key: cord-0799638-3xup6glg
authors: Giamarellos-Bourboulis, Evangelos J.; Tsilika, Maria; Moorlag, Simone; Antonakos, Nikolaos; Kotsaki, Antigone; Domínguez-Andrés, Jorge; Kyriazopoulou, Evdoxia; Gkavogianni, Theologia; Adami, Maria-Evangelia; Damoraki, Georgia; Koufargyris, Panagiotis; Karageorgos, Athanassios; Bolanou, Amalia; Koenen, Hans; van Crevel, Reinout; Droggiti, Dionyssia-Irene; Renieris, George; Papadopoulos, Antonios; Netea, Mihai G.
title: ACTIVATE: RANDOMIZED CLINICAL TRIAL OF BCG VACCINATION AGAINST INFECTION IN THE ELDERLY
date: 2020-09-01
journal: Cell
DOI: 10.1016/j.cell.2020.08.051
sha: 4f843a4f7590f70caa34be52035457332a725c17
doc_id: 799638
cord_uid: 3xup6glg

BCG vaccination in children protects against heterologous infections and improves survival independently of tuberculosis prevention. The phase III ACTIVATE trial assessed whether BCG has similar effects in the elderly. In this double-blind, randomized trial, elderly patients (n=198) received BCG or placebo vaccine at hospital discharge, and were followed for 12 months for new infections. At interim analysis, BCG vaccination significantly increased the time to first infection (median 16 weeks compared to 11 weeks after placebo). The incidence of new infections was 42.3% (95% CIs 31.9-53.4%) after placebo vaccination and 25.0% (95% CIs 16.4-36.16%) after BCG vaccination; most of the protection was against respiratory tract infections of probable viral origin (hazard ratio 0.21, p: 0.013). No difference in the frequency of adverse effects was found. Data show that BCG vaccination is safe and can protect the elderly against infections. Larger studies are needed to assess protection against respiratory infections, including COVID-19.

Infection by the novel SARS-CoV-2 virus (also termed COVID-19) has a severe impact on both the health of the populations around the globe, and on the world economy. Many countries are in lockdown, with a third of the world population in some form of movement restrictions, which brings serious financial and societal consequences. The urgent need for the reversal of this situation can only be met through the generation of an immune defense shield to protect the populations from SARS-CoV-2 infection. Many efforts for the development of a vaccine are under way, but it is likely that at least 12 to 24 months will be needed till an effective vaccine could be available.

Interestingly however, trained immunity induced by some already available vaccines such as Bacille Calmette Guérin (BCG), oral polio vaccine (OPV), or measles vaccine have been suggested to be used as a potential protective approach against COVID-19 to bridge the period until a specific vaccine is developed (Netea et al., 2020) . Trained immunity is the process of epigenetic, transcriptional and functional reprogramming of innate immune cells (such as myeloid cells or NK cells), leading to an increase in the cytokine production capacity and their antimicrobial function (Kleinnijenhuis et al., 2012; Netea et al., 2016) . In models of experimental human infections such as yellow fever vaccine virus (Arts et al., 2018) or human experimental malaria (Walk et al., 2019) , BCG vaccination was able to induce a nonspecific protection.

These experimental data are accompanied by epidemiological studies in children and adults showing non-specific protection against infections and mortality by BCG vaccination. BCG vaccination reduced the incidence of respiratory syncytial virus infection in children in Africa (Stenballe et al., 2005) and protected elderly J o u r n a l P r e -p r o o f against respiratory tract infections in Indonesia (Wardhana et al. 2011) and Japan (Ohrui et al. 2005) . Finally, the concept was also successfully tested in healthy volunteers that were vaccinated with placebo or BCG vaccine, and 14 days later received a tri-valent influenza A vaccine. Volunteers previous vaccinated by BCG developed significantly greater titers against hemagglutinin A of the influenza A virus whereas their circulating monocytes were more potent for the production of interferon-gamma (Leentjens et al. 2016) .

ACTIVATE (A randomized Clinical trial for enhanced Trained Immune responses through Bacillus Calmette-Guérin VAccination to prevenT infections of the Elderly) is a randomized trial in which hospitalized elderly patients were vaccinated on the day of hospital discharge with single doses of placebo or BCG. Patients were under follow-up for 12 months, with the last visit of the last patient scheduled for August 2020. However, the pressure rising from the need of protection of the elderly who are considered susceptible to infection by SARS-CoV-2 (Guan et al. 2020; Huang et al. 2020) led to an interim analysis of the results of the study. Results of this interim analysis clearly showed protection of the elderly from new infections with major effect on the prevention of respiratory infections. (Table 1 and supplementary Table S1 ).

Regarding the primary endpoint of the study, BCG vaccination significantly increased the time to first infection: median 16 weeks after BCG vaccine compared to 11 weeks after placebo administration. The incidence of a new infection during the 12-month period of follow-up after vaccination was also significantly decreased; the statistically significant hazard ratio (HR) of 0.55 corresponds to 45% reduction in the risk of a new infection in the BCG group compared to the placebo group ( Figure 2A ).

The incidence of new infection was 42.3% (95% confidence intervals [CIs] 31.9-53.4%) in the placebo group and 25.0% (95% CIs 16.4-36.16%) in the BCG group.

The difference in the incidence according to the type of infection showed most of the benefit on the prevention of respiratory infections of probable viral origin ( Figure 2B ); the hazard ratio in this case was 0.21 (95% CI 0.06-0.72) corresponding to 79% decrease in the risk for the BCG group in comparison to the placebo group. An analytical presentation of the efficacy of BCG vaccination for all primary and secondary study outcomes is shown in Table 2 .

Sensitivity analysis was done for the total of 198 patients taking into consideration that the time of 12-month follow-up has not been completed for 48 patients (Figure 2C and supplementary Table S2 ). This sensitivity analysis confirmed the results of the primary outcome presented in Figure 2A . The confirmation of the primary endpoint of the interim analysis by the sensitivity analysis establishes the absence of any violations on the time-to-event analyses since individuals that are J o u r n a l P r e -p r o o f censored have the same probability of experiencing a subsequent event as individuals that remain in the study. The proportionality of the hazards over the total time period of follow-up was validated by plotting the Schoenfeld residuals (supplementary Figure S1 ).

Stepwise Cox regression analysis showed that BCG vaccination was an independent protective factor from the incidence of new infection until month 12 (HR 0.56; 95% CI, 0.32 to 0.99; p: 0.048) ( Table 3) .

Major benefit from BCG vaccination was observed in the main secondary endpoint patient-infections per year. This was 57.7 per 100 patients in the placebo group and 33.3 per 100 patients in the BCG group (p: 0.003) ( Table 2) . No difference in the other secondary endpoints was found between the two groups (Table 2 and supplementary Figures S2 and S3 ).

In a sub-group of 57 volunteers (31 placebo and 26 BCG vaccinated), we assessed production of innate immune responses at 2 time points (before and 3 months after vaccination) in peripheral blood mononuclear cells (PBMCs).

Heterologous production of tumour necrosis factor-alpha (TNFα), IL-1β and IL-10 (trained immunity induction) (Figures 3A to E), but not of interleukin (IL)-6 (data not shown) by PBMCs after stimulation with non-mycobacterial ligands was amplified among BCG-vaccinated individuals compared to placebo-vaccinated individuals. A trend towards amplified interferon-gamma (IFNγ) (heterologous T-cell responses) responses was also found ( Figure 3F ). Unfortunately, the number of BCG-vaccinated J o u r n a l P r e -p r o o f individuals in which cytokine data are available are too small to permit the prediction of trained immunity responses as correlates of protection.

Various studies have shown that the increased cytokine responses upon BCG vaccination are the result of epigenetic reprogramming of monocytes (Arts et al, 2018; Kleinnijenhuis et al, 2012) . In order to examine potential differences in the epigenetic profile between BCG-vaccinated individuals and controls, we determined at pro-inflammatory genes the level of histone H3 acetylation at lysine 27 (H3K27ac), a mark of active promoters and enhancers. In line with previous findings, we observed increased levels of H3K27ac at the regions of IL-6 and TNFα in BCGvaccinated individuals as compared to individuals that received placebo, suggestive of epigenetic reprogramming upon BCG vaccination ( Figure 3G ).

To further validate the solidity of the observation that BCG induces trained immunity responses in the elderly, we assessed immune responses before BCG vaccination, two weeks and three months after vaccination in fourteen healthy volunteers aged 55 years or older that took part in an independent BCG-vaccination study (300BCG cohort, www.humanfunctionalgenomics.org). All individuals in this cohort were vaccinated with the same BCG strain used in the ACTIVATE trial and PBMCs were isolated and stimulated ex vivo either with Staphylococcus aureus, LPS or Mycobacterium tuberculosis, before and after vaccination to assess the magnitude of the immune memory responses. We observed a significant increase in IFNγ upon stimulation with M. tuberculosis after BCG vaccination ( Figure 3H ), indicative of induction of adaptive immune memory response. In addition, cytokine production also significantly increased in elderly when cells were exposed to non-mycobacterial stimuli such as S. aureus and LPS ( Figure 3I and supplementary Figure S4A ), indicative of induction of trained immunity. Furthermore, we observed long-term J o u r n a l P r e -p r o o f changes in neutrophil phenotype three months upon BCG vaccination as compared to baseline ( Figure 3J ). Together, these findings indicate sustained trained immunity responses in the elderly, and support our previous observation of non-specific beneficial effects against unrelated infections in elderly upon BCG vaccination. In addition, we employed a targeted proteome platform to measure 92 inflammatory markers before and after BCG vaccination, which revealed no significant changes in the concentrations of circulating inflammatory proteins, including IL-6 and IL-18 after BCG (supplementary Figure S4B and supplementary Table S3 ). Similarly, no significant changes in monocyte, granulocyte or lymphocyte count were observed upon vaccination (supplementary Figures S4C to S4E ). This demonstrates that while BCG vaccination induces trained immunity and cell responsiveness, it is not followed by excessive systemic inflammation.

A trend for lower serious adverse events was recorded in the BCG vaccination group than in the placebo group (Table 4) . Moreover, the incidence of non-serious adverse events did not differ between the two groups. None of the adverse events were related to the study intervention. None of the patients developed tuberculosis.

The ACTIVATE study was conducted from 2017 with the aim to assess the potential of BCG vaccination to protect elderly with an increased risk for infection against new infectious episodes. As a target population we have chosen to investigate elderly patients returning home from a hospital admission, as it is known that this population is at a high risk to develop infections (Bender, 2003) . This J o u r n a l P r e -p r o o f approach using BCG vaccination is justified due to the increasing number of experimental and epidemiological studies suggesting that BCG can protect against respiratory infections in general, and viral infections in particular (Moorlag et al., 2019) . Indeed, the data shown here demonstrate that BCG vaccination led to a lower number of infections of all causes, and especially respiratory tract infections, arguing for a protective effect.

Epidemiological data suggest beneficial effects of BCG on all-cause mortality in children in countries with high infectious pressure. This protection has been attributed to lower incidence of neonatal sepsis and respiratory tract infections (Garly et al. 2003) , which in children are often viral as etiological cause. This assumption is also supported by the data indicating protective effects of BCG vaccination against RSV infection (Stensballe et al. 2005) . The protection in children was also complemented more recently by studies showing protective effects of BCG vaccination against respiratory tract infections in adolescents (Nemes et al., 2018) and in elderly individuals (Wardhana et al., 2011) . In line with this, the incidence of infection in the ACTIVATE trial was significantly lower in the elderly individuals vaccinated with BCG, compared to the non-vaccinated volunteers. Moreover, this protection was mainly due to respiratory tract infections of probable viral origin, with hazard ratio 0.21 in the BCG vaccinated group, which is in line with the 70-80% reduction in respiratory tract infections in studies done in Indonesia and Japan (Ohrui et al., 2005; Wardhana et al. 2011 ).

An important aspect that should be mentioned is that this interim analysis was performed earlier than the final visit of the planned study, resulting in 78 patients in the placebo group and 72 patients in the BCG group being able to complete the 12months follow-up. The reason for this interim analysis that has been approved by J o u r n a l P r e -p r o o f National Ethics Committee and by the National Ethics Committee of Greece was the emergence of the COVID-19 pandemic and the initiation of several major studies on the effect of BCG on the infection with SARS-CoV-2. In addition to the effectiveness aspect, another important question that needed to be urgently answered was that of the safety of BCG vaccination is the setting of COVID-19. An exaggerated inflammatory reaction has been described to contribute to severity and mortality in some patients with COVID-19 (Huang et al. 2020 ) raising concerns in the community that BCG vaccination may have deleterious effects due to the enhancement of innate immune responses. Indeed, circulating concentrations of proinflammatory cytokines are increased in severely ill COVID-19 patients (Huang et al., 2020) . On the other hand, it can be also argued that vaccination with BCG leading to activation of antiviral mechanisms will lead to decreased viral loads and thus lower systemic inflammation, with milder disease and quicker recovery. This model is supported by our earlier studies showing that BCG vaccination decreased viral loads and systemic inflammation after yellow fever vaccine administration (Arts et al., 2018) . It is reassuring to observe that this hypothesized protective effect of BCG vaccination is supported by the clinical data in ACTIVATE trial where most of benefit was observed for the incidence of respiratory infection of probable viral origin. Importantly, we observed no increase in the concentrations of pro-inflammatory proteins in the circulation of BCG-vaccinated individuals as compared to concentrations before vaccination, demonstrating that steady-state levels of inflammation are not increased by BCG.

The mechanism of protection induced by BCG vaccination could be either through heterologous T-cell responses (Welsh et al., 2010) , or though induction of trained immunity (Kleinnijenhuis et al., 2012) . Our data showing enhanced cytokine J o u r n a l P r e -p r o o f responses to non-mycobacterial stimuli and epigenetic reprogramming of monocytes in BCG vaccinated individuals point towards the induction of trained immunity, although it is likely that a combination between innate and heterologous T-cell immunity is responsible for the entire clinical effect.

The main limitations of the trial are: a) the relatively small sample size of our cohort that will need additional validation in a larger study; b) the lack of repeat IGRA after vaccination, c) the absence of serological information on the incidence of various respiratory infections; and d) the lack of information on BCG vaccination at birth. It needs, however, to be mentioned that despite the small sample size significant differences were found.

The number of individuals participating in the trial is too low to permit us drawing any conclusions regarding the effect of BCG vaccination on coronaviruses in general, or COVID-19 in particular. For that, either much longer follow-up, or much larger studies, are necessary. Indeed, several clinical trials have been started or are under preparations to test this hypothesis. The majority of these trials are studying the protective effect of BCG in healthcare workers, and a synopsis of these trials is provided in supplementary Table S4 . Serological assessment of the prevalence of antibodies against respiratory coronaviruses is also warranted in future studies.

In conclusion, in the present study we demonstrate that BCG vaccination is safe and decreases the number of infections in an elderly population at risk. While these data need to be interpreted with caution, they support the hypothesis that BCG-induced trained immunity protects against new infections, mostly against respiratory tract infections. Although results argue that BCG vaccination could be used to bridge the period until a specific vaccine for SARS-CoV-2 is developed and J o u r n a l P r e -p r o o f produced, larger randomized clinical trials to study the impact of BCG vaccination on morbidity and mortality due to SARS-CoV-2 infection are needed.

The study was funded in part by the Horizon 2020 grant ImmunoSep (#847422) and in part by the Hellenic Institute for the Study of Sepsis. Participants in the ACTIVATE trial vaccinated with placebo or with BCG Panels A to F report on PBMCs isolated at baseline (month 0) and 3 months after vaccination. PBMCs were stimulated for cytokine production. Blood sampling was not done for all participants after three months either because some individuals had died or because they were hospitalized at other study sites or because of denial for blood sampling.

A) Percentage of patients vaccinated with placebo and BCG with more than 30% increase of the production of TNFα after stimulation with Pam3Cys. J o u r n a l P r e -p r o o f Table 3 for a complete list of circulating markers that were measured. 

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Evangelos J. Giamarellos-Bourboulis (egiamarel@med.uoa.gr)

This study did not generate new unique reagents.

Data of this study are available after communication with the Lead Contact. A material transfer agreement will be needed. On the day of hospital discharge and after careful recording of detailed medical history and laboratory examinations for the inclusion and exclusion criteria, patients remaining eligible underwent IGRA test. Those who had negative IGRA were allowed to be enrolled in the study. Participants were randomized to one intradermal vaccination with 0.1ml of sodium chloride 0.9% or with 0.1ml of BCG (BCG vaccine strain 1331; Intervax). Simple randomization was performed in a 1:1 ratio by a biostatistician and delivered to the investigators as an electronic file for treatment allocation at randomization. Blind administration was secured by two study pharmacists; one who was preparing the vaccine and another who was delivering the preparation to the investigators for vaccination. enlargement of liver or of spleen found on deep abdominal palpation; and absolute lymphocyte count more than 4,000/mm 3 (Antonopoulou et al., 2012) . Bloodstream infection was defined as at least one positive blood culture for a pathogen not related to infection at other site (Calandra & Cohen, 2005) . Community-acquired pneumonia was defined as any new or evolving infiltrate on chest X-ray in a patient without any contact with the hospital environment the last 90 days and who was presenting with at least two of the following: new onset or worsening of cough; body temperature above 38 0 C; dyspnea; purulent tracheobronchial secretions; auscultatory findings compatible with pulmonary consolidation; and procalcitonin ≥0.25 ng/ml or absolute total white blood cell count ≥12,000/mm 3 (Christ-Crain et al., 2006) . Hospital-acquired J o u r n a l P r e -p r o o f pneumonia was defined as any new or evolving infiltrate on chest X-ray in a patient starting at least 48 hours after hospital admission and who was presenting with at least two of the following: new onset or worsening of cough; body temperature above 38 0 C; dyspnea; purulent tracheobronchial secretions; auscultatory findings compatible with pulmonary consolidation; and procalcitonin ≥0.25 ng/ml or absolute total white blood cell count ≥12,000/mm 3 (Kalil et al., 2016) . The urinary tract infection was defined as the presence of ≥10 leukocytes/high power field in urine sediment or of positive urine culture with ≥10 5 colonies/ml in a patient who was presenting with at least two of the following: body temperature above 38 0 C; dysuria, increased urinary frequency or urgency; flank pain or lumbar pain at palpation; and ultrasound findings compatible with acute pyelonephritis (Pinson et al., 1997) . volunteers gave written informed consent before any material was taken.

For the IGRA test, venous blood was collected in four heparinized tubes, namely: one negative control; two tubes containing M. tuberculosis-specific antigens PBMCs were isolated from the whole blood of participants of the 300BCG study as described above. PBMCs were stimulated at counts of 5 x 10 6 /ml in 96-well plates by 5 µg/ml of heat-killed M. tuberculosis H37Rv, 5x10 6 cfu/ml heat-killed S.

aureus or 10 ng/ml of E. coli LPS. After 24 hours and 7 days of incubation at 37˚C, supernatants were collected and stored at -20˚C until analysis. Concentrations of IL-1β and IL-6 were measured in 24-hour supernatants and of IFNγ in 7-day supernatants by the enzyme immunosorbent assay described above.

Using whole blood of participants in the 300BCG study, blood cells were counted using a Coulter Ac-T Diff® cell counter (Beckman Coulter, Brea, USA).

Following red blood cell lysis, white blood cells were washed twice with PBS and re- Table S3 for an overview of all inflammatory markers that were analyzed.

The trial sample size was calculated assuming the median time to new infection would be 4 months in the placebo group and 7 months with BCG vaccination. To achieve so with 90% power at the 5% level of significance, 100 patients were allocated to each arm. Under these prerequisites, the study is sufficiently powered to prove that differences in the first time incidence of infection between the placebo and the BCG group of the order of 20% will be statistically significant.

Baseline qualitative data were presented as percentages and CIs and compared by the Fisher's exact test. Baseline quantitative data were presented as J o u r n a l P r e -p r o o f mean and standard deviation and compared by the Student' s "t-test" for variables that followed normal distribution; they were presented as mean and standard error and compared by the Mann-Whitney U test for variables that did not follow normal distribution.

The primary outcome was the time, in weeks, of the appearance of one first new infection, censored at twelve months after vaccination. Differences between the placebo and BCG vaccination groups were assessed with the HR of the Cox proportional hazards regression model with its 95% CIs. The corresponding p-values were also reported. The effects of other confounders, both at the univariate and the multivariate Cox model, were also assessed with the corresponding HR. Only variables found to be significant in the univariate analysis entered in the stepwise multivariate analysis and they were retained in the model only if they had a significant effect after adjusting for the other effects. The proportionality of the hazard function at different levels throughout the follow-up period was assessed with the Schoenfeld residuals method (Xue et. al., 2013) . Since this was an interim analysis, a sensitivity analysis was also performed for the total number of participants with the primary aim to show that individuals that are censored have the same probability of experiencing a subsequent event as individuals that remain in the study. The number of infections in each group was expressed as patient-infections per year. The same analysis was done for the secondary endpoints. The frequency of adverse events was compared by the Fisher's exact test.

The interim analysis included only patients with completed 12-month of followup. In order to preserve the overall Type I error rate at 5%, an adjustment of the level of significance of the interim and final analyses was done by O'Brien-Fleming strict alpha adjustment. This adjustment provides significance a=0.0054 at interim and J o u r n a l P r e -p r o o f a=0.0492 at final (deMets, & Gordon Lan, 1994) . The purpose of using this seemingly unattainable level of significance at interim analysis was to allow the study to conclude, at the same time providing evidence that the required level of significance will be attained at the final stage. Statistical analysis was performed with the IBM SPSS Statistics for Windows, version 25 (IBM Corp., Armonk, N.Y., USA) and corroborated with the R statistical package (R Core Team, 2013) (see supplementary Statistical Analysis plan).

Cytokine data were expressed for each group of vaccination as the ratio of the cytokine production at month 3 versus the production at month 0 (before vaccination). Receiver operator characteristic curve analysis was done to discriminate the ratio of each cytokine that can better differentiate the two groups of vaccination. The best cut-off of this ratio was selected by the co-ordinate points of the curve using the Youden index. Patients above and below this cut-off were compared between groups by the Fisher's exact test.

Cytokine and protein data of participants in the 300BCG study were expressed as means ± SE. Data were analyzed using the Wilcoxon's rank sum test.

Analysis was conducted using IBM SPSS Statistics v. 25.0. All p values were two-sided and any p value <0.05 was considered as statistically significant.

Ongoing interventional clinical trials studying the protective effect of BCG vaccination on Covid-19 were reviewed using the Clinicaltrials.gov registry of trials (see supplementary Table S4 ).

The study is registered at ClinicalTrials.gov NCT03296423 J o u r n a l P r e -p r o o f

Role of tumor necrosis factor gene single nucleotide polymorphisms in the natural course of 2009 influenza A H1N1 virus infection

BCG cytokines associated with trained immunity

Homogenous 96-plex PEA immunoassay exhibiting high sensitivity, specificity, and excellent scalability

Infectious disease risk in the elderly

The international sepsis forum consensus conference on definitions of infection in the intensive care unit

Procalcitonin guidance of antibiotic therapy in community-acquired pneumonia. A randomized trial

Interim analysis: the alpha spending function approach

BCG scar and positive tuberculin reaction associated with reduced child mortality in West Africa. A non-specific beneficial effect of BCG?

Clinical characteristics of coronavirus disease 2019 in China

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Management of adults with hospitalacquired and ventilator-associated pneumonia: 2016 clinical practice guidelines by the Infectious Diseases Society of America and the

Bacille Calmette-Guerin induces NOD2-dependent nonspecific protection from reinfection via epigenetic reprogramming of monocytes

BCG vaccination enhances the immunogenicity of subsequent influenza vaccination in healthy volunteers: a randomized, placebo-controlled pilot study

Nonspecific effects of BCG vaccine on viral infections

Prevention of M. tuberculosis infection with H4:IC31 vaccine or BCG revaccination

Trained immunity: A program of innate immune memory in health and disease

Trained immunity: a tool for reducing susceptibility and severity of SARS-CoV-2 infection

Prevention of elderly pneumonia by pneumococcal, influenza and BCG vaccinations

Fever in the clinical diagnosis of acute pyelonephritis

R: A language and environment for statistical computing. R Foundation for Statistical Computing

Current and future trends in antibiotic therapy of acute bacterial skin and skin-structure infections

The Third International Consensus Definitions for Sepsis and Septic Shock

Acute lower respiratory tract infections and respiratory syncytial virus in infants in Guinea-Bissau: a beneficial effect of BCG vaccination for girls community based case-control study

Outcomes of controlled human malaria infection after BCG vaccination

The efficacy of Bacillus Calmette-Guerin vaccinations for the prevention of acute upper respiratory tract infection in the elderly

Heterologous immunity between viruses

Testing the proportional hazards assumption in case-cohort analysis