Khoj: Eradicating Tuberculosis and Redefining Epidemic Management

Thank you for your question. The short answer is that we will use a two-pronged approach to differentiate latent TB from active TB - a clinical diagnosis and a confirmatory test after a few days - this would be the same test we first used to diagnose TB.

We will distinguish between subjects at high and low risk for at over-diagnosis by looking at both the bacterial load (which will be higher for those at low risk of over-diagnosis) and clinical parameters (symptoms, history of contact with active TB).

To elaborate further- we are gearing our solution to have maximum sensitivity for diagnosis using advanced imaging techniques and machine learning. TB in India is a notifiable disease, which implies that any diagnosed cases have to be reported to the government after which strict/rigid clinical protocols are followed by the government. Current detection technologies are limited in detecting high bacterial loads and since testing is limited it is usually deployed after clinical diagnosis to detect cases. As a consequence, almost all the cases detected are of active TB, with the majority of cases detected having very high bacterial loads. The clinical protocols defined for these cases are rigorous and unnecessary to follow for latent TB.

Since our solution promises higher accuracy and much wider deployment than traditional technologies, we expect that we will detect both active and latent TB cases. Since the clinical protocol and follow-up is defined for active TB but not for latent TB, we would like to add granularity to the reporting of cases, when reporting to the administration.

We would therefore take steps to avoid reporting latent TB till granularity is added to the treatment and response side. To distinguish between latent TB and active TB we will use confirmatory clinical diagnosis based on both histories of contact with active TB and clinical exhibition of TB in diagnosed cases.

Over time we hope to be able to define threshold ranges of bacterial load and patterns around it by monitoring changes in bacterial loads over time using additional tests to be able to distinguish latent and early TB with late-stage TB without clinical intervention.

Our team was largely involved in the first wave as we aided made projections and aided internal efforts in the logistics side. We also educated policy makers with Covid basics and evidence based interventions.

However we were not involved in the projections in the second wave until April, this is because there was not enough serological data, or rather enough high quality truly randomised serological data for COVID to be able to understand the susceptible population parameter. This inhibited any serious model making.

Despite not being involved except for the last month in the effort for the second wave, the factors behind the wave and this humanitarian disaster are abundantly clear.

The obvious ones are-

1. Covid innappropriate behaviour and poor policy making- after the December wave most epidemiologists (including us) and the government let down their guard. This was informed by poorly conducted serological surveys which showed upto 70% population infected in some places. On careful analysis we found that these surveys were not truly randomised with biases towards areas with high covid incidence. Certain large gatherings (involving millions of people) along with the elections in at least 7-10 regions in India also caused this massive spike.

2. Wanning Immunity- add to these factors, waning immunity. Antibody prevalence in the population that blocked the first wave from growing any further reduced over the course of the last 6-8 months. This meant more people were transmissive than at the end of the first wave.

3. Mutations- as is the case with any virus, the virus mutated to multiple variants. Multiple significant mutations have been recorded and with a population, landscape, and variation of strains afforded by India, these mutations are all ever so slightly different from each other, populating different strains in silos which mixed. We suspect that even though most of these mutations will not make the virus deadlier or more transmissive, they will provide for an immune escape from an epidemiological perspective. Antibodies tuned to protect from the earlier strains of the virus will not provide the same level of protection (as is also seen in the case of vaccine-generated antibodies and mutations related to immune escape).

4. Low testing, near-zero randomized prevalence monitoring- In the first wave some states upon our recommendation implemented randomised prevalence and randomised sentinel monitoring (sentinels being high contact individuals), this let them get ahead of spread well in time. However in the second wave this was not done as one tended to believe that we ld be able to face any peak that would arise out of waves subsequent to the first having scaled up medical facilities significantly.

5. Reluctance to lockdown even at a local level- The central body (this time not informed by our projections) in charge of the projections was fairly confident that we would peak at 100,000 cases per day in their worst-case scenario back in March when the wave was picking up. This rang false and their projections turned out to be too low. While on the other hand in the government, there were skeptics who didnt even think we would reach 100K cases per day. While a national lockdown didnt make sense at any point (even in retrospect) as it would cause more damage to life and livelihood like last time around, localized restrictions could have been enforced. However due to elections and economic concerns governments refused to enforce this. As a classic example in the first wave before mumbai even had 100 cases a day, the city was completely shuttered. In the second wave, Mumbai local trains- the densest mode of transport-were running even when Mumbai had 2000 cases per day.

A combination of these factors lead India to the second wave. To be honest, the first wave was more of an anomaly as we expected COVID in a population of 1.3 billion with the densest cities in the world to be extremely devastating. However, the early lockdown, the low variance of strains, high contact tracing, the low spread of the virus to rural areas worked in our favor. In the second wave complacence led to a disaster.

Universal randomised testing, sentinel surveillance, large swathes of data such as the ones our technology will enable not only in TB but in all emerging infectious diseases can stop in tracks any epidemic, converting what would look like the second wave in India to what was the first wave at the very least.

I hope we get a chance to elaborate more in the next round as a comment here in reply would not do the question justice.