered as non-polio AFP cannot be considered as trivial. There is thus impelling reason to try and understand the underlying causes for the surge in non-polio paralysis numbers.
The AFP rates are different in the various states and Union Territories of India. In this study we examined the factors that correlate with the non polio AFP rates in the states, to discern possible explanations for the increasing incidence of non-polio AFP. Data on AFP rate from the National Polio Surveillance Programme (NPSP) in India over the period 2000 – 2012 was used. Differences between states and changes over time with respect to non-polio AFP were examined. Association with demographic factors and polio programme parameters were assessed.
States reporting more wild polio cases are targeted for increased vaccine coverage. For example it was reported that during the year 2005, children under 5 years in UP and Bihar received on average 15 doses of trivalent oral polio vaccine (tOPV), compared with 10 doses in the rest of India (9). Confounding is a possibility. On cursory examination of the NPSP data it appears that the states with the highest polio rates are also the ones with higher non-polio AFP rates. It is conceivable that the factors that result in higher polio incidence in these states, also promote non-polio AFP. Grassly and colleagues suggest that high population densities and poor sanitation explain the persistence of polio (9). We aimed to see whether various factors like population density, literacy and poverty could explain the non-polio AFP rate.
Material and Methods
The data on AFP, polio and non-polio AFP and number of polio rounds were examined in each state in each year from 2000 to 2012. Data from the National Polio Surveillance web site (5) was used. The raw data, as extracted from the web site, has been uploaded as supplementary file (and also available at http://bit.ly/npsi_data). Data on numbers of polio rounds during the years 2003 and 2012 was incomplete and so figures for these years were not included in the general analysis. However data from 2012 on polio rounds for some states were available and they were used in analysis specific for those states. When different areas within a state received different numbers of doses, the arithmetic mean of doses was taken as the representative dose for that state. Normal linear regression analysis was carried out, taking the non-polio AFP rate as the outcome variable and the number of OPV doses as the explanatory variable. Both these were treated as continuous variables. Multiple linear regression analysis was carried out to adjust for region/state, literacy rate, female literacy rate and per capita GDP. Pearson’s correlation coefficient of non-polio AFP rate with the per capita income (10) and population density (11) of each state and union territory were looked into. According to Rosser and colleagues, sanitation is closely linked to female literacy in a range of Asian countries. (12) We could not find authentic data on sanitation in the different states and hence we used female literacy as a surrogate for social development and also of the general level of sanitation and hygiene in each state. Data on female literacy and overall literacy was obtained from the Census of India (13). We also included overall literacy because some communities tend not send their girl children to school and we expected overall literacy may have a positive effect on the evolving social developments of the area, which may not be reflected entirely and accurately by female literacy in that area.
We explored whether any correlation exists between the non-polio AFP rates and the number of doses of polio vaccine used in the state in that year. Further, we tried to look for cumulative effect by adding data on vaccine doses from previous years.
The non-polio AFP rate per 100,000 increased with increased number of OPV doses during 2000-2011, irrespective of time and region. Figure 1 shows the trend of non-polio AFP with doses of OPV. As can be seen, the relationship is curvilinear with a more steep increase in non-polio AFP beyond 6 doses of OPV. To demonstrate this association, the Pearson correlation was calculated. This association is statistically highly significant (R2 = 32.1%; P < 0.001).
Looking at data up to 2012 from Uttar Pradesh and Bihar, where the maximum doses of OPV were used, the R2 was 51.9% (P < 0.001). Regression analysis indicated that for an increase of one dose of OPV, the non-polio AFP rate increased on the average by 3.7 per 100,000 populations under 15 (95% CI: 2.1 - 5.3). The relation is shown in Figure 2. To examine the trends over time we looked at the relationship of the non-polio AFP rate each year to the number of doses of OPV received in that year separately for UP and Bihar. Figure 3 shows the non-polio AFP over the years in the states of UP and Bihar alongside the 4 year cumulative OPV doses. The fall in NPAFP rate for the first time in 2012 with a decrease in the OPV doses is noted.
When the effect of cumulative doses over the previous years was examined, the non-polio AFP rate in 2011 best correlated to the cumulative doses received in the previous 4 years. Association (R2) of non-polio AFP rate with OPV doses received in 2011 was 52.6%. It increased to 56.9.6%, when we looked at AFP rate in 2011 related to total doses received in the years 2010 and 2011. Adding up doses received in 2009 to that in 2010 and 2011 further increased the association (R2 = 54.4%) and adding up doses received between 2009 and 2011 the regression coefficient rose to 62.5. All these correlations were highly significant (P < 0.001). The association showed no further improvement with addition in OPV doses from earlier years.
Per capita income of the state, female literacy and overall literacy showed negative correlation with the non-polio AFP (R2 = 4.1%, P < 0.001; R2 = 13.0%, P < 0.001; and R2 = 12, P < 0.001 respectively). Population density did not show any association with the non-polio AFP (R2 = 0.0%, P = 0.91).
A multivariable model to adjust simultaneously the influence of factors like overall literacy, female literacy and per capita GDP, confirmed the significant positive association between the number of OPV doses and non-polio AFP rate. Per capita income, female literacy and overall literacy did not show significant association with non-polio AFP rate once we considered the number of doses of OPV. After adjusting for these factors, the average increase in the non-polio AFP rate was 1.30 per 100,000 (P < 0.001, 95% CI: 1.09-1.51) with each dose of OPV. The results of this multiple regression analysis are presented in Table 1.
Our results indicate that the incidence of non-polio AFP was strongly associated with the number of OPV doses delivered to the area. We also observed a dose response relation with cumulative doses over the years, which further strengthen the hypothetical relationship between polio vaccine and non-polio AFP. The cumulative dose received in the previous 4 years related best to the non-polio AFP rate in 2011. Children over 5 are not vaccinated and the vaccine naïve newborns are added to the pool each year and this is perhaps the reason cumulative doses beyond 4 years did not improve the strength of association. The fall in the NPAFP rate in Bihar and UP for the first time in 2012, with a decrease in the number of OPV doses delivered, is further corroborative evidence of a causative association between OPV doses and the NPAFP rate
A regression analysis reveals an association but does not prove a causal role. Ecological fallacies must be borne in mind meaning that correlation of aggregate variables take into account cross sectional effects which are not relevant at the individual level. The likelihood of confounding also needs to be kept in mind. Poor sanitation can cause spread of entero-pathogens (like Campylobacter jejuni) which cause non polio flaccid paralysis. The same sanitation problem can result in increased polio in that area and this will trigger more frequent polio immunization rounds in the locality. Thus, a spurious relationship between OPV doses and non-polio flaccid paralysis may be inferred when poor sanitation is the real culprit - the confounding variable – causing the non-polio AFP.
Data from the WHO / UNICEF Joint Monitoring Programme (JMP) for Water
Supply and Sanitation, suggests that states of Delhi and Kerala, with a female literacy rate of 75% and 88% respectively have achieved millennium development goals (MDGs) on sanitation already, while Assam with 64% female literacy and Arunachal with 55% female literacy will achieve it in 10 years, and Madhya Pradesh and Orissa with 50% and 51% female literacy respectively, will achieve the MDG only in the next century. (14) This further supports the observation that sanitation is linked to female literacy in many Asian countries. (12) In the absence of hard data on the number of households with toilets for safe disposal of excreta and free running water, we tried to remove the confounding effect of poor sanitation, by using female literacy as a surrogate for social development in the state and its level of sanitation and hygiene. We did not find an association of non-polio AFP rates with female literacy in the multivariable analysis. Admittedly although female literacy is often used as a surrogate for social development, it need not be an adequate surrogate for sanitation in an area. However poor sanitation by itself cannot explain why the incidence of non-polio AFP should increase year to year in the same area, in proportion to doses of polio vaccine administered here, unless sanitation in the area is deteriorating each year, coinciding with increased vaccine doses administered.
Another possible explanation for the apparent association could be increased surveillance of AFP in areas where the polio related AFP is high, and this may be detecting more cases of NPAFP. Though the surveillance has improved over time, there is no evidence to suggest that there are differences in the quality of surveillance between states in the same year. The difference in non-polio AFP rate between states in any particular year cannot therefore be explained on the basis of improved surveillance alone. In fact it is the states with the best health indicators and therefore presumably the best surveillance, (for example Goa and Kerala) that have some of the lowest non-polio AFP rates. The case definition of AFP has also changed over the years but the definition at any point of time has been uniform all over the country and so this cannot explain the differences in the rates of non polio AFP seen in different states in the same year.
Our findings point to the need for a critical appraisal to find the factors contributing to the increase in non-polio AFP with increase in OPV doses – perhaps looking at the influence of strain shifts of enteropathogens induced by the vaccine given repatedly. The clear dose response relationship indicates that this relationship in not a spurious one. In 1994 the Institute of Medicine of the National Academy of Sciences, Washington had noted that the evidence available to them favors acceptance of a causal relation between Guillain-Barre syndrome (a major component of the non polio AFP rate) and OPV. (15) All these factors need to be examined.
Further studies perhaps including a rapid epidemiological appraisal exploring the hypothesis of association between number of doses of OPV and NPAFP by case control studies, individual children with NPAFP being the cases and matched healthy children as the controls are called for.
Our observation showed positive association between the non-polio AFP rate and the number of OPV doses. Though there is a possibility that the apparent association between the increasing trend of non-polio AFP and OPV doses could be a statistical artifact due to confounding factors, the magnitude of non-polio AFP incidence and its increasing trend are public health problems in themselves and need to be looked into. We hope our findings will stimulate further work to bring down the non-polio AFP incidence and if needed, may also be used to rationalize and optimize the dose schedule of the OPV.
1. Ahmad A Rehman A. One year surveillance data of acute flaccid paralysis at Bahwal Victoria Hospital Bahawalpur Pakistan Journal of Medical Sciences 2007;23:308-12.
2. Marx et al., Epidemiologic Reviews, 2000, Vol 22(2): 298-316.
3. Canadian Pediatric Society. Surveillance Canadian Pediatric Surveillance Programme Acute Flaccid Paralysis. http://www.cps.ca/english/Surveillance/CPSP/Studies/acute.htm
4. Alcala H. The differential diagnosis of poliomyelitis and other flaccid paralysis. Bio Med Infant Mex 1993; 50: 136-144.
5. http://www.npspindia.org/bulletin.pdf accessed on 19/3/11
7. Grassly NC, Wenger J, Bahl S, Sutter RW, Aylward RB. Protective efficacy of a monovalent oral type 1 poliovirus vaccine. Lancet 2007; 370: 129-130.
8. Puliyel JM, Gupta MA, Mathew JL. Polio Eradication & the future for other programmes: Situation analysis for strategic planning in India. Indian J Med Res 2007; 125: 1-4.
9. Grassly NC, Fraser C, Wenger J, Deshpande JM, Sutter RW, Heymann DL, Aylward RB. New Strategies for the Elimination of Polio from India. Science 2006; 314: 1150 – 1153.
10. Gupta S. States Performance in Per capita Income Growth ASSOCHAM ECO PULSE January 2008 www.assocham.org/arb/aep/states-per-capita-income.doc Accessed on 28/2/2011
11. Census of India. "Area and Population". Government of India (2001). http://www.censusindia.gov.in/Census_And_You/area_and_population.aspx. Quoted in Wikipedia: List of states and union territories of India by population. http://en.wikipedia.org/wiki/List_of_states_and_union_territories_of_India_by_population
12. Rosser JB, Rosser MV. Comparative Economics in a Transforming World Economy, 2nd Edition Massachusetts MIT Press ISBN 0-262-18234-3 (2004) Page 476
13. Census of India 2001-- State wise population totals
14. WHO / UNICEF Joint Monitoring Programme (JMP) for Water
Supply and Sanitation. http://www.wssinfo.org/data-estimates/introduction/ Accessed on 15/5/12
15. Stratton KR, Howe CJ, Johnston RB. Adverse events associated with childhood vaccines other than pertussis and rubella. Summary of a report from the Institute of Medicine. JAMA. 1994;271:1602-5.